CN111076161A

CN111076161A - Method and device for determining drum water level in subcritical boiler of coal-fired unit

Info

Publication number: CN111076161A
Application number: CN201911241456.6A
Authority: CN
Inventors: 刘磊; 康静秋; 杨振勇; 郑重; 张红侠; 程亮; 解冠宇; 李永富
Original assignee: Inner Mongolia Daihai Electric Power Generation Co ltd; State Grid Corp of China SGCC; North China Electric Power Research Institute Co Ltd
Current assignee: Inner Mongolia Daihai Electric Power Generation Co ltd; State Grid Corp of China SGCC; North China Electric Power Research Institute Co Ltd
Priority date: 2019-12-06
Filing date: 2019-12-06
Publication date: 2020-04-28
Anticipated expiration: 2039-12-06
Also published as: CN111076161B

Abstract

The invention provides a method and a device for determining drum water level in a subcritical boiler of a coal-fired unit, which optimize a drum water level test value through a linear regression algorithm in machine learning, and the optimized drum water level data can more accurately reflect the real situation of the boiler water level, thereby solving the problem of drum water level control of the unit under the deep peak regulation working condition, providing favorable guarantee for the safe production of the unit, providing a better solution for process control on the basis of ensuring the safe operation of the unit, improving the variable load capacity and adaptability of the unit, improving the regulation performance of various indexes of the unit, and simultaneously meeting the requirements of an electric network on AGC and primary frequency modulation management and examination under the deep peak regulation condition. Therefore, the safety and the economical efficiency of the operation of the unit are improved, the environmental protection index of the unit is ensured, and the economic benefit and the social benefit of the thermal generator set participating in power grid examination are enhanced.

Description

Method and device for determining drum water level in subcritical boiler of coal-fired unit

Technical Field

The invention relates to the field of boiler water level measurement, in particular to a method and a device for determining drum water level in a subcritical boiler of a coal-fired unit.

Background

With the change of global energy patterns, particularly the change of new energy policies in China, a large number of new energy power stations are connected to a power grid like bamboo shoots in spring after rain. Therefore, energy consumption is reduced, but due to the problems that the thermal power generating unit is influenced by environmental factors and the like, a certain thermal power generating unit needs to be matched and supported necessarily.

Under the background, the energy bureau provides a scheme for flexibly transforming the thermal power generating unit. Therefore, higher requirements are put forward for the common coal-fired thermal power generating units. Namely, under the condition that the load of the power grid is in a low valley, the coal-fired unit is required to be adjusted to a lower load downwards under the condition that the new energy unit is ensured to generate electricity normally, and the requirements of the power grid and users are met.

On the premise of ensuring the operation safety, the coal-fired unit simultaneously meets the dispatching requirement of AGC, so that the coordination control and all related subsystems thereof are required to have the deep peak regulation capability. This requires more demanding capabilities including plant performance, operational tuning, and control logic optimization. The normal load adjustment range of the conventionally designed coal-fired thermal power generating unit is between 50% Pe and 100% Pe, and the peak regulation range of the unit with deep peak regulation capability is 40% -100% Pe, even lower to 30% or even 20% Pe. When the unit is reduced to a lower load range, the characteristics of the unit, particularly the boiler, are greatly changed, and the characteristics of the auxiliary machine are also greatly changed, so whether the unit can have deep peak regulation capacity or not and the peak regulation limit is determined according to an actual operation test. And meanwhile, when any one operation parameter exceeds a normal operation range, the deep peak regulation capacity of the unit can be limited. For subcritical drum boilers, the control of drum water level is excellent and directly influences the deep peak regulation capability of the unit.

Disclosure of Invention

In order to solve at least one of the problems, the invention provides a method and a device for determining the drum water level in a subcritical boiler of a coal-fired unit.

An embodiment of one aspect of the invention provides a method for determining a drum water level in a subcritical boiler of a coal-fired unit, which comprises the following steps:

acquiring a plurality of water level measurement data of two sides of a boiler barrel in a subcritical boiler;

inputting a plurality of water level measurement data serving as independent variables into a preset linear regression function to obtain steam drum water level optimization data;

and inputting the steam drum water level optimization data into a control device, so that the control device adjusts the steam drum water level in the subcritical boiler according to the steam drum water level optimization data.

In certain embodiments, further comprising:

generating a training sample set according to historical water level measurement data, wherein the training sample set comprises a plurality of training samples, the training samples comprise a plurality of water level measurement data in each measurement, the median of the plurality of water level measurement data, and optimized water level data labels corresponding to the plurality of water level measurement data in each measurement and the median of the plurality of water level measurement data;

and applying the training sample set to train the linear regression function.

In some embodiments, before training the linear regression function, the drum level determination method further comprises:

establishing the linear regression function according to the quantity of the water level measurement data obtained by each measurement, wherein the linear regression function comprises independent variables with corresponding quantity;

and constructing a loss function of the linear regression function according to the distance between each independent variable and the median of the independent variables.

In certain embodiments, further comprising:

and constructing a weight constraint function according to the influence of the boiler evaporation capacity on each weight in the linear regression function.

In some embodiments, the constructing the loss function of the linear regression function comprises:

constructing a first function structure according to the distance between each independent variable and the median of the independent variables;

constructing a second function structure according to the weight function of the linear regression function and the regular term coefficient;

summing the first and second function structures to form the loss function.

In some embodiments, said applying said training sample set to train said linear regression function comprises:

executing a first iteration operation, executing an optimization operation on a group of training samples under an initial regular term coefficient based on a gradient descent method, generating the weight of each independent variable, and replacing the initial regular term coefficient with the adjusted regular term coefficient to execute the optimization operation until the sum of the weights of each independent variable is smaller than a set range;

executing a second iteration operation, calculating the mean square error of the loss values of all the training samples under the weight of each independent variable output by the first iteration operation, and reselecting a group of training samples to execute the first iteration operation until the mean square error of the loss values of all the training samples is lower than a set threshold value;

and outputting the final weight of each independent variable to obtain a linear regression function after training is finished.

In some embodiments, the linear regression function includes a plurality of linear regression functions, each linear regression function corresponding to a predetermined operation load segment, and the method further includes:

determining an operation load section of the water level measurement data;

the inputting of the plurality of water level measurement data as independent variables to a preset linear regression function comprises:

and inputting a plurality of water level measurement data serving as independent variables into the linear regression function of the corresponding operation load section.

In certain embodiments, further comprising:

generating a plurality of training sample sets according to historical water level measurement data, wherein the training sample sets correspond to each operation load section one by one, each training sample set comprises a plurality of training samples, and each training sample comprises a plurality of water level measurement data in each measurement under the corresponding load section, the median of the plurality of water level measurement data, and an optimized water level data label corresponding to the plurality of water level measurement data in each measurement and the median of the plurality of water level measurement data;

and applying each training sample set to train the linear regression function of the corresponding load segment.

Another embodiment of the present invention provides a device for determining a drum water level in a subcritical boiler of a coal-fired unit, including:

the system comprises an acquisition module, a data acquisition module and a data processing module, wherein the acquisition module is used for acquiring a plurality of water level measurement data of two sides of a boiler barrel in a subcritical boiler;

the input module is used for inputting the water level measurement data serving as independent variables into a preset linear regression function to obtain steam drum water level optimization data;

and the control module inputs the steam drum water level optimization data into a control device so that the control device adjusts the steam drum water level in the subcritical boiler according to the steam drum water level optimization data.

In certain embodiments, further comprising:

the training sample set generating module is used for generating a training sample set according to historical water level measurement data, wherein the training sample set comprises a plurality of training samples, the training samples comprise a plurality of water level measurement data in each measurement, the median of the plurality of water level measurement data, and optimized water level data labels corresponding to the plurality of water level measurement data in each measurement and the median of the plurality of water level measurement data;

and the training module is used for applying the training sample set to train the linear regression function.

In some embodiments, the drum level determining apparatus further comprises:

the linear regression function establishing module is used for establishing a linear regression function according to the quantity of water level measurement data obtained by each measurement, and the linear regression function comprises independent variables with corresponding quantity;

and the loss function building module is used for building the loss function of the linear regression function according to the distance between each independent variable and the median of the independent variables.

In certain embodiments, further comprising:

and the weight constraint function construction module is used for constructing a weight constraint function according to the influence of the boiler evaporation capacity on each weight in the linear regression function.

In some embodiments, the loss function building module comprises:

the first function structure construction unit is used for constructing a first function structure according to the distance between each independent variable and the median of the independent variables;

the second function structure construction unit is used for constructing a second function structure according to the weight function of the linear regression function and the regular term coefficient;

and the adding unit is used for adding the first function structure and the second function structure to form the loss function.

In certain embodiments, the training module comprises:

the first iteration operation execution unit executes first iteration operation, performs optimization operation on a group of training samples under an initial regular term coefficient based on a gradient descent method, generates the weight of each independent variable, and replaces the initial regular term coefficient with the adjusted regular term coefficient to execute the optimization operation until the sum of the weights of each independent variable is smaller than a set range;

the second iteration operation execution unit is used for executing second iteration operation, calculating the mean square error of the loss values of all the training samples under the weight of each independent variable output by the first iteration operation, and reselecting a group of training samples to execute the first iteration operation until the mean square error of the loss values of all the training samples is lower than a set threshold value;

and the independent variable weight output unit outputs each final independent variable weight to obtain the linear regression function after the training is finished.

In some embodiments, the linear regression function includes a plurality of linear regression functions, each linear regression function corresponding to a predetermined operation load segment, and the apparatus further includes:

the operation load section determining module is used for determining the operation load section of the water level measurement data;

the input module inputs the water level measurement data as independent variables to the linear regression function of the corresponding operation load section.

In certain embodiments, further comprising:

the training sample set generating module is used for generating a plurality of training sample sets according to historical water level measurement data, wherein the training sample sets correspond to each operating load section one by one, each training sample set comprises a plurality of training samples, and each training sample comprises a plurality of water level measurement data in each measurement of the corresponding load section, the median of the plurality of water level measurement data, and an optimized water level data label corresponding to the plurality of water level measurement data in each measurement and the median of the plurality of water level measurement data;

and the training module is used for applying each training sample set to train the linear regression function of the corresponding load segment.

A further embodiment of the present invention provides an electronic device, which includes a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the drum level determining method as described above.

A further aspect embodiment of the invention provides a computer readable storage medium having stored thereon a computer program which, when being executed by a processor, carries out the steps of the drum level determination method as described above.

The invention has the following beneficial effects:

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.

Fig. 1a shows one of the schematic diagrams of the cross fluctuation of the drum level on both sides due to combustion, hydrodynamic forces, etc.

Fig. 1b shows a second schematic diagram of the cross fluctuation of the drum level on both sides due to combustion, hydrodynamic forces, etc.

Fig. 2 shows a schematic flow chart of a method for determining drum water level in a subcritical boiler of a coal-fired unit in the embodiment of the invention.

Fig. 3 is a schematic structural diagram showing a drum water level determining device in a subcritical boiler of a coal-fired unit in the embodiment of the invention.

FIG. 4 shows a schematic structural diagram of an electronic device suitable for use in implementing embodiments of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

Because the boiler type of a power station boiler, particularly a coal-fired unit is complex, the operation working condition is complicated, and the interference factors are more, the steam drum water level is more influenced by the interference factors, not only the combustion characteristic of the boiler is influenced, but also the boiler combustion is influenced by a plurality of factors, such as the change of coal quality, the change of air volume, the change of parameters of a pulverizing system and the like; but also influenced by boiler water power, and the boiler water power characteristics are restricted by the boiler structure, combustion condition and the change of water supply system parameters. The change in drum level is therefore a result of a combination of factors. The first technical scheme is a classical steam drum water level control scheme, requirements of working conditions of most existing units can be met, however, the existing thermal power generating units are forced to be transformed in necessary flexibility along with large-scale access of new energy power stations to a power grid, the load of the units is required to be reduced to be lower at a load valley section of the power grid, accordingly, the operation working conditions of main and auxiliary machines of the units are changed greatly, and the basic three-impulse and single-impulse adjusting scheme cannot meet the water level control requirements.

In the second technical scheme, the control of the steam drum water level by the large load change of the unit in a certain range is met to a certain extent, the requirement on the safe operation of the unit is met to a certain extent, and the fluctuation of the steam drum water level and the requirement on the self characteristic of a water supply system are not deeply considered on the premise of deep peak shaving of the unit, so that the high control requirement cannot be adapted under the condition that the unit is low in load.

In the third technical scheme, due to the adoption of a more advanced algorithm, the problems can be solved to a certain extent. However, most of the control systems of thermal power generating units adopt a Distributed Control System (DCS), and most of the control systems do not have such algorithms. The algorithm cannot be realized by the original control system, so that the algorithm cannot be applied in a large range.

Based on the situation, the method is suitable for the requirement of a power grid on deep peak regulation of the thermal power unit under the condition of new energy access under the condition of meeting the existing conditions and facing various predicaments to the thermal power unit. In particular, an integral solution is provided for controlling the water level of a subcritical boiler drum.

The subcritical boiler drum water level is an important parameter index for unit operation. The method marks the excellent running state of the unit and represents an important mark for judging whether the boiler hydraulic power circulation is normal or not. Factors for generating water level changes are complex, such as the condition of hydrodynamic circulation of a unit, the change of combustion of a boiler furnace, the disturbance of a water supply system, the change of steam side parameters, the structural characteristics of the interior of a steam drum, the installation position and the installation precision of a steam drum water level measuring device, the precision of a measuring transmitter and the like. Therefore, the water level of the steam drum is an important protection parameter of the boiler in general. The control effect of the system directly influences the safe operation of the unit. Especially under the operating mode of the unit in the degree of depth peak regulation, the operating condition of the main auxiliary engine of the unit is more severe than under the normal operating mode, the factors such as huge change of boiler combustion characteristics and the like all generate huge disturbance to the water level, the water level change is violent, and the unit can not operate safely and stably. Therefore, the invention provides a safer and more reliable control scheme for controlling the water level of the steam drum under the working condition. The condition that the water level fluctuation is severe can be effectively improved, so that the water level of a steam drum can be ensured to be in a safe range, the unit can operate in a lower-load interval, the unit has the capacity of deep peak regulation, certain economic benefit is brought to a power plant, and more favorable support is provided for the deep peak regulation of a power grid.

A boiler barrel in a subcritical boiler is intermediate equipment which is connected with a water-cooled wall and a downcomer and plays a role in steam-water separation, wherein the water-cooled wall is arranged inside a hearth, a working medium which absorbs heat after the boiler is started is reduced in density due to vaporization, and the downcomer is generally arranged outside and is basically unchanged in density. The working medium of the natural circulation boiler generates natural water circulation due to the difference between the density of the working medium of the water wall tube and the density of the working medium of the downcomer. The steam is separated to a superheater at the rear end in the steam drum, and low-temperature water supplied by the water supply pump enters the steam drum as make-up water through the economizer. Therefore, the water level of the steam pocket of the subcritical natural circulation steam pocket boiler marks the comprehensive mark of the stability of the working medium in the circulation and also marks the quality balance of the working medium. Therefore, the stability of the water level of the steam drum is controlled, the stable operation of the boiler is realized, and the safe operation of the unit plays a vital role.

The measurement of the drum water level of the existing coal-fired unit generally adopts a differential pressure type measuring element which is arranged on two sides of a boiler barrel at the same time, and generally adopts a mode of two on one side. The regulated quantity of the classical three-impulse control is a mode of taking three measuring points, namely two measuring points at one side and one measuring point at one side.

In the prior art, a three-in-one processing mode is adopted in the control logic, when one measuring point has a problem, the other two measuring points can be automatically selected and averaged to be used as the controlled quantity. The processing mode can represent the real water level condition of the system under most conditions, and the condition of extreme value selection can be avoided by selecting the middle value from the three values. However, the water level is affected by various factors, including the combustion condition of the boiler, the change of the hydrodynamic circulation of the boiler, the change of the evaporation capacity and the feed water flow, and the like, especially under the condition that the unit operates at low load or even ultra-low load, the combustion condition inside the boiler changes, the change of the characteristics of the feed water system is severe, and great influence is caused on the change of the water level, and particularly, the cross phenomenon of the water levels at two sides is frequent, and the treatment mode in the third extraction has unstable effect on the PID control. The specific phenomenon is shown in fig. 1a and 1 b. In fig. 1a and 1b, three of the

drum water levels

1, 2, 3 are simulated as

signals

1, 2, 3, and the selected drum water level 4 changes to a selected signal 4. When the values of the

signals

1 and 3 are larger than the value of the signal 2, namely the curves of the

signals

1 and 3 are above the signal 2, the small value in the

signals

1 and 3 is selected as the signal 1; when the

signal

1, 3 values are smaller than the signal 2 values, i.e. the

signal

1, 3 curves are below the signal 2, the larger of the

signals

1, 3 is selected, in the figure signal 3. The process of selecting signal 2 occurs at the time between the two case changes, i.e., during the crossing of

signals

1, 3 and signal 2. The calculation in the three selections shows a signal selection process, wherein the signal 1- > signal 2- > signal 3- > signal 2- > signal 1 is repeatedly circulated as long as the water level is crossed. In this case, the operation result of the PID is analyzed. The PID algorithm itself is simple and has the basic formula:

where error is the PID controller inlet offset, i.e., the difference between the set point and the measured value. Assuming that the water level set value is a fixed value, for example, -1, the median value selected from the simplified diagram between time 2 and time 3 is signal 1, and gradually approaches the set value, and the proportional action term in the output of PID is that action is generated at this time, and because signal 1 is approaching the set value downwards, the increment generated at each time is gradually smaller than that generated at the previous time, that is, the output is changed towards the increasing direction. While the direction of the integral term deviation does not change, the output is reduced in the direction of the decrease, but the increment is reduced (the derivative effect is less used and is not considered here).

When reaching between times 3-4, signal 2 is selected as output during this time because of the logical middle of three,

signal

1, 3 is overridden by signal 2. At the same time, the inlet deviation value of PID will change, and the original signal falling direction is changed into increase, and the proportion function item of PID, namely the direction change of output decrease. The deviation direction of the integral term is not changed, so that the output acts in the direction of reduction, and the increment is increased.

After time 4, the value of signal 2 exceeds

signals

1 and 3, and the triple-decimation algorithm selects signal 3 as output. The inlet bias direction of the PID is similar to that at time 2-3.

From the above analysis, it can be seen that the change at time 3-4 is quite opposite to the change at time 2-3, and the obtained median value is opposite to the effect of PID, and is a step disturbance caused by instantaneous switching. In this case, the algorithm in the third-order process has an adverse effect on the PID control output, and when the signal crossing process is long or the variation amount is large, the PID control output amount generated by the PID is increased under the condition that the parameter is not changed. Under extreme conditions, problems may arise that are detrimental to system stability.

Fig. 2 illustrates a method for determining a drum water level in a subcritical boiler of a coal-fired unit according to an embodiment of an aspect of the present invention, which specifically includes:

s1: acquiring a plurality of water level measurement data of two sides of a boiler barrel in a subcritical boiler;

s2: inputting a plurality of water level measurement data serving as independent variables into a preset linear regression function to obtain steam drum water level optimization data;

s3: and inputting the steam drum water level optimization data into a control device, so that the control device adjusts the steam drum water level in the subcritical boiler according to the steam drum water level optimization data.

According to the method for determining the drum water level in the subcritical boiler of the coal-fired unit, the drum water level test value is optimized through a linear regression algorithm in machine learning, the optimized drum water level data can reflect the real situation of the boiler water level more accurately, the problem of drum water level control of the unit under the deep peak regulation working condition is solved, the favorable guarantee is provided for the safe production of the unit, a better solution is provided for the process control on the basis of guaranteeing the safe operation of the unit, the variable load capacity and the adaptability of the unit can be improved, the regulation performance of various indexes of the unit is improved, and meanwhile the requirements of an electric network on AGC and primary frequency modulation management and examination under the deep peak regulation condition are met. Therefore, the safety and the economical efficiency of the operation of the unit are improved, the environmental protection index of the unit is ensured, and the economic benefit and the social benefit of the thermal generator set participating in power grid examination are enhanced.

In the invention, the number of the plurality of water level measurement data can be determined according to specific requirements, generally speaking, the current common water level measurement data is three impulse data, namely the number of the water level measurement data is three, two measurement elements are respectively arranged on one side of the boiler barrel, and one measurement element is arranged on the other side of the boiler barrel. That is, in the aspect of data acquisition, the existing three-impulse mode is preferably adopted, and the number of the water level measurement data measured each time is three, which is not described herein again.

In step S1, the water levels on both sides of the drum are measured by the differential pressure type measuring element to obtain three water level measurement data, and the three water level measurement data are transmitted by a data transmission method (wired or wireless).

In step S2, the three water level measurement data are taken as independent variables into a preset linear regression function to generate optimized data of drum water level. Specifically, the linear regression function should be a combined structure including a plurality of arguments and corresponding weights, i.e., y ═ k₁x₁+k₂x₂+k₃x₃…, wherein the weights thereof satisfy a weight function k₁+k₁+k₃… is equal to 1 and the weighting function is used as a constraint for k value. In addition, each linear regression function necessarily corresponds to a loss function for fitting the linear regression function.

Further, the linear regression function can be more fit by a training mode, that is, the invention further comprises the following steps:

s01: generating a training sample set according to historical water level measurement data, wherein the training sample set comprises a plurality of training samples, the training samples comprise a plurality of water level measurement data in each measurement, the median of the plurality of water level measurement data, and optimized water level data labels corresponding to the plurality of water level measurement data in each measurement and the median of the plurality of water level measurement data;

s02: and applying the training sample set to train the linear regression function.

The optimized water level data labels in the training samples are used for marking the training data by using the known optimized water level data as the data labels. The training data samples are primarily derived from raw data that is transmitted to the system through the transmitter measurements in the actual field. The method also comprises data of boiler evaporation capacity, the operating condition of the boiler is judged according to the boiler evaporation capacity, and data information is respectively marked according to the marks. For example: when the evaporation capacity of the boiler is in the interval of 20-25% Pe, the original steam drum water level data is marked as 20% Pe data (30% Pe and 40% Pe in the same way). And after finishing the classified collection of the data, processing the data.

The data processing comprises the following parts, namely the elimination of the first interference data, the separation and the cross validation of the second training data and the validation data set. For example, if the total number of data after interference extraction after 20% Pe data are removed is 8000, the data are randomly divided into 4 groups of data such as A, B, C, D, etc., each group of data is 2000 information points, 200 points are extracted from each group of information points to be used as a verification data set, and the rest 1800 points are used as a training data set. After each group of data training is finished, after the group meets corresponding indexes after verification, other groups of verification data sets are full of the verification data sets. For example, after training and verification of the data A, the verification data set of B, C, D is used to evaluate the model. If any group of model verification indexes is unqualified, classifying 8000 information points again, and repeating the above process until the information points are qualified.

By adopting the random cross validation method, the dependence of the model on the training data set is effectively avoided, and the problem that the algorithm falls into local optimization is solved to a certain extent.

Training the linear regression function with training sample set, the essence of which is to continuously optimize respective variablesk value (i.e., the weight of the independent variable), in the case of three impulses, under constraint k₁+k₁+k₃When the function value of the loss function is minimized, the optimal k is obtained by combining the preset loss function under 1₁、k₁And k₃Combinations of (a) and (b).

The linear regression function in the invention can be established on line or off line, that is, the invention can utilize the established linear regression function, and can also be realized by establishing and reusing the linear regression function.

The invention provides the following steps of establishing a linear regression function and a corresponding loss function, which comprise:

s11: establishing the linear regression function according to the quantity of the water level measurement data obtained by each measurement, wherein the linear regression function comprises independent variables with corresponding quantity;

s12: and constructing a loss function of the linear regression function according to the distance between each independent variable and the median of the independent variables.

In some embodiments, the present invention further includes a step of constructing a weight constraint function, that is, in the present invention, further including:

s03: and constructing a weight constraint function according to the influence of the boiler evaporation capacity on each weight in the linear regression function.

S04: the weight constraint function takes into account the effect of boiler boil-off on each weight in the linear regression function.

There are many factors that affect the drum level, such as boiler load (boiler evaporation), changes in combustion, changes in pressure, and so forth. The influence between water levels is far from enough to be considered, but the influence factors are considered too much, so that the data volume is huge, and a regression function corresponding to the algorithm cannot be obtained. The main influencing factors are taken into account, i.e. the pointer corrects the boiler evaporation (boiler load) as the main influencing factor for different stages of the weight of the linear regression function. The method has the advantages of ensuring the real-time performance of the algorithm and objectively covering the influence of the change of the operation working condition of the whole boiler on the change of the water level.

In some embodiments, the step S11 is measured in a three-impulse manner, that is, the water level measurement data includes three, and the linear regression function is a function of three independent variables, and each independent variable corresponds to a weight, that is, the calculation expression is:

y＝k₁lvl1+k₂lvl2+k₃lvl3

in some embodiments, the distance of each argument from the median of the arguments in step S12 may be calculated by the following formula:

further, in a preferred embodiment, the loss function is constructed by considering the influence of the boiler evaporation amount on each weight in the linear regression function and the distance of the signal in the three-tap, and the construction steps include:

s001: constructing a first function structure according to the distance between each independent variable and the median of the independent variables;

s002: constructing a second function structure according to the weight function of the linear regression function and the regular term coefficient;

s003: summing the first and second function structures to form the loss function.

Specifically, the loss function of the present invention may be:

wherein lvl1, lvl2 and lvl3 are actual water level values (i.e. multiple water level measurement data in the present invention) obtained by an actual DCS system, k₁,k₂,k₃Is the parameter (i.e., corresponding weight) being optimized.

Wherein lvl1⁽⁰⁾A value at time 0, lvl1, representing the measured value of water level 1⁽⁵⁰⁰⁾The data value at time 500.

Representing the true value, replaced by the median value at time 0-500, respectively.

The loss function considers the distance of the three-hit signal on one hand, combines the influence of the boiler evaporation capacity on the weight function, and simultaneously adds a constraint term k in order to ensure that the optimizing data is in an allowable range and prevent the occurrence of the overfitting phenomenon of the data₁+k₂+k ₃1, α (k) in the loss function₁+k₂+k₃-1)²Wherein α is a regular term coefficient, the value between 1-5 is adjusted according to the optimization case.

In a further preferred embodiment, the regularization term coefficients α may be iteratively optimized nested with the independent variable weights, in particular, training a linear regression function includes:

For example, mutually nested iterations in a specific scenario specifically include the following steps:

1, acquiring original data in DCS, wherein the original data mainly comprises original measured values of three water levels, numerical values in the three water levels and operation loads (mainly electric loads) corresponding to the unit. Under the working conditions of 20% Pe, 30% Pe and 40% of the three load sections, various parameters are stabilized, and the data about 10min are taken out.

And 2, classifying the original data, namely respectively taking out three groups of corresponding data in corresponding time sequence according to the electric load data, and then taking out 50 data randomly and correspondingly as test data and taking the others as training data.

3: and normalizing the data, namely respectively performing normalization processing on the three measured values and the obtained value of the water level, and normalizing the normalized values to the interval of 0-100.

4 setting the regularization term hyperparameter α, initially at 1.

5: the superparameters of the gradient descent method are set, for example, the step δ is 0.01 and the initial value is 0. However, the selection of the step length and the initial value needs to fully consider the algorithm to prevent the local optimal solution from entering, so a grid searching method is adopted, namely, the initial value and the final value of the step length (delta is 0.01-0.7) are set, and the increment of the step length is 0.01; the initial value and the final value of the initial value are 0-50, and the step length is 1. Two nested loops are utilized to form a grid-like hyper-parameter distribution form, a group of hyper-parameters are formed at grid intersection points, optimization is carried out under the hyper-parameters, and a group of optimal hyper-parameters is selected from the group.

6: and (3) judging Mean Square Error (MSE) after obtaining an optimization result (each independent variable weight), namely judging that the mean square error of the test data is in a small range, otherwise, reselecting a hyper-parameter to perform optimization again, and performing the next step after the MSE meets the requirement (the MSE meets the requirement of selecting the model data with the minimum MSE in the optimization process and is the most global optimal).

7: determining whether the optimizing result satisfies a constraint condition, i.e. k₁+k₂+k₃And in the range of 1, otherwise, the regularization parameters are readjusted and the optimization is carried out again. And (5) outputting a result if the condition is met.

The regularization term is actually a constraint term of the model parameters, prevents over-fitting and under-fitting of the model, and is generally determined according to the physical characteristics and equipment properties of the system and engineering experience.

Further, in a preferred embodiment, each of the operation load segments has a difference, and specifically, the present invention also performs a targeted processing on each of the operation load segments, that is, data of each of the operation load segments is used to train a linear regression function and a corresponding regular term coefficient, so as to obtain a plurality of linear regression functions, where each of the linear regression functions corresponds to one of the operation load segments.

Specifically, the method of the present invention further comprises:

s05: determining an operation load section of the water level measurement data;

step S2 is specifically to input a plurality of water level measurement data as arguments to the linear regression function of the corresponding operation load segment.

Further, in this embodiment, the training for the linear regression function is also classified historical data, that is, the training step in the present invention specifically includes:

It can be understood that the invention correspondingly trains a linear regression function for each operation load segment, and is more targeted, since k searched in different load segments is adopted₁,k₂,k₃And correcting according to the load or the boiler evaporation capacity by using a smooth function, and revising the selection logic of the water level according to the parameters after perfection.

The following table shows the data obtained from the above procedure on a 600MW unit

	k₁	k₃	k₂
				20％Pe	0.457	0.328	0.197
30％Pe	0.368	0.501	0.131
				40％Pe	0.320	0.418	0.262

From the data in the above table, it can be seen that k is more severe due to low load level crossing₁，k₃The weight of (2) is larger, when the load is increased, the cross phenomenon is gradually weakened, the weights of the three coefficients are gradually balanced, and finally the weights may tend to be average. The problem of sudden direction change of the PID control algorithm caused in the crossing process of the steam drum water level after the method is adopted is solved.

Through the detailed description of the embodiment, the steam drum water level test value is optimized through a linear regression algorithm in machine learning, the optimized steam drum water level data can more accurately reflect the real situation of the boiler water level, the problem of steam drum water level control of a unit under a deep peak regulation working condition is further solved, favorable guarantee is provided for unit safety production, a better solution is provided for process control on the basis of guaranteeing unit safety operation, the variable load capacity and adaptability of the unit can be improved, the adjustment performance of each index of the unit is improved, and meanwhile, the requirements of an electric network on AGC and primary frequency modulation management and examination under the deep peak regulation condition are met. Therefore, the safety and the economical efficiency of the operation of the unit are improved, the environmental protection index of the unit is ensured, and the economic benefit and the social benefit of the thermal generator set participating in power grid examination are enhanced.

Based on the same inventive concept, another embodiment of the present invention provides a drum level determining apparatus in a subcritical boiler of a coal-fired unit, as shown in fig. 3, comprising:

an obtaining module 110, which obtains a plurality of water level measurement data of two sides of a boiler barrel in a subcritical boiler;

the input module 120 is used for inputting the water level measurement data serving as independent variables into a preset linear regression function to obtain steam drum water level optimization data;

the control module 130 inputs the drum water level optimization data to a control device, so that the control device adjusts the drum water level in the subcritical boiler according to the drum water level optimization data.

Based on the same reason, the steam drum water level determining device in the subcritical boiler of the coal-fired unit optimizes the steam drum water level test value through the linear regression algorithm in machine learning, the optimized steam drum water level data can reflect the real situation of the boiler water level more accurately, the problem of steam drum water level control of the unit under the deep peak regulation working condition is further solved, the favorable guarantee is provided for the safe production of the unit, a better solution is provided for process control on the basis of guaranteeing the safe operation of the unit, the variable load capacity and the adaptability of the unit can be improved, the adjusting performance of various indexes of the unit is improved, and meanwhile, the requirements of an electric network on AGC and primary frequency modulation management and examination under the deep peak regulation condition are met. Therefore, the safety and the economical efficiency of the operation of the unit are improved, the environmental protection index of the unit is ensured, and the economic benefit and the social benefit of the thermal generator set participating in power grid examination are enhanced.

Based on the same inventive concept, in an embodiment, the method further includes:

Based on the same inventive concept, in an embodiment, the drum level determining apparatus further includes:

Based on the same inventive concept, in an embodiment, the loss function constructing module includes:

Based on the same inventive concept, in an embodiment, the training module includes:

Based on the same inventive concept, in an embodiment, the linear regression function includes a plurality of linear regression functions, each linear regression function corresponds to a preset operation load segment, and the apparatus further includes:

An embodiment of the present invention further provides a specific implementation manner of an electronic device, which is capable of implementing all steps in the method in the foregoing embodiment, and referring to fig. 4, the electronic device specifically includes the following contents:

a processor (processor)601, a memory (memory)602, a communication interface (communications interface)603, and a bus 604;

the processor 601, the memory 602 and the communication interface 603 complete mutual communication through the bus 604;

the processor 601 is configured to call the computer program in the memory 602, and the processor executes the computer program to implement all the steps of the method in the above embodiments, for example, when the processor executes the computer program, the processor implements the following steps:

From the above description, the electronic device provided by the invention optimizes the steam drum water level test value through the linear regression algorithm in the machine learning, the optimized steam drum water level data can more accurately reflect the real situation of the boiler water level, further the problem of steam drum water level control of the unit under the deep peak regulation working condition is solved, the favorable guarantee is provided for the safe production of the unit, a better solution is provided for the process control on the basis of ensuring the safe operation of the unit, the variable load capacity and the adaptability of the unit can be improved, the regulation performance of each index of the unit is improved, and meanwhile, the requirements of the power grid on AGC and primary frequency modulation management and examination under the deep peak regulation condition are met. Therefore, the safety and the economical efficiency of the operation of the unit are improved, the environmental protection index of the unit is ensured, and the economic benefit and the social benefit of the thermal generator set participating in power grid examination are enhanced.

An embodiment of the present invention further provides a computer-readable storage medium capable of implementing all the steps of the method in the above embodiments, wherein the computer-readable storage medium stores thereon a computer program, which when executed by a processor implements all the steps of the method in the above embodiments, for example, the processor implements the following steps when executing the computer program:

From the above description, the computer-readable storage medium provided by the invention optimizes the steam drum water level test value through a linear regression algorithm in machine learning, and the optimized steam drum water level data can more accurately reflect the real situation of the boiler water level, so that the problem of steam drum water level control of a unit under a deep peak regulation working condition is solved, a favorable guarantee is provided for the safe production of the unit, a better solution is provided for process control on the basis of ensuring the safe operation of the unit, the variable load capacity and adaptability of the unit can be improved, the regulation performance of each index of the unit is improved, and the requirements of AGC and primary frequency modulation management and examination under the deep peak regulation condition of a power grid are met. Therefore, the safety and the economical efficiency of the operation of the unit are improved, the environmental protection index of the unit is ensured, and the economic benefit and the social benefit of the thermal generator set participating in power grid examination are enhanced.

The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the hardware + program class embodiment, since it is substantially similar to the method embodiment, the description is simple, and the relevant points can be referred to the partial description of the method embodiment. Although embodiments of the present description provide method steps as described in embodiments or flowcharts, more or fewer steps may be included based on conventional or non-inventive means. The order of steps recited in the embodiments is merely one manner of performing the steps in a multitude of orders and does not represent the only order of execution. When an actual apparatus or end product executes, it may execute sequentially or in parallel (e.g., parallel processors or multi-threaded environments, or even distributed data processing environments) according to the method shown in the embodiment or the figures. The terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, the presence of additional identical or equivalent elements in a process, method, article, or apparatus that comprises the recited elements is not excluded. For convenience of description, the above devices are described as being divided into various modules by functions, and are described separately. Of course, in implementing the embodiments of the present description, the functions of each module may be implemented in one or more software and/or hardware, or a module implementing the same function may be implemented by a combination of multiple sub-modules or sub-units, and the like. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form. The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks. As will be appreciated by one skilled in the art, embodiments of the present description may be provided as a method, system, or computer program product. Accordingly, embodiments of the present description may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, embodiments of the present description may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and so forth) having computer-usable program code embodied therein. The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment. In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of an embodiment of the specification. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction. The above description is only an example of the embodiments of the present disclosure, and is not intended to limit the embodiments of the present disclosure. Various modifications and variations to the embodiments described herein will be apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the embodiments of the present specification should be included in the scope of the claims of the embodiments of the present specification.

Claims

1. A method for determining drum water level in a subcritical boiler of a coal-fired unit is characterized by comprising the following steps:

2. The drum level determining method according to claim 1, further comprising:

and applying the training sample set to train the linear regression function.

3. The drum level determination method according to claim 2, wherein prior to training the linear regression function, the drum level determination method further comprises:

4. The drum level determining method according to claim 3, further comprising:

5. The drum level determination method of claim 3, wherein the constructing the loss function of the linear regression function comprises:

summing the first and second function structures to form the loss function.

6. The drum level determination method of claim 5, wherein the applying the training sample set to train the linear regression function comprises:

7. The method of claim 1, wherein the linear regression function comprises a plurality of linear regression functions, each linear regression function corresponding to a predetermined operational load segment, the method further comprising:

determining an operation load section of the water level measurement data;

8. The drum level determining method according to claim 7, further comprising:

9. A drum level determination apparatus in a subcritical boiler of a coal-fired unit, comprising:

10. The drum level determining apparatus according to claim 9, further comprising:

11. The drum level determination device according to claim 10, further comprising:

12. The drum level determining apparatus according to claim 11, further comprising:

13. The drum level determining apparatus according to claim 11, wherein the loss function constructing module comprises:

14. The drum level determination device of claim 13, wherein the training module comprises:

15. The drum level determining apparatus of claim 9, wherein the linear regression function comprises a plurality of linear regression functions, each linear regression function corresponding to a predetermined operation load segment, the apparatus further comprising:

16. The drum level determining apparatus according to claim 15, further comprising:

17. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor, when executing the program, carries out the steps of the drum level determination method according to any of claims 1 to 8.

18. A computer-readable storage medium, having stored thereon a computer program, characterized in that the computer program, when being executed by a processor, is adapted to carry out the steps of the drum level determination method as defined in any one of the claims 1 to 8.