WO2004015333A2

WO2004015333A2 - Combustion control system with virtual lambda sensor

Info

Publication number: WO2004015333A2
Application number: PCT/IB2003/003321
Authority: WO
Inventors: Marcello Venanzoni; Lorenzo Marra; Guido Dubielzig
Original assignee: Merloni Termosanitari SpA
Current assignee: Merloni Termosanitari SpA
Priority date: 2002-08-05
Filing date: 2003-07-25
Publication date: 2004-02-19
Anticipated expiration: 2005-02-05
Also published as: ITAN20020038A0; CN100570218C; ITAN20020038A1; WO2004015333A3; EP1527303A2; AU2003247123A1; CN1675503A

Abstract

This patent describes a combustion control method for gas combustion units with automatic pre-mixing of the air/gas mixture. Said control method implies calculating the estimated air excess value λ, by using at least two physical magnitudes of the combustion status (J, θ; J1, J2; θl, θ2) close to the flame point during standard operation in specific positions (p.l, p.2); wherein said magnitudes are ionization currents (J) and flame temperatures (θ). The method also includes, a) the processing of a mathematical method suited to evaluate the excess air value λ, b) the determination of the operating conditions at values λ that represent optimal, briefly tolerable and intolerable parameters, regardless of the values of the mathematical model, c) the regular recalibration of the mathematical model The patent also describes the means to implement said method.

Description

COMBUSTION CONTROL SYSTEM WITH VIRTUAL λ SENSOR

DESCRIPTION Introduction

This invention refers to gas combustion units with automatic pre-mix of an air/gas mixture. The text that follows often quotes the term "boiler" to refer to both gas boilers and water heaters, specifically "instantaneous" models, in which combustion processes are exactly the same as the ones employed in gas boilers.

The adjustment of the air/gas ratio, usually referred to as λ, is a very important characteristic because it permits the definition of the operating ranges characterized by low polluant emissions and high yield. State of the art The current technology of gas boilers, based on "pneumatic pre-mixing" consists in producing the combustion of an air/gas mixture and in conveying it as a whole to the burners, as opposed to using a combustion chamber as occurs in atmospheric systems. A fan supplies the required flow rate of comburent air, generating, through a series of load losses, a pressure differential signal (Δp) transmitted to the gas valve. The latter is an "air/gas" valve, which means that the gas flow rate varies according to the signal Δp. The "gas flow rate" ratio = f(Δp) is mechanically determined by means of a nozzle that establishes this ratio for each operating condition. In other words, this pneumatic system determines the volume ratio between the air and gas flow rates. The excess of air in combustion is referred to as the ratio between the "relative air mass" used for combustion and the "relative air mass" required for the stoichiometric combustion, where the "relative air mass" represents the ratio between the air mass and the unit mass of fuel. The ratio between the excess of air λ and the volume ratio of the air and gas flow rates varies according to the fluids involved, which vary according to the varying temperatures of air and gas.

The so-called "mechanical" ratio between the volume flow rate of air and gas is set to a value capable of yielding a specific λ value. In these conditions, the comburent air contributes to the mixture with a temperature that is generally higher than that under cold start conditions; at start-up, the combustion occurs with a high excess of air, especially if the ambient air temperature is low (which determines a high fluid density) and can therefore be difficult and noisy. The supply of fuel gas to utilities is guaranteed by the gas supply system operating within a specific chemical composition range and by the universally known "Wobbe index". The mechanical adjustment of the ratio of the air/gas volume flow rate with pneumatic pre- mixing is not able to adapt to these variations because it is calibrated to guarantee an optimal performance in average conditions. The determined mechanical ratio does not permit arbitrary change in λ that would instead be useful, if we consider that the optimal lambda value of the thermal flow rate modulation (in terms of noxious emissions and yield) may not be constant.

The air and gas flows of combustion systems with "electronic pre-mixing" are controlled individually to guarantee a greater flexibility as compared to the pneumatic systems described above. For cold starts, for example, it is possible to reduce the volume air flow rate in order to attain a more effective combustion or at least make sure that combustion occurs at a configurable λ and basically more suitable for the situation. A variation in the incoming gas determination produces a variation of λ, which can be at least approximately corrected by adjusting the air and/or gas flow rates. The independent control of the fan and gas valve should permit the desired air excess to be attained within an interval of sufficient amplitude in the whole thermal power modulation range.

In systems with electronic pre-mixing, the fan is no longer expected to generate a pressure differential Δp for the gas valve, thus enabling the selection of a more economic model with a lower head. This solution also reduces the cost of the gas valve, which does not require a complex pneumatic regulation and compensation system and has a far simpler configuration. To coordinate the fan speed and the gas valve opening commands in a simple manner and to control combustion it is obviously necessary to control and measure, to a certain extent, the excess of air λ. As this value is not immediately accessible, it is necessary to estimate it approximately using indirect methods.

It is known that the gas molecules that participate in combustion undergo a ionization process that makes the gases electrically conductive; the actual amount of ionization changes according to combustion conditions and more specifically to the excess of air λ. If an electrode is placed close to the flame and an electric differential is applied between the electrode and a metal part, also exposed to the flame, the ionized gases are crossed by a current, known as ionization, which shall be identified with symbol J. The qualitative variation of this ionization current in function of λ has been known for several years. The measurement of value J has been used to estimate the value of λ, though the exact amount can be measured directly only using laboratory instruments.

Some of the combustion control systems developed so far control and adjust the value of λ using the measurement and adjustment of the ionization current value, assuming that this represents the excess of air λ. This implies setting a reference value for current J that represents the reference value for λ. Further information is provided in patents: GB2001426 priority of 25 7 1977; GB2185810 priority of 28 1 1986; DE3937290, priority of 10.11.1988; EP0617234 priority of 24.3.1993; DE4433425 priority of 20.9.1994; US5924859 priority of 25.10.1995; DE19502901 priority of 31.1.1995; DE19502905 priority of 31.1.1995; DE19618573 priority of 9.5.1996; EP0821198 priority of 25.7.1996; DEI 9631821 priority of 7.8.1996; EP0806610 that bases its priority on three documents of 1996; DE19726169 priority of 20.6.1997; EP0909922 priority of 17.10.1997; EP0962703 priority of 2.6.1998; DE19828111 priority of 24.6.1998; DE19835790 priority of 7.8.1998; DE3937290 priority of 10.11.1988; EP1002997 priority of 20.11.1998; DE19958340 priority of 3.12.1999; DE10003819 priority of 28.1.2000.

The most important detail is that J can acquire a maximum value of J.s for an almost exact stoichiometric air/gas ratio, i.e. for λ « 1. Therefore values of J below the maximum value of J.s can indicate the lack (λ < 1) or excess of air (λ > 1) in combustion; there are however known methods that use the measurement of J < J.s to determine whether the amount of air is in excess or default; see, for example, EP0806610 quoted above for further information. The relation between the ionization current J, measured close to the flame, and the excess of air is far more complex as it is significantly influenced by several factors, and specifically by:

- the nature of the device that supplies the electrode and measures current J; - the position of the electrode within the chamber;

- the physical dimensions of the electrode and its degree of wear, oxidation and dirtiness;

- the type of gas, i.e. the Wobbe index value and its chemical composition;

- the heating power generated; - the quality of comburent air;

- the actual flow rate of air that is also influenced by greater or minor obstructions in the stack and in the air suction duct.

The first two factors are set by the manufacturer according to the type or model of boiler by means of an initial calibration, while those remaining undergo variations that cannot be anticipated or that can be measured only in part. Although existing systems compare the difference (calibration) between the ionization value measured with the unit excess of air along with other additional information, for example related to the air flow rate (N. EP 1002997), the combustion control process based on the measurement of the ionization current in a specific position can be critical because there is no univocal correspondence between the excess of air and value J, regardless of where it is measured. The varying operating conditions (quality of comburent air, variations of Wobbe's index, partial obstruction of the stack, reduction of the gas supply pressure) as compared to the rated example influence the trend of the ionization current J. Therefore, this value, when taken as reference for the adjustment of the air in excess, can determine a shift of the working point, caused by the automatic regulation system, towards a value of λ that significantly differs from the that required, as qualitatively described in Figure 3, which will be illustrated in detail. Scope of the invention The main scope of this invention is to illustrate methods and means that allow a more accurate measurement of the actual value of the instantaneous excess of air λ as compared to that offered by traditional methods. A further scope of this invention is to illustrate methods and means to verify, with a higher degree of precision in comparison to known techniques, that the operation of the boiler is within a range that guarantees low and acceptable emissions of noxious gases and/or safety conditions and/or maximum efficiency.

A further scope of this invention is to describe a method suitable to define more accurate mathematical relations, with a varying degree of precision, among easily measurable physical magnitudes using the instruments provided with the boiler and the value of the air in excess λ.

A further scope of this invention is to offer a more accurate modulation of the boiler power. A further scope of this invention is to extend the power modulation range so that it can be used in safe and efficient combustion conditions.

Means and methods of the invention

These and other scopes are attained using the method described in the invention and the related means that provide an indirect control of the air in excess λ, which are based on a direct control of at least two physical magnitudes regarded as representative of the combustion status close to flame. These are represented by at least two ionization currents

J, as described above, which are measured at different distances from the burner or by a ionization current J and a flame temperature θ.

The method and means of the invention shall be described in further detail in this document, which illustrates some of the typical configurations of the invention, and in the drawings and enclosed claims, which form integral part of this document.

The suggested method therefore consists in estimating the excess of air λ, starting from at least two flame signals; the reference variable does not refer only to the ionization current

J measured so far, but also the estimate of λ. Therefore, this parameter is the direct object of the automatic regulation.

It has been determined in fact that the use of more than one flame signal, measured in each of the points that will be described below, provides much more accurate information on the actual status of combustion, thus enabling to mediate the effects of the deviations and variations of the factors involved, including those that are unpredictable. This results in a more limited deviation between the actual value of λ and that estimated in this invention in a wide range of operating conditions different from rated conditions A useful variant of the invention also controls and guarantees that combustion remains within the range that corresponds to standard operating conditions, thus preventing the unit from reaching values that are typical of the measured signals and/or their combinations, which indicate deviations to areas characterized by an exorbitantly high or low excesses of air (that could be potentially dangerous due to the high content of unburnt material) in a muchshorter time compared to the time ranges available according to known techniques. Description of the invention

The paragraphs that follow describe the invention making reference to some of the preferred variants and the drawings listed below. Fig. 1 represents a possible regulation layout for a combustion unit with pre-mixing, according to the invention, comprising a CONTROL SYSTEM. As the elements illustrated are purely symbolical, the dimensions and positions must be regarded as indicative. Fig. 2 represents the main logical blocks of the CONTROL SYSTEM shown in Figure 1 , according to the invention, along with a possible flow diagram of the signals exchanged among these logical blocks and the other components of the combustion unit.

Fig. 3 shows a diagram λ - J qualitatively illustrating the variation of the curve that regulates the relation between the ionization current J and the excess of air λ during the change of one or more physical parameters representing the combustion change. Figure 4 shows a diagram similar to the previous one that qualitatively illustrates the curve that regulates the relation between the two ionization currents Jl and J2, measured in two different flame positions, and the excess of air λ, with equivalent physical combustion parameters.

Figure 5 shows a diagram similar to the previous one that qualitatively illustrates the trend of the difference ΔJ between these ionization currents Jl and J2 with equivalent physical combustion parameters.

Figure 6 shows a diagram similar to the previous one that qualitatively shows the trend of the RJ ratio between these ionization currents Jl and J2 with equivalent physical combustion parameters. Figure 7 shows a diagram λ - θ that qualitatively illustrates the curve that regulates the relation between the two flame temperatures θl and Θ2, measured close to two different flame positions, and the air in excess λ with equivalent physical combustion parameters. Figure 8 shows a diagram similar to the previous one that qualitatively illustrates the variations of the curve that regulates the relation between the flame temperature Θ2 and the excess of air λ in function of the thermal power generated by the boiler.

Figure 9 shows a diagram similar to the previous one that qualitatively illustrates the trend of the difference Δθ between the flame temperatures θl and Θ2 with equivalent physical combustion parameters.

Figure 10 shows a diagram similar to the previous one that qualitatively illustrates the variation of the curve that regulates the relation between this difference Δθ and the excess of air λ in function of the thermal power generated by the boiler. Figure 11 is a time diagram t - λ that shows the progressive regulation ramp that must be followed to gradually reach the preset air excess, starting from a generic value, determined by a device of the CONTROL SYSTEM called SUPERVISOR.

Figure 12 shows a diagram λ - J that compares the actual values of J that can be attained with λ using the estimated values λ using a possible linear mathematical model J = f (λ) described in the invention.

Figure 13 graphically represents the intermediate limits of the air excess values λ, within which the CONTROL SYSTEM, designed according to the invention, is able to operate, and the extreme values that, on the contrary are unacceptable.

Figure 14, which is based on the indications shown in Figure 1, shows a diagram that illustrates the critical flame temperature values to be stored, which are representative of unacceptable combustion conditions according to varying opening degrees of the gas valve.

Figure 15 shows a relation between the percentage opening of the gas valve and the flame temperature below which combustion produces unacceptable conditions. Figure 16, which is based on the diagram of Figure 5, shows the acceptable limits and the unacceptable combustion limits.

Figure 1 shows a fan 2 that blows air into the pre-mixing chamber 1.1 of the burner 1; the gas valve 3 conveys the gas towards the pre-mixing chamber 1.1 if safety valve 4 is open.

Alternatively, mixing can occur directly within the fan if the gas is conveyed to the opening of the air inlet of the fan. The air/gas mixture continues towards the boiler 1 where combustion occurs. In the configuration defined in the invention, the combustion chamber 1.2 contains one or more "flame sensors" that detect certain physical magnitudes typical of combustion. More specifically, these two or more sensors can consist in one or two electrodes sjl and sj2 that detect one or two ionization currents, respectively Jl and J2, and in one or two sensors θl and sθ2 that detect one or two flame temperatures, respectively θl and Θ2. Index 1 identifies the two sensors sjl and sθl that are situated close to the outlet openings of burner 1 (i.e. situated in position p.l), while index 2 identifies the two sensors sj2 and sθ2 that are situated at a greater distance from the outlet openings (i.e. situated in position p.2). Positions p.l and p.2 of these sensors are not shown in scale in Figure 1 ; in reality, all sensors are positioned so that the signal is adequately influenced by combustion conditions. Furthermore, sensors with index 2 are placed at a sufficient distance from the sensors with index 1 so that they can detect the ionization currents J2 and/or flame temperature Θ2 values that are substantially different from those detected by sensors with index 1. These values change significantly according to the point in which they are measured. By way of example, for combustion units employed in domestic heating boilers with methane gas, with a maximum power of about 30 kW, position p.l of sensors with index 1 is located at a distance of 2 - 3 mm from the gas outlet openings of burner 1, while position ρ.2 of sensors with index 2 is situated at a distance of about 40 mm. Figure 1 shows the heat exchanger 5 with the thermal carrying fluid flowing in the direction shown by arrows 6. The figure also shows a CONTROL SYSTEM 7, designed according to the invention, equipped with electronic storage and calculation functions. This CONTROL SYSTEM 7 is suited to acquire signals that represent at least some of the following physical magnitudes, along with the following variants: the actual delivery temperature T.out and the return temperature T.in of the thermal carrying fluid, respectively measured downstream and upstream from the heat exchanger 5; the delivery temperature required in a specific instant by the boiler, the temperature, i.e. the set point T.out.sp; the signals that represent the ionization currents Jl and J2; the signals that represent the flame temperatures Θl and Θ2; a GV signal that represents the degree of opening of gas valve 3; and finally a signal that represents the realspeed measurement, RPM.m, of fan 2; according to the variants of CONTROL SYSTEM 7 designed in accordance with the invention, with varying degrees of accuracy, as described below, some of these input signals may not be present. CONTROL SYSTEM 7 processes these input and outputs signals at least for the following parameters: an RPM.sp signal used to adjust the speed of the fan to the value requested at a specific instant; a GV.sp signal used to adjust the opening of gas valve 3 to the value requested at a specific instant; an SGV signal used to allow to maintain in open position safety valve 4. These signals T.out, T.in, GV and RPM.m constitute the status signals of the combustion unit.

Thus, CONTROL SYSTEM 7 acts on the "adjustable" components of the boiler, i.e. fan 2, gas valve 3 and safety valve 4 to allow:

- the thermal carrying fluid to be supplied at the requested temperature, T.out.sp, through the opening of valve 3 at the degree required to make sure that the supplied thermal power is suitable to satisfy the condition T.out = T.outsp;

- a combustion with optimal air excess λ.sp, by adjusting the flow rate through the regulation of the speed of fan 2 and, if necessary, also through the regulation of the degree of opening of gas valve 3; - the continuation of combustion by maintaining safety valve 4 open, if the air excess value λ is maintained within acceptable values. As explained above, the excess air λ cannot however be directly measured inside the boiler with the instruments that can normally be fitted; its value is therefore determined as output of a function block that acquires as input one or more of the following parameters: - the signals that represent the ionization currents Jl and J2, measured in the combustion chamber;

- the signals that represent temperatures θl and Θ2, measured in the combustion chamber;

- the measurement of the fan speed RPM.m; - the GV value that represents the degree of opening of gas valve 3.

Figure 2 shows a layout of CONTROL SYSTEM 7.

This comprises the following main components.

The "virtual sensor" of λ" indicated with 8 (hereinafter referred to as VIRTUAL SENSOR

8) that is used to supply a sufficiently accurate estimate λ.c of the air in excess λ, according to two or more input parameters representing the combustion status.

SUPERVISOR 9 is used to evaluate certain conditions that delimit the efficient, safe and hygienic combustion area.

SUPERVISOR 9 receives as input the physical magnitudes that represent the operating conditions of the boiler and outputs some reference signals. SUPERVISOR 9 continuously sets and signals the value of λ considered optimal for each situation, hereinafter referred to as λ.sp. SUPERVISOR 9 is also able to regularly control the combustion conditions, in the modes and for the purposes described below, by acting directly on the gas valve 3 opening modulation, on the speed of fan 2 and on the opening acknowledgement sent to safety valve 4. CONTROLLER 10 continuously controls the combustion conditions (thermal power generated and excess of air λ) as soon as it receives as input the deviation between the actual delivery temperature T.out, the set point temperature T.outsp and the deviation between the excess air value λ.c calculated by the VIRTUAL SENSOR 8 and the optimal air in excess λ.sp (equivalent, for example, to 1.3) transmitted by SUPERVISOR 9. The CONTROLLER 10 outputs a signal A, for example through a P.I.D. signal processor PID- A l l, which receives directly even signal RPM.m, increasing the speed of the fan 2 to the RPM.sp value required to attain a correct λ. CONTROLLER 10 then sends directly to gas valve 3 a signal G that is used to open the same gas valve 3 with the value GV.sp required to attain the delivery temperature T.outsp set at a specific time. SUPERVISOR 9 can regularly directly force, with signals A' and G', the RPM.sp and/or GV.sp values by switching change-over switches 12 and/or 13 by means of the application of switching signals 12.1 and 13.1.

VIRTUAL SENSOR 8 receives as input two or more of the above-described signals: GV, Jl, J2, Θl, Θ2, RPM.m and outputs a signal that represents the calculated value λ.c. In addition VIRTUAL SENSOR 8 can receive from SUPERVISOR 9 a corrective value λ._os and a command that "stores typical values" m.v.c, with the modes and for the purposes that are described below.

SUPERVISOR 9 receives as input the same two or more signals received by VIRTUAL SENSOR 8, along with value λ.c, commands A, G and RPM.sp and value T.in, when required. The signals output from SUPERVISOR 9 include those described above along with the SGV signal that determines whether safety valve 4 should be maintained open or not. It is useful to notice that the value of T.outsp, required to establish the instantaneous thermal power required by the utility, is generated externally by CONTROL SYSTEM 7, with known methods and means, thus excluding the processing modes described in this document. This does not, however, forbid these methods from being implemented in CONTROL SYSTEM 7 or from incorporating these means in the system. Information provided bv the measurement of the examined magnitudes Before describing in detail the method illustrated in the invention, it is appropriate to examine how the excess of air λ of some typical parameters of combustion, like the ionization currents J or the flame temperatures θ, are influenced or substantially independent from other parameters like the positions in which these magnitudes are measured, the thermal power supplied, the thermal and chemical characteristics of the burnt gas, etc.

Appropriate graphs are shown in figures from 3 to 10 for this purpose. These show, by way of example, also the numerical values measured experimentally on a boiler of the type indicated, i.e. a methane gas boiler, with a maximum power of about 30 kW and with positions p.1 and p.2 respectively situated at a distance of about 3 and 40 mm from the burner gas outlet orifices. It is obvious that positions p.l and p.2 must be selected experimentally for each model of combustion unit in order to attain significant measurements of J and/or θ, according to the requirements of this invention. It is known, as described above, that the trend of the ionization current J increases for increasing air excess values λ < 1 , becomes equivalent to the maximum value with λ = 1 and decreases once more when λ increases and is > 1. Optimal combustion conditions occur when the value of λ.o is >1 (the preferred value is λ.o = 1.3). There is also a range within which combustion can be defined as normal, although the regulation system should be activated in order to substantially reset combustion to a value of λ.o. This range is delimited by the values here defined as λ.min.n and λ.max.n, which are obviously both below >1. It is also necessary to define minimum and maximum acceptance limits λ.min.e and λ.max.e, which when exceeded will cause an irreperable operating fault and the immediate closure of safety valve 4. By way of example and for the purposes of this invention, it is possible to provide the following progression values: λ.min.e « 1.15 < λ.min.n <λ.o « 1.3 < λ.max.n < λ.max.e « 1.45.

As all these values λ are clearly below > 1; the graphs of these figures qualitatively show the curve that regulates the relation between J and λ for the involved area only (indicated on the X axis by the diagrams identified as "λ > 1") In this interval, or at least within the range of ordinary combustion, the value of a ionization current Jl, measured in an area sufficiently close to burner 1, has, as known, a substantially linear trend and can therefore be adequately approximated with a linear mathematical model.

The relation between J and λ is not, as known, unique in the range λ > 1. Figure 3 shows how the curve that regulates the relation between J and λ changes in function of the above- mentioned parameters, i.e. at least in function of the following parameters:

- nature of the supply and measurement loop;

- physical dimensions of the electrode and related status of wear, oxidation and dirt;

- position of the electrode within the chamber and nature of flame; - type of combustion gas and value of Wobbe's index;

- power burnt; quality of comburent air. This explains why an optimal value of J.sp may not correspond to the desired λ.sp that will obviously be set to λ.o, which is considered an optimal value. Yet, the control of at least two physical parameters typical of combustion, measured in appropriate positions, provides much more information on the actual conditions of combustion, and permitting a more accurate calculation of the actual value of λ and controlling that combustion does not exceed the limit acceptance values λ.min.e e λ.max.e Measurement of the two ionization currents Jl andJ2 (see Figure 4) If the remaining conditions are equivalent, the measurement of the two ioniziation currents Jl and J2, close to positions p.l and p.2, provides trends and values that can be qualitatively different if the value of the air in excess changes: the resulting trends of currents Jl and J2 are similar to those shown in Figure 4. Figure 4 also shows possible numerical values: close to λ = 1.3, the values of Jl and J2 are significantly different (respectively 17 and 6 μA), but tend to have the same value close to λ = 1.45. These trends provide information on the status of the flame. It is useful to notice that while the trend of Jl is substantially linear as compared to λ, the trend of J2 is substantially hyperbolic. It is known that the position in which the flame stabilizes varies in function of the value of the air in excess, due to the combination between the flame propagation speed and the output speed of the air/gas mixture. The flame tends to be close to the burner when the λ values are low and more distant from the burner when the values are higher. This means that at a given power, the intensity of ionization current measured in function of the flame, Jl, increases when the flame is close to the burner and lower when the flame is more distant. Current J2 is instead measured at a distance from the burner that is only partially influenced by the flame and tends to lower and intersect the other trend, as soon as the value of λ increases, that is when the distance of the flame from the burner is increased. The value of the two currents changes also according to the thermal capacity, although the amount of the deviation does not allow a sufficiently accurate evaluation of power. Difference ΔJ between the ionization currents Jl - J2 (see Figure 5)

If the remaining conditions are stable ΔJ = Jl - J2 presents a trend of λ, as shown in Figure

5.

ΔJ acquires the maximum value close to the maximum vertical distance between the curves of Jl and J2 and returns to zero in the point in which these two currents have the same value. Experimentation has discovered that, although the relation between curves Jl and J2 and λ is influenced by several factors, including unpredictable ones, the maximum and zero values of ΔJ are related to values of λ that are substantially not influenced by these factors and that are respectively equivalent to about 1.15 and 1.45, for the reference boiler. These two points can constitute the limit values λ.min.e and λ.max.e that should never be exceeded during ordinary operating conditions. The maximum value of ΔJ is reached when the operating range shifts towards excess air values that are considered too low, while value zero indicates that the combustion is occurring at values of λ that are definitely too high. It is also useful to notice that the two values of λ that correspond to the maximum value and to zero, can also be used, as explained below, to regularly calibrate VIRTUAL SENSOR 8, both on the input side by dividing all the magnitudes J and θ used by VIRTUAL SENSOR 8 by the value that these acquire in the maximum and zero points of the Jl - J2 difference, and on the output side, by simply adding to the estimated value of λ a value λ.os equivalent to the difference between the value of λ, which are measured in maximum and zero pointsof Jl - J2, and the value which should result according to VIRTUAL SENSOR 8. Ratio RJ between the two ioniziation currents Jl / J2 (see Figure 6)

Examining the trends of currents Jl and J2, shown in Figure 4, one notices that the ratio RJ =J1 / J2 is reached in the hygienic combustion range, i.e. in the range between the maximum and zero value of ΔJ shown in Figure. 5. Even for ratio RJ, experimentations have determined that the value of λ reaches its maximum independently from Wobbe's indexes, at least for the values that range from the minimum and maximum values that suppliers of fuel gas can use for gases of Class II (i.e. natural gas within the Ws range from 54.76 to 40.90 MJ/Nm³ at 15°C and 1013 mbar (dry gases), corresponding to gases of Group E, in accordance with EN 437) and from the chemical composition of the gas, provided that the gas valve reaches a specific degree of opening. The same occurs for gases of Class III (i.e. liquid gas with Wobbe's index Ws within the range from 72.86 to 87.33 MJ Nm3 at 15°C and 1013 mbar (dry gas)). Furthermore, this maximum value is reached with a value of λ very close to 1.3, i.e. at an optimal λ.o combustion value, or at any rate within the range of hygienic combustion. However, if the value of λ corresponding to the position of this maximum is known, it is possible to use this value, as explained below, to regularly calibrate VIRTUAL SENSOR 8, both on the input side by dividing all the magnitudes J and θ used by VIRTUAL SENSOR 8 by the value they acquire in the maximum point of J1/J2, and on the output side, by simply adding to the estimated value of λ a value of λ.₀s equivalent to the difference between the value λ in which RJ actually reaches its maximum value and the value in which it should theoretically reach the maximum value according to VIRTUAL SENSOR 8.

Thus, by using the measured values of the ionization current J in two different positions p.l and p.2, appropriately selected, and by adequately processing them, or rather specifically calculating the difference ΔJ and the ratio RJ between Jl, J2, it is possible to have sufficient elements to control the combustion more accurately and more efficiently as compared to previous methods.

Temperatures inside the combustion chamber (see Figure 7) By measuring the temperatures θl and Θ2, as defined above, in equivalent conditions, the related trends for λ, can be obtained as shown in Figure 7.

Figure Θ2 shows that the decreasing trend is substantially linear as λ increases (i.e. when the amount of air used to dilute combustion products increases), while θl has a curved decreasing trend due to the disabling of the flame, in presence of high air excess values, as this contributes to reduce the temperature close to the burner.

Temperature θl, which is therefore more sensitive to the excess of air and to preset low temperatures, can be used to indicate that the operating point is moving to areas with a high λ. Information on temperature Θ2 (see Figure 8)

It is useful to notice that the curve influenced by temperature Θ2 da λ, maintains its linearity but varies significantly according to the thermal power generated, as expected. Therefore, the trend of temperature Θ2 can provide, at a given gas valve opening or gas flow rate and in presence of a specific fuel, an immediate indication of value λ. Vice versa, if value λ is known, temperature Θ2 can be used to obtain values on the thermal capacity, though this is significantly .influenced by the type of gas. Difference Δθ between the temperatures of chambers Θ2 - Θl (see Figure 9) If other conditions and specifically the thermal power generated are equivalent, value Δθ = θl - Θ2 is represented by an increasing curve, whose slope significantly increases as soon as the zero is crossed. A preset maximum difference value Δθ can be used as limit signal before the operating areas with an excessively high λ. Influence ofΔθby the supplied power (see Figure 10)

It has been experimentally determined that value Δθ substantially remains unaltered with all gases, while the curve related to Δθλ, maintains a stable trend but significantly changes according to the supplied power. Therefore, if value λ is known, value Δθ can be used to estimate the thermal power value; vice versa, if the thermal power is estimated, value Δθ yields the value of λ.

A close examination of the diagrams of Figures from 4 to 10 shows that the measurement of one or more magnitudes typical of combustion, and consisting in the measurement of the ionization currents J and/or of the flame temperatures θ in one or more positions p.l or p.2 (or in all the other positions) provides information on the combustion status that a measurement, based on known methods, of a single ionization current J cannot provide. It is evident that this information does not only provide information on the limit values of λ, which should not be exceeded, but also permits the formulation of mathematical models suitable to calculate value λ more accurately as compared to known models.

Said mathematical models may also take into account, as the sections below will explain, further typical operating parameters, which can be easily measured on the boiler or have already been measured for other purposes. We have already seen that there are several sets of parameters that can be used as input values for these mathematical models, just as there are several mathematical models that can be employed (linear or non linear), which cannot be listed in full but that are included within the methods used to estimate the value of λ according to the invention. Description of the virtual sensor This section describes in detail VIRTUAL SENSOR 8. This employs one or more mathematical models that define the law used to discipline the relation between value λ and input parameters. The mathematical relation between inputs and outputs can be determined, for example, using the known model identification theory (see: Sergio Bittanti, "Identificazione dei Modelli e Controllo Adattativo", Pitagora Ed., 1997). To determine the relation between the inputs and the output of the model contained in VIRTUAL SENSOR 8, it is possible to follow the procedure described below.

1) it is necessary to design the trends for A and G (see Figure 2) that enable the boiler to work in the conditions resulting from the Cartesian product of the following ranges:

- all the adjustment range of the thermal capacity involved ;

- a range of λ that extends over wider limits as compared to acceptable ones; for example from 1.0 to 1.5;

- the emissions of CO are on average always within 1000 ppm;

- the application uses both the reference gas and the two limit gases of the family in question.

A model designed according to these requirements is able to estimate, with an accuracy that varies according to the models, the value of λ within the indicated operating areas, thus automatically compensating the thermal flow rate variations and the type of fuel gas. The number of ionization currents (one or two) that will be measured and used by VIRTUAL SENSOR 8 determine the type of procedure that can be used to identify the model. If a single current J is used, it is necessary to perform procedure 2.a); while when two currents Jl and J2 are used, it is possible to choose any of the following procedures: 2.a), 2.b), 2.c), 2.d). Procedure 2.a) The operator:

- uses the reference gas of the desired class (i.e. G20) to operate the test boiler with a specific valve opening degree, which should be equivalent to the one that is regarded more significant for the measurement of the values that follow, and adjusts the value of

A so that λ remains > 1 ;

- reduces the speed of the fan in order to determine the maximum value of the only ionization current J or of the ionization current chosen between Jl or J2 (that occur at a λ.s point equivalent to about 1), storing with the test system the following values: 1. current J.s or ionization currents Jl .s and J2.s;

2. flame temperature θ.s or the two flame temperatures θl.s, Θ2.s (where "s" stands

- operates the boiler within the time ranges A and G, designed as described above, measuring and storing the λ of the boiler, one or more ionization current values J, Jl, J2; one or more flame values θ, θl, Θ2; the speed of the fan RPM.m and the degree of opening GV of the gas valve. Procedure 2.b) The operator:

- uses the reference gas of the desired class (i.e. G20) to operate the test boiler with a specific valve opening degree, which should be equivalent to the one regarded as the most significant for the measurement of the values that follow, and adjusts the value of A so that λ remains > 1 ;

- changes the value of command A in order to reach the maximum value of the RJ ratio between the measurement of current Jl and of current J2 (that occurs when λ.b is within the hygienic combustion range), and allows the test system to store the values of: 1. the two ionization currents Jl .b and J2.b;

2. the temperature θ.b or the two flame temperatures θl.b and Θ2.b, when measured (where "b" indicates a "good combustion");

— operates the boiler within the time ranges A and G, designed as described above, measuring and storing the λ of the boiler, the values of the two ionization currents Jl and J2; the values of the flame temperatures Θ, Θl, Θ2; the speed of the fan RPM.m and the degree of opening GV of the gas valve. Procedure 2.c) The operator: - uses the reference gas of the desired class (i.e. G20) to operate the test boiler with a specific valve opening degree, which should be equivalent to the most significant for the measurement of the values that follow, and adjusts the value of A so that λ remains

> i;

— changes the value of command A in order to reach the maximum difference ΔJ = Jl - J2 between the measurement of current Jl and of J2 (that occurs when λ.m > 1), and allows the test system to store the values of:

1. the two ionization currents J 1.m and J2.m;

2. the temperature θ.m or the two flame temperatures Θ 1.m and Θ2.m, when measured (where "m" stands for "maximum"); — operates the boiler within the time ranges A and G, designed as described above, measuring and storing the λ of the boiler, the values of the two ionization currents Jl and J2; the values of the flame temperatures θ, Θl, Θ2; the speed of the fan RPM.m (where "m" stands for "measured") and the degree of opening GV of the gas valve. Procedure 2.d The operator:

— uses the reference gas of the desired class (i.e. G20) to operate the test boiler with a specific valve opening degree, which should be equivalent to the most significant for the measurement of the values that follow, and adjusts the value of A so that λ remains

> i; - changes the value of command A in order to reach the zero point of the ΔJ = Jl - J2 difference between the measurement of current Jl and of J2 (that occurs when λ.z > 1), and allows the test system to store the values of:

1. the two ionization currents J 1.z and J2.z;

2. the temperature θ.z or the two flame temperatures θl.z and Θ2.z, when measured (where "z" stands for "zero"); - operates the boiler within the time ranges A and G, designed as described above, measuring and storing the λ of the boiler, the values of the two ionization currents Jl and J2; the values of the flame temperatures θ, θl, Θ2; the speed of the fan RPM.m and the degree of opening GV of the gas valve.

At this point, it is possible to use known calculation methods to determine the relation between inputs and outputs that minimizes the average quadratic deviations between the sequences of the measured values of λ and those calculated by means of the relation. The set of inputs consists in a combination, as described below, of one or more of the standardized currents, of the measurements of one or more standardized chamber temperatures, of the measurement of the fan speed RPM.m and of the gas valve opening GV. The term "standardized" indicates that the measured values of currents J and of temperatures θ are respectively divided by values J.s (or Jl.s and J2.s), θ.s (or θl.s and Θ2.s) when procedure 2.a) is used, or respectively by values Ji.b, J2.b, θ.b (or θl.b and Θ2.b) when procedure 2.b) is used, or respectively by values Jl.m, J2.m, θ.m (or θl.m and Θ2.m) when procedure 2.c) is used or respectively by values Jl.z, J2.z, θ.z (or θl.z and Θ2.z) when procedure 2.d) is used. It is necessary to decide which model to use for the selected set of inputs (for example a model - relation that receives as input the two ionization currents Jl and J2 and the fan speed RPM.m, or a model that uses a current J and a temperature Θ, or a model that acquires as input the two currents Jl and J2 and a temperature θ, and so on). The resulting relation has to be then validated on a set of inputs/outputs different from those used for identification purposes.

It is possible to implement several mathematical models in the same VIRTUAL SENSOR 8 used for several models of boilers. During the boiler manufacturing phase, it is necessary to select the set of inputs of the VIRTUAL SENSOR 8 of λ, prepare the related sensors and electronically implement the model-relation that refers to the chosen set of inputs. The invention foresees the following input configurations: - two currents (Jl and J2); - one current and one temperature (J, θ);

- one current and two temperatures (J, θ 1 , Θ2);

- two currents and one temperature (Jl, J2, θ);

- two currents and two temperatures (Jl , J2, θ 1 , Θ2); - one of the above parameters in addition to the speed, RPM.m, of the fan 2 and/or the GV gas valve opening value 3. The VIRTUAL SENSOR 8 stores, depending on the configuration selected, the following default values:

- Jl.s, J2.s or Jl.b, J2.b or Jl.m, J2.m or Jl.z, J2.z; - J.s, θ.s;

- J.s, θl.s, Θ2.s;

- Jl.s, J2.s, θ.s or Jl.b, J2.b, θ.b or Jl.m, J2.m, θ.m or Jl.z, J2.z, θ.z;

- Jl.s, J2.s, θl.s, Θ2.s or Jl.b, J2.b, θl.b, Θ2.b or Jl.m, J2.m, θl.m, Θ2.m or Jl.z, J2.z, Θl.z, θ2.z. In standard/normal operating conditions, VIRTUAL SENSOR 8 uses the input data and standardizes them (as explained above) before estimating the value of λ. During the initial start-up, the standardization parameters are those with subscript "s" or "b" or "m" or "z", depending on configurations. These factors are regularly updated during the operation of the boiler in accordance with the procedure described below, in order to compensate the possible variations of the parameters that change in percentage the measured values of current and temperature, in equivalent conditions. Updated factors are stored in memory and used for standardization up until the new update occurs. The procedure also includes a calibration of the output of VIRTUAL SENSOR 8, which is carried out measuring the deviations between the estimated value λ.c and the value considered in some specific conditions.

Double calibration (proportional for inputs and additional for outputs) is particularly suited to the structure of the mathematical model and to account for the degradation of sensors. The wear, ageing and dirt of ionization sensors cause a decrease of the measured value that can be easily described as a percentage derating. The measured current may become, after a specific number of working cycles, equivalent to 90% of its initial value. Basically, the percentage (or proportional) effect is the one that best describes the decrease of the measured value within the whole range of measured values.

A close examination of the trend of the current Jl in function of λ will lead to a maximum value in point λ ~ 1 with a value of Jl.max and the generic value Jl with a given value of λ. If the current undergoes a percentage reduction in the whole range of λ, the new values will be equivalent to:

Jl.max' = 0.9 * Jl.max and J1' = 0.9 * J1 that results in 0.9, by equating the two equations:

Jl.max' / Jl.max = Jl ' / Jl and also Jl' / Jl.max' = J1 / Jl.max .

Thus, the ionization current values, which have been standardized as compared to maximum values, remain unaltered even if a percentage variation of the current J occurs. In this way the mathematical model remains independent from the percentage input variations. We have also determined that the standardized current curves overlap for the different gases (at least in the range of Wobbe's indexes referred to for the diagram of Figure 6) if the gas valve opening degree is stable. Thus, the trends of standardized currents also permit the compensation of the variations of the type of gas, to signal the limits for unacceptable combustion limits, and to make sure that these signals are substantially not influenced by the type of gas.

While the calibration of the inputs influences the products, the calibration of outputs influences the sums, as it adds an addendum for the correction of the estimated value of λ. If the mathematical model is simply parametric and linear, i.e. equivalent to "λ(t) = k.ji x Jl(t-1) + k._J2 x J2(t-1) + k._θι x θl(t-l) + k._θ2 x Θ2(t-1) + etc." (where "x" represents the multiplication signal; "t" is a time variable, "t-l" indicates the time instant that precedes "t" used to determine the values of J, θ , etc. and to calculate λ for instant t), the values of the inputs are multiplied by appropriate parameters and summed.

Although the relation between J2 e λ and θl and λ is clearly not linear, the parametrized linear mathematical models built as described above have offered an estimate of λ.c of the actual λ, which is satisfactory at least in the operating range examined (specifically for those that take into account the number of revolutions RPM.m). It would be possible however to build more accurate mathematical models (suitable to be managed by means of ordinary microprocessors despite the complexity), like "λ = k.jι(J2) x Jl + k.e₂(θl) x Θ2 + etc.", characterized by a linear relation between λ and Jl and Θ2, while the corresponding coefficients k.jι and k.ø₂ are determined according to the values of J2 and θl (or calculated or read by a memory table).

Alternatively, it is possible to develop a mathematical model that takes into account also or only a combination of the input variables like RJ o Δθ. Description of the supervisor

SUPERVISOR 9 receives as input at least the following signals: - the same signals acquired as input by VIRTUAL SENSOR 8;

- temperatures T.out and T.in, which respectively represent the delivery and return temperatures of the heating circuit;

- the RPM.sp signal set according to the speed of the fan;

- signals A and G processed by CONTROLLER 10; - the value λ.c calculated by VIRTUAL SENSOR 8.

Depending on the number and quality of input signals for each possible variant, SUPERVISOR 9 calculates, as described below, at least some of the relations referred to in paragraph "Information resulting from the measurement of the examined magnitudes^'". The quantity and type of signals used for the estimate of λ and, therefore, the type of relation-model used enable to attain the following information:

- one current J and one temperature θ indication of the high values of λ (see Figures 4, 7, 8)

- one current J and both temperatures θl and Θ2 as above, along with information on the thermal capacity (see Figures 7 and 10) - two currents Jl and J2 information on the limit values of the operating range of λ (J1-J2) and of the reference point for λ in the hygienic combustion area (J1/J2) (see Figures 5 and 6)

- two currents and one temperature as above (see Figures 5, 6 and 7) - two currents and both temperatures as above, along with information on thermal capacity (see Figures 5, 6 and 10). The SUPERVISOR 9 carries out appropriate commands on A and G only in certain situations, leaving the VIRTUAL SENSOR 8 and the CONTROLLER 10 to manage standard operations. Standard operating cycle As soon as the boiler is started, SUPERVISOR 9 performs a switching operation on the change-over switches 12 and/or 13 by means of the switching signals 12.1 and 13.1, replacing CONTROLLER 10, and sends signals A' and G' respectively to the fan and to the gas fan, thus generating a pair of values that is regarded as optimal for the injection and that enables to attain, after the start-up time of the boiler, a value λ above 1, which is regarded adequate to guarantee a safe and noiseless start-up with any Wobbe index included within the range indicated. If the boiler does not start after the first injection, the system will attempt a second start (after a specific interval of time, i.e. 10 s) and a third one, if necessary, following the known procedures and gradually reducing the fan speed.

As soon as the boiler turns on, the flame is allowed to stabilize so that it is able to maintain commands A' and G' constant for a specific interval of time (i.e. 15 s).

This operation is carried out by SUPERVISOR 9 as follows: the supervisor acquires as input the estimate of the actual λ.c from VIRTUAL SENSOR 8; it then processes, for λsp (see Figure 11) , a ramp with a length equivalent to a specific amount of time (i.e. 10 s) which ranges from current value λ.c to the final desired rated value of λsp (for example λ = 1.3). Thus, before the command is switched over to CONTROLLER 10, the estimated of λ.c calculated by VIRTUAL SENSOR 8 and the value λ_SP set by SUPERVISOR 9, which corresponds to the initial point of the ramp, coincide.

After the command is switched over to CONTROLLER 10, SUPERVISOR 9 sends increasing signals of λsp following the ramp up to the desired value (i.e. λsp = 1.3). The differences between the current values λ.c, of T.out, of λsp and of T.out.sp for the delivery temperature are transmitted to CONTROLLER 10 that adjusts commands A and G in order to rectify these differences. This occurs without sudden alterations of the speed of fan 2 and of the opening of gas valve 3, as the processed ramp gradually leads to the desired operating speed λsp, starting from the initial value of λsp, which is equivalent to the current value of λ.c.

Calibration of the virtual sensor by means of the sampling of characteristic data VIRTUAL SENSOR 8 is able to compensate the variations of the parameters (that influence the measurements of currents J and of temperatures Θ) by calibrating both the inputs and outputs. Sampling of characteristic data At preset intervals during the standard operation of the boiler, when the gas valve reaches the preset degree of opening that the manufacturer deems relevant for the measurement of the data that followed (the same used in the previously described identification procedures), SUPERVISOR 9 can assume the control of the fan 2 and of gas valve 3 by switching the change-over switches 12 and 13, as described above. The control starts by sending signals A' and G' identical to signals A and G that CONTROLLER 10 transmitted before the switching, generating λsp and value λ.c calculated, instant by instant, by VIRTUAL SENSOR 8 in order not to influence the combustion status. Soon after, while G' is maintained constant, SUPERVISOR 9 starts changing signal A' following four alternative modes. Mode a) is applied when one ionization current J is measured; modes a), b), c) or d) can be used when it is necessary to measure two currents Jl and J2. Mode a)

A c λ > 1 S

iAentϊfv t e τnnχ.tτπτm vpl"-* n-F+1-ι-a ioniziation current J or find the most significant value between the ionization currents Jl and J2. As soon as it identifies value A', which represents this condition, SUPERVISOR 9 uses the "typical data storage" command, m.v.c, to make sure that VIRTUAL SENSOR 8 records the current values of data J and θ received as input (which may include, as explained above, one or more of the following values: Jl, J2, θl, Θ2

Mode b)

SUPERVISOR 9 changes A' in order to make sure that the combustion conditions correspond to the maximum value of the RJ = J1/J2 ratio (see Figure 6). As soon as value

A', which represents this condition, is reached, SUPERVISOR 9 uses the "typical data storage" command, m.v.c, to make sure that VIRTUAL SENSOR 8 records the current values of data J and θ received as input.

Mode c) SUPERVISOR 9 changes A' in order to make sure that the combustion conditions correspond to the maximum value of the ΔJ = J1-J2 difference (see Figure 5). As soon as value A', which represents this condition, is reached, SUPERVISOR 9 uses the "typical data storage" command, m.v.c, to make sure that VIRTUAL SENSOR 8 records the current values of data J and θ received as input. Mode d SUPERVISOR 9 changes A' in order to make sure that the combustion conditions correspond to the zero of the ΔJ = J1-J2 difference (see Figure 5). As soon as value A', which represents this condition, is reached, SUPERVISOR 9 uses the "typical data storage" command, m.v.c, to make sure that VIRTUAL SENSOR 8 records the current values of data J and θ received as input. In one of the variants of the sampling modes described for the typical values, SUPERVISOR 9 first brings the gas valve 3 to the degree of opening that is regarded as most significant for the measurement of the described values, using an adequate value of signal G'. This option is useful to make sure that it is possible for the gas valve 3 to reach the optimal degree of opening in all sampling conditions. The sampling operation is so short (it lasts a few seconds) that it does not interfere with the regular operation of the boiler, even if the optimal opening corresponds to the maximum degree of opening. Calibration of inputs

The values stored as above represent the new factors used for the standardization of the measured values acquired as input by VIRTUAL SENSOR 8. Calibration of the output

All the sampling modes a), b), c) or d), permit a combustion condition to be reached in which the value of λ = λ.typical corresponds to the typical value, i.e. to value known to the manufacturer. The calibration of outputs consists in comparing the value λ.c calculated by VIRTUAL SENSOR 8 in specific conditions with the given value and in adding to λ.c a corrective offset value λ.os if the typical estimated error λ.c - λ.typical exceeds the tolerated value.

The procedure varies slightly according to the procedure employed for the sampling. If the sampling has been performed using Mode a), the system stores the value J.s of the ionization current that offers a typical value λJs equivalent to approximately 1 and known to the manufacturer. If a linear model has been used to estimate λ, the value λ.c output by VIRTUAL SENSOR 8 for J - J.s could significantly differ from the actual value (Figure 12). The amount of the deviation increases as the maximum value decreases; although this does not mean that the mathematical model is inaccurate in the combustion range that has a significant operating relevance. As the deviation between the actual curve J = f(λ) and that of the mathematical model J = f(λ.c) is known to the manufacturer, it is possible to: multiply the value J.s measured for an experimentally determined coefficient in order to obtain the correct value of J.s.corr (see Figure 12), close to which the mathematical model must generate an output value λ.c = λjs and, if the estimated error λ.c - λJs exceeds the tolerated value, SUPERVISOR 9 determines a correct offset value λ.os ⁼ λJs - λ.c. For the boilers described above, the acceptable maximum error value is 0.07. If the sampling is performed with Mode b , the system stores the ionization currents Jl.b and J2.b in the condition in which the ratio R.max = Jl / J2 corresponds to the maximum value. This always occurs with a value λ, equivalent to λ.Rma_x known to the manufacturer and very close to 1.3 in the optimal combustion area. In this area, the mathematical model, including the linear model, provides a sufficiently accurate approximation of the actual curve J = f(λ). Thus, the comparison between λ.c and λ._Rnιax can be performed directly without using the corrective factors for Jl and/or J2. As above, if the estimated error λ.c - λ. _max exceeds the tolerated value, SUPERVISOR 9 determines a corrective offset value λ.os = λ._Rmax - λ.c For the boilers described above, the acceptable maximum error value is 0.05. If the sampling is performed with Mode c), the ionization currents Jl.m and J2.m are stored in the condition in which the combustion condition corresponds to the maximum value of the Δ.max = J1-J2 difference. This always occurs with a lambda value equivalent to λ.Λma that is known to the manufacturer. Thus, the comparison between λ.c and λ._Λmax can be performed directly because even in this case the mathematical model, even when linear, provides an adequate and sufficiently accurate approximation of the actual curve J = f(λ). As explained above, if the estimated error λ.c - λ.Λmax exceeds the tolerated value, SUPERVISOR 9 determines an offset corrective value of λ.os = λ.Λma - λ.c. For the boilers described above, the acceptable maximum error value is 0.05. If the sampling is performed with Mode d , the system stores the ionization currents Jl.z and J2.z in the combustion condition where the difference Δ.z = J1-J2 is equivalent to zero. This always occurs with a lambda value of λ.Δ_Z that is known to the manufacturer. The comparison between λ.c and λ._Δz can be performed directly because even in this case, the mathematical model, even when linear, provides and adequate and sufficiently accurate approximation of the actual curve J = f(λ). As explained above, if the estimated error λ.c - λ._Δz exceeds the tolerated value, SUPERVISOR 9 determines a corrective offset value of λ.os = λ._Δz - λ.c. For the boilers described above, the acceptable error value is 0.05.

After configuring the VIRTUAL SENSOR for the "typical data storage", m.v.c, and determining the value of λ.os, SUPERVISOR 9 returns the combustion control to CONTROLLER 10. Therefore, SUPERVISOR 9 resets A' and G', when necessary, to the A and G values imposed on CONTROLLER 10 just before interruptionof the control functions in order to calculate, using the procedures described, a ramp whose length is determined by λ_Sp because, in the meantime , the value λ.c generated by VIRTUAL SENSOR 8 could have differed from from the optimal value, λ_Sp. In fact, VIRTUAL SENSOR 8 is already evaluating the new input values and dividing them by the updated standardized values and could therefore generate a new value of λ.c if the actual combustion conditions remain stable. The new ramp serves as guide for CONTROLLER 10 and is used to gradually reestablish optimal combustion conditions.

SUPERVISOR 9 instantly calculates also a ramp for signal λ₀s, that is sent to VIRTUAL SENSOR 8. The ramp that has a specific duration (i.e. 20 s) starts with the offset value used for the previous output calibration and reaches a value that is equivalent to the sum of the new and old offset values λ.₀s- This operation leads to the sum, instant by instant of the offset value and the value of λ, calculated using the mathematical models, which ensures that the output of VIRTUAL SENSOR 8 is always equivalent to the expected value of λ.c, which has already been corrected with the output calibration procedure. The offset value is stored in memory by SUPERVISOR 9, continuously applied to VIRTUAL SENSOR 8 after a ramp of λ₀s and changed only after the subsequent calibration of the output. Detection of abnormal operating conditions In standard operating conditions, the value of λ is maintained very close to the desired value (i.e. 1.3) thanks to the control system supported by VIRTUAL SENSOR 8, although there are some cases in which the system is not able to maintain the value of λ close to the setpoint λsp. To prevent the occurrence of events out of the combustion range and detect anomalies in the operation of gas valve and fan, it is possible to set two safety levels for value λ, in accordance with the invention. The set point of λ is positioned, for example, at λsp = 1.3. The first safety level (see Figure 13) is close (compatibly with the accuracy of VIRTUAL SENSOR 8 and CONTROLLER 10) to a condition of standard operation, ranging from λ.min.n <λ.o to « 1.3 < λ.max.n. If the system falls below or exceeds this range, it is possible to decide to tolerate the deviation for a limited interval of time, for example 5 sec, in order to attempt to reset value λ within the range. If this does not occur, SUPERVISOR 9 turns the boiler off generating an error code, if required, which can be accessed by the technical assistance by disabling the safety valve.

The second level of safety (see Figure 13) has operating values of λ and λ.min.e / λ.max.e external to those of the previous range and causes the immediate shutdown of the burner in the event of exceedance (with the generation of an error code, when required). These thresholds, set in accordance with the invention, are configured using the trends described in paragraph "Information provided by the measurement of the examined magnitudes". As explained at the beginning of the description of SUPERVISOR 9, each configuration of inputs with at least two flame sensors (sensors detecting current J or temperature θ) always has at least one threshold that signals excessive λ values. If the measurement of a single current J is used in the combustion chamber, then the presence of high λ values can be detected by using current J or temperature θ. In this case (see Figure 14) the system detects the minimum temperature value θ (or of one of the temperatures, preferably value θl measured at the base of the flame, which is more sensitive to the air in excess), signalling for all types of gases and for each opening of the gas valve the excess air values that are regarded excessively high (i.e.λ = 1.45). The excess air value used for the measurement of this temperature is variable in this case, as it changes according to the opening of the valve, though it still remains an effective indicator of the threshold. If this value is reached during the operation of the boiler (after the expiry of the time required to heat the combustion chamber), the boiler arrest is imposed (with the generation of an error code). To compensate the variability of the λ measured, it is possible to adjust all the temperature values reached in point λ, chosen as limit threshold (i.e. λ = 1.45), according to the command transmitted to the gas valve (see Figures 14 and 16). Similarly it is also possible to use the trend of the temperature differences Δθ = Θ2 - θl (illustrated above) and define the values beyond which the boiler must be shutdown using the same procedure based on the use of a single temperature (see Figure 14).

The standardized ionization current J is substantially independent from power; thus the maximum threshold for the excess of air can be set following the same procedure used for the example based on a single temperature by adapting the standardized current values measured to the varying opening degrees of the gas valve. If the two ionization currents Jl and J2 are used, it is possible to set both thresholds for the excessively high and low values of λ. The trend of the difference between the two ionization currents is characterized, as explained above, by a maximum and by a zero value (though it is alternatively possible to set another threshold value in order to have another reference for the excessively high or low values of air in excess. These two points can be used as thresholds (as in the example of Figure 16). As soon as one of these two typical points is reached, SUPERVISOR 9 commands the boiler arrest (generating an error ^r4de in some casεs\

Conclusions

The variants of the method described in the invention described so far are only some of the preferred versions. It is clear from the graphs of Figures from 3 to 10, that much data can be acquired by monitoring at least two physical magnitudes J and/or θ that characterize the flame and by also using other values that are indicative of the status of combustion with pre-mixing, like the degree of opening of gas valve 3 and the speed of fan 2. Having determined that two or more physical magnitudes J and/or θ can establish quantitatively more accurate relations between all the parameters typical of combustion, it is possible to state that any technician of the field is able to determine, using known procedures, mathematical relations to indirectly establish the parameters that cannot be acquired through a direct measurement and thus indirectly acquire other parameters.

As for as the type of mathematical models employed, the degree of accuracy is limited only by the effectiveness of the calculation and memory of the microprocessors with which it is economically reasonable to equip the boilers. The description has clearly underlined the advantages offered by the method in terms of capacity of establishing acceptable operating limits in terms of safety, hygiene and effectiveness of combustion. It is useful to notice, in particular, that these reference values are totally independent from the mathematical model employed to the point of being actually used for the calibration of the mathematical model. It is also important to take into account that these reference values are substantially independent or made independent from the parameters that cannot normally be measured in normal operating conditions, like the degree of dirt/deterioration of the sensors, the exact characteristics of the fuels, provided that these are within the acceptable limits established by standards. It is also worth noticing that value λ.o, which represents the optimal combustion value, often indicated as equivalent to about 1.3, changes in reality according to the thermal load. Thus, with very low thermal loads, close to the maximum threshold defined by the project, it is more difficult to maintain combustion within acceptable limits. This is why a more accurate estimate of λ permits the control and maintainance of good combustion within modulation ranges that are much wider than those used in the past. As a matter of fact, the experimentations of the method based on this invention have shown that it is possible to adjust the power with a range from 1 to above 6; for example from 4 to 27 kW. It is also important to underline that the methods and means used in this invention can be applied also to combustion units that employ liquid fuels, by simply applying the necessary changes that can be easily applied by expert technicians using the descriptions provided above.

Claims

Claim 1 Combustion control method consisting of gas combustion units with automatic pre-mixing of the air/gas mixture, wherein said control foresees the estimation of the value of λ characterized by the fact that the air in excess (λ.c) of said λ, in at least one working condition, is estimated by using at least two physical magnitudes representative of the combustion status (J, θ; Jl, J2; θl, Θ2), measured close to the flame in standard operating conditions. Claim 2 Combustion control method according to the previous claim characterized by the fact that the aforesaid value (λ.c) of said λ is also estimated using one or more status signals of the combustion unit (T.out; T.in; GV; RPM.m). Claim 3 Combustion control method as per any of the previous claims characterized by the fact that said estimated value (λ.c) is represented by the result of the mathematical model that employs, as input variables, said two or more of the above-described physical magnitudes representative of the combustion status (J, θ; Jl, J2; θl, Θ2) and, if necessary, said one or more status signals of the combustion unit (T.out; T.in; GV; RPM.m)

Claim 4 Combustion control method as per previous claim characterized by the fact that said mathematical model is experimentally determined by the manufacturer by using known methods, by measuring and by registering in laboratory conditions, for the selected models and in several combustion conditions, the values of the said two or more above-mentioned physical magnitudes of the combustion status (J, θ; Jl, J2; θl, Θ2) and of one or more of said above-mentioned status signals of the combustion unit (T.out; T.in; GV; RPM.m), and by recording also the resulting λ values. Claim 5 Combustion control method as per previous claim characterized by the fact that said experimental combustion conditions include:

- all the adjustment range of the thermal capacity involved ;

- a range of λ that extends beyond those actually acceptable; i.e.from 1.0 to 1.5; - the use of both the reference gas and of the two extreme limit gases of the group involved. Claim 6 Combustion control method as per previous claim characterized by the fact that the emissions of CO are maintained within the limit of 1000 ppm during said experimental measurements.

Claim 7 Combustion control method as per previous claims 5 and 6 characterized by the fact that said experimental combustion conditions include at least one specific combustion condition that produces values of λ (λ.s; λ.b; λ.m; λ.z), which can be easily identified by the reading of characteristic values (J.s; Jl.b, J2.b; Jl.m,

J2.m; Jl.z, J2.z) of the ionization currents J or of their combinations (Jl.b / J2.b;

T1 _m . IT m- T1 r, _ T? -Λ

Claim 8 Combustion control method as per claim 7 characterized by the fact that said two or more of the physical magnitudes related to combustion (J, θ; J, Θl,

Θ2) include at least one ionization current (J) and that said specific combustion condition corresponds to that in which the maximum value of J = J.s is obtained which yields a known value of λ - λ.s, substantially equivalent to the stoichiometric value (= 1) and stable as compared to the other parameters that influence combustion.

Claim 9 Combustion control method as per claim 7 characterized by the fact that said two or more physical magnitudes of the combustion status (J, θ; J, θl, Θ2) include at least two ionization currents (Jl, J2) and that said specific combustion condition corresponds to the condition that yields the maximum value of RJ of ratio Jl / J2 that offers a value of λ= λ.b, substantially unchanged as compared to the other parameters influenced by combustion and typically very close to the optimal air excess value λ.o. Claim 10 Combustion control method as per claim 7 characterized by the fact that said two or more physical magnitudes of the combustion status (J, θ; J, Θl, Θ2) include at least two ionization currents (Jl, J2) and that said specific combustion condition corresponds to the condition that yields the maximum value of the difference Jl - J2 = ΔJ that corresponds to a value of λ= λ.m = λ.min.e substantially remains unaltered as compared to the other parameters influenced by combustion, and that is typically indicative of combustion conditions with an excess of air λ.min.e at a value below the lower limit of the tolerated level. Claim 11 Combustion control method as per claim 7 characterized by the fact that said two or more physical magnitudes of the combustion status (J, θ; J, θl, Θ2) include at least two ionization currents (Jl, J2) and that said specific combustion corresponds to the condition that yields a null value for the difference Jl - J2 = ΔJ that corresponds to a value of λ= λ.z = λ.max.e that substantially remains unaltered as compared to the other parameters influenced by combustion, and that is typically indicative of combustion conditions a with an excess of air λ.max.e at a value above the higher limit of the tolerated leveL Claim 12 Combustion control method as per claims from 3 to 11 characterized by the fact that said two or more physical magnitudes of the combustion status (J, θ; J, θl, Θ2) are used in the standardized mathematical model after being divided; i.e. for standardization factors that assume these values (J.s, θ.s; J.b, Θ.b; J.m, θ.m; J.z, θ.z) close to the values of λ (λ.s; λ.b; λ.m; λ.z); being easily identified as said values are stored and used until they are updated. Claim 13 Combustion control method as per claim 12 characterized by the fact that said standardization factors are represented by the values (J.s, θ.s) acquired by J and θ close to the value of λ = λ.s. Claim 14 Combustion control method as per claim 12 characterized by the fact that said standardization factors are represented by. values (J.b, θ.b) acquired by J and θ close to the value of λ = λ.b. Claim 15 Combustion control method as per claim 12 characterized by the fact that said standardization factors are represented by values (J.m, θ.m) acquired by J and θ close to the value of λ = λ.m.

Claim 16 Combustion control method as per claim 12 characterized by the fact that said standardization factors are represented by values (J.z, θ.z) acquired by J and θ close the value of λ = λ.z. Claim 17 Combustion control method as per claims from 3 to 16 characterized by the fact that the input variables of the mathematical model may alternatively include:

- two currents (Jl, J2),

- one current and one temperature (J, θ), - one current and two temperatures (J, θl, Θ2),

- two currents and one temperature (Jl, J2, θ),

- two currents and two temperatures (Jl, J2, θl, Θ2),

- any of the above values above the measured speed (RPM.m) of the fan (2) and/or the opening value (GV) of the gas valve (3). Claim 18 Combustion control method as per one or more claims from 3 to 17 characterized by the fact that it is possible to process and implement in a single control system several mathematical models (7) close to different sets of input variables and the manufacturer determines, for each boiler model, which control systems must be enabled and/or considered relevant for the boiler model, using physical magnitudes of the combustion status (J, θ; Jl, J2; θl, Θ2) and of the status signals of the combustion unit (T.out, T.in, GV, RPM.m). Claim 19 Combustion control method as per one or more claims from 3 to 18 characterized by the fact that said mathematical model is simply linear and parametrised. Claim 20 Combustion control method as per one or more claims from 3 to 18 characterized by the fact that said multiplying coefficients (k.π. k.₀₂) of at least some variables (Jl, Θ2) of this mathematical model are selected according to other variables (J2, Θl). Claim 21 Combustion control method as per one or more claims from 3 to 20 characterized by the fact that said variables included in this mathematical model include combinations of input variables (RJ or Δθ). Claim 22 Combustion control method as per one or more claims from 3 to 21 characterized by the fact that value λ.c of λ, calculated using the mathematical model in one specific instant

(t), according to the variables read in previous instants (t - 1 ), is taken as actual value of λ close to the current (t) instant. Claim 23 Combustion control method as per claims 1 or 2 characterized by the fact that the calculated value λ (λ.o) is considered optimal for current operating conditions. Claim 24 Combustion control method as per previous claim characterized by the fact that the basic parameters used to calculate the said optimal value of λ (λ.o) include: - the physical magnitudes of the combustion status (J, θ; J1, J2; θl, Θ2);

- the status signals of the combustion unit (T.out; T.in; GV; RPM.m);

- the thermal power request signals (T.outsp). Claim 25 Combustion control method as per claims 23 or 24 characterized by the fact that - an appropriate value of λ (λ.sp) is calculated;

- appropriate signals (A, G) are sent to the fan (2) and to the gas valve (3), to make sure that the calculated value λ (λ.o) is equivalent to said appropriate value of λ (λ.sp). Claim 26 Combustion control method as per claims from 23 to 25 characterized by the fact that said appropriate value of λ (λ.sp) coincides with the value of λ that is considered optimal for current combustion conditions (λ.o). Claim 27 Combustion control method as per claims 1 or 2 characterized by the fact that a sampling is carried out in specific instants by changing the combustion conditions so that the air in excess λ reaches values of λ (λ.s; λ.b; λ.m; λ.z), which can be easily identified by the reading of characteristic values (J.s; Jl.b, J2.b; Jl.m, J2.m; Jl.z, J2.z) of the related ionization currents J or of their combination (Jl.b / J2.b; Jl.m - J2.m; Jl.z - J2.z) whereas - said characteristic values (J.s; Jl.b, J2.b; Jl.m, J2.m; Jl.z, J2.z) are stored

- said sampling instants are determined on the basis of criteria determined by the manufacturer. Claim 28 Combustion control method as per claim 27 characterized by the fact that said one or more characteristic values (J.s; Jl.b, J2.b; Jl.m, J2.m; Jl.z, J2.z) correspond to operating conditions with typical air excess λ values (λ.s; λ.b; λ.m; λ.z) known to the manufacturer. Claim 29 Combustion control method as per the previous claim characterized by the fact that certain typical air excess values λ (λ.m; λ.z) are regarded as acceptable minimum and maximum limit values (λ.min.e; λ.max.e) for the air excess λ. Claim 30 Combustion control method as per claim 28 characterized by the fact that the typical air excess λ (λ.b) is regarded as reference value (λ.o) for a good combustion.

Claim 31 Combustion control method as per claim 28 characterized by the fact that the temporarily acceptable lower and upper values for the excess air λ (λ.min.a; λ.max.a) are respectively determined as equivalent to the intermediate value between the minimum acceptable value (λ.min.e) and the optimal combustion value (λ.b), and to the intermediate value between the optimal combustion value

(λ.b) and the maximum acceptable value (λ.min.a). Claim 32 Combustion control method as per claims 27 and 12 characterized by the fact that one or more of said typical values (J.s; Jl.b, J2.b; Jl.m, J2.m; Jl.z, J2.z) and of the corresponding typical flame temperature values (θs; θl.b, Θ2.b; θl.m, Θ2.m; θ 1.z, Θ2.z) are used for a regulator restandardization. Claim 33 Combustion control method as per claim 27 and as one of the claims from 3 to

22 characterized by the fact that one or more of said typical values (J.s; Jl .b, J2.b; Jl .m, J2.m; Jl .z, J2.z) are used to recalibrate the input parameter of the mathematical model. Claim 34 Combustion control method as per claims 27 and/or 28 and one of the claims

characterized by the fact that one or more of said typical λ air excess values (λ.s; λ.b; λ.m; λ.z) is used to recalibrate (λ.os) the output parameter of this mathematical model. Claim 35 Combustion control method as per claims 1 or 2 characterized by the fact that the measurement, close (p.l) to the burner (2), of a sufficiently low flame temperature θ (Θl) is considered indicative of intolerably high air excess values

(λ.max.e). Claim 36 Combustion control method as per claims 1 or 2 characterized by the fact that the measurement of a sufficiently high preset flame temperature difference (Δθ) is considered indicative of intolerably high air excess values (λ.max.e).

Claim 37 Combustion control method as per claims 1 or 2 characterized by the fact that the flame temperature value Θ2 in position (p.2), sufficiently different from the boiler (2), is used to estimate the generated thermal power. Claim 38 Combustion control method as per claims 1 or 2 characterized by the fact that the flame temperature difference (Δθ) is used to estimate the thermal power generated. Claim 39 Combustion control method as per any of the above-mentioned claims characterized by the fact that the J value measured close to the point that corresponds to the maximum value

(J.s) is not used directly for the mathematical model, but first multiplied by an experimentally defined parameter in order to attain the correct value (J.s.corr). Claim 40 Control system (7) characterized by the fact that it is suitable for the implementation of one or more of the methods described in previous claims.

Claim 41 Control system (7) as per previous claim characterized by the fact it includes electronic calculation and storage means. Claim 42 Control system (7) as per claims 40 or 41 characterized by the fact that it includes at least the following logical blocks:

- a virtual sensor of λ (8) suitable to calculate the value of λ (λ.c)

- a supervisor (9) suited at least to

- execute the sampling of typical data (J.s; Jl.b, J2.b; Jl.m, J2.m; Jl.z, J2.z), - use specific signals (v.m.c, λ.os) to correct the processed values of the virtual sensor of λ (8),

- determine the target value of λ (λ.sp) during current operation,

- regularly determine, particularly during start-up, specific operating conditions, - detect critical (λ.min.e; λ.max.e; λ.min.a; λ.max.a) or optimal (λ.b) operating conditions, - monitor the operating parameters and block the unit if abnormal conditions occur. Claim 43 Boiler characterized by the fact that it is suitable to implement one or more of the methods described in claims from 1 to 39.

Claim 44 Boiler characterized by the fact that it includes one or more of the means described in claims from 40 to 42. Claim 45 Boiler designed according to one or more of the previous claims, characterized by the fact that it employs natural gas. Claim 46 Boiler designed according to one or more claims from 1 to 44, characterized by the fact that it employs liquid gas. Claim 47 Boiler designed according to one or more claims from 1 to 44, characterized by the fact that it employs liquid fuels.