Energy efficiency optimization method for coal-fired power generation unit

Abstract

The invention relates to the technical field of energy efficiency management and control, and particularly discloses a coal-fired generator set energy efficiency optimization method, which comprises the following steps: acquiring historical data, screening and cleaning the historical data, and eliminating invalid data; segmenting the data from which the invalid data is removed to obtain historical data of different steady-state working conditions; according to historical data of different steady-state working conditions, the boiler entering coal quality and the output enthalpy value are calculated in real time according to coal quality online detection equipment, and the boiler efficiency is calculated in a positive balance calculation mode; inverting related parameters based on the boiler efficiency and determining parameter reference values of the related parameters; and training a regression algorithm according to the parameter reference values of the related parameters. The invention provides an intelligent energy efficiency optimization scheme, an operator can find a boiler combustion optimization adjustment rising space point according to the reference value of the current working condition, the operation level of a unit is improved, and economic benefits are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of energy efficiency management and control, and in particular to a method for optimizing the energy efficiency of a coal-fired power generation unit. Background Art

[0002] Coal-fired power generation units are mainly composed of a combustion system (with a boiler as the core), a steam-water system (mainly composed of various pumps, feedwater heaters, condensers, pipelines, water-cooled walls, etc.), an electrical system (mainly steam turbine generators, main transformers, etc.), and a control system.

[0003] Under the background of deep peak regulation of thermal power units, when coal-fired power generating units are not operating under rated conditions, once the operating personnel lack experience in low-load operation, it will cause errors in the coal supply data. Large errors in the coal supply data will lead to inaccurate calculation of coal consumption for power generation, affecting the cost accounting and data analysis of the power plant. Summary of the invention

[0004] The purpose of the present invention is to provide a method for optimizing the energy efficiency of a coal-fired power generation unit to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A method for optimizing energy efficiency of a coal-fired power generation unit, the method comprising:

[0007] Obtain historical data, filter and clean the historical data, and eliminate invalid data;

[0008] The data after invalid data is removed is segmented to obtain historical data of different steady-state conditions;

[0009] For historical data of different steady-state conditions, the coal quality entering the furnace and the output enthalpy value are calculated in real time based on the online coal quality detection equipment, and the boiler efficiency is calculated using the positive balance calculation method;

[0010] Inversely analyze relevant parameters based on boiler efficiency and determine parameter reference values ​​of relevant parameters;

[0011] The regression algorithm will be trained based on the parameter benchmark values ​​of the relevant parameters.

[0012] As a further solution of the present invention: the step of segmenting the data after removing invalid data to obtain historical data of different steady-state conditions includes:

[0013] Query the time tag of the data and divide the data into winter heating data and non-heating data; among them, the winter heating data needs to deduct the heating network data to obtain the non-heating data;

[0014] Query the main steam flow rate of the data, divide the working conditions according to the main steam flow rate, and obtain the historical data of different steady-state working conditions;

[0015] The meaning of steady-state operating condition is that the load maintains the main steam flow rate oscillating within 1.5% of the rated operating condition within 30 minutes.

[0016] As a further solution of the present invention: the steps of calculating the boiler efficiency using a positive balance calculation method by using historical data of different steady-state conditions and calculating the coal quality entering the furnace and the output enthalpy in real time according to the coal quality online detection equipment include:

[0017] Calculate the total heat input to the boiler Qtotal, Qtotal = Qnet × B, Qnet is the calorific value of the raw coal, B is the amount of coal fed;

[0018] Calculate the effective utilization heat Q1, Q1 = D1 × Δh1 + D2 × Δh2 + D3 × Δh3 + ΔQ; where D1 is the superheated steam flow rate, Δh1 is the difference between the superheater outlet enthalpy and the economizer inlet enthalpy, D2 is the reheated steam flow rate, Δh2 is the difference between the reheater outlet enthalpy and the inlet enthalpy, D3 is the reheater desuperheating water flow rate, Δh3 is the difference between the reheater outlet enthalpy and the reheater desuperheating water flow rate, ΔQ is the sum of the enthalpies of soot blowing, tracing, and blowdown;

[0019] Use the reverse balance calculation method to verify the positive balance calculation results and calibrate the coal feed rate; the calculation process is:

[0020] Q total reaction = Q1 + Q2 + Q3 + Q4 + Q5 + Q6, of which the effective utilization heat is Q1, the smoke loss is Q2, the chemical incomplete combustion loss is Q3, the mechanical incomplete combustion loss is Q4, the heat dissipation loss is Q5, and the ash physical heat loss is Q6;

[0021] The calibration logic is:

[0022] If Qtotal reverse ≈ Qtotal is established, it proves that the positive balance calculation result is correct; under the premise that Qtotal reverse ≈ Qtotal is established, if Qtotal reverse / B ≈ Qnet, then the verification result shows that the coal feeding amount is correct. If Qtotal reverse / B is obviously larger than the error range of Qnet, it can be concluded that the coal feeding amount B is too small, otherwise the coal feeding amount is too large; under the premise that Qtotal reverse ≈ Qtotal, the actual coal feeding amount is obtained through Qtotal reverse / Qnet to correct the original coal feeding.

[0023] As a further solution of the present invention: the step of inverting relevant parameters based on boiler efficiency and determining parameter reference values ​​of relevant parameters comprises:

[0024] The boiler efficiency η under different steady-state conditions is discretized into intervals of 2%, and the discrete interval of 2% of the higher data peak is used as the target interval for operation optimization adjustment under the condition;

[0025] Use heat maps to mine parameters whose correlation coefficients with boiler efficiency reach preset thresholds;

[0026] The association model uses an iterative method of layer-by-layer search to find the relationship between item sets in the database to form rules. The process consists of connection and pruning. There are multiple combinations of association parameters of boiler efficiency η under various working conditions. Each combination is regarded as an item set. If an item set is frequent, then the subset of the item set is also frequent. Frequent item sets are mined, and strong association rules are generated from these item sets to obtain item sets that meet the requirements.

[0027] Taking the target range of boiler efficiency under this operating condition as the benchmark value, the model is used to obtain the parameter benchmark values ​​of the parameters in the same period as the boiler efficiency benchmark value.

[0028] As a further solution of the present invention: the step of using the thermal map to mine parameters whose correlation coefficient with boiler efficiency reaches a preset threshold value includes:

[0029] Assume there are n samples and p variables in total; for the i-th variable, its value is x1 i, x2i, ..., xn i, and for the j-th variable, its value is y1 j, y2j, ..., ynj;

[0030] Calculate the mean value of each variable, recorded as xi and yj respectively, which can be calculated by dividing the cumulative value of n samples by the number of samples;

[0031] Calculate the standard deviation of each variable, denoted as σi and σj, which can be calculated by (∑(xi-xi)2 / (n-1))^0.5;

[0032] Calculate the covariance, denoted as sigma(i,j), which can be calculated using the following formula:sigma(i,j)=∑(xi-xi)(yj-yj) / (n-1)where xi, xi, yj, yj are the values ​​and averages of the i-th and j-th variables respectively;

[0033] The Pearson correlation coefficient, denoted as ρ(i,j), can be calculated using the following formula: ρ(i,j) = s igma(i,j) / (σiσj) where s igma(i,j) is the covariance between the ith and jth variables, σi and σj are the standard deviations of the ith and jth variables respectively;

[0034] The strength of the correlation between variables can be judged based on the size of the Pearson correlation coefficient. Usually, the following indicators can be used: when |ρ(i,j)|>0.8, it means that the correlation between the two variables is very strong;

[0035] When 0.6≤|ρ(i,j)|≤0.8, it means that the correlation between the two variables is relatively strong;

[0036] When 0.3≤|ρ(i,j)|≤0.6, it means that the correlation between the two variables is general;

[0037] When |ρ(i,j)|<0.3, it means that the correlation between the two variables is weak or there is no obvious correlation;

[0038] Draw a heat map based on the calculated correlation coefficient and set the corresponding color label.

[0039] As a further solution of the present invention: the step of training the regression algorithm according to the parameter reference values ​​of the relevant parameters includes:

[0040] If it is linear regression, the model is converted into a matrix form y=Xβ+ε, where y represents a vector of n rows and 1 column, X is a matrix of n rows and p+1 columns, β represents the model coefficient, and ε represents the error term;

[0041] Our goal is to find the best β by the least squares method so that the sum of squared errors is minimized:

[0042] min(ε'ε)=min(y-Xβ)'(y-Xβ) Expand the above formula, you can get: min(ε'ε)=min(y'y-2y'Xβ+β'X'Xβ) Take the derivative of β, you get: d(ε'ε) / dβ=-2X'y+2X'Xβ=0;

[0043] The solution is: β = (X'X)-1X'y; where X' represents the transpose of X, and (X'X)-1 represents the inverse matrix of X'X; finally, by calculating this formula, we can get the value of the coefficient β, and thus get the linear regression model fitting curve;

[0044] If it is a polynomial regression, the model is fitted by a p-order polynomial equation y=β0+β1x+β2x^2+...+βpx^p, that is, the higher-order terms of the independent variable x are introduced into the fitting equation; similarly, we convert the model into a matrix form, that is: y=Xβ+ε; where y represents a vector of n rows and 1 column, X is a matrix of n rows and p+1 columns, β represents the model coefficient, and ε represents the error term; our goal is to find the best β through the least squares method to minimize the sum of squares of errors: min(ε'ε)=min(y-Xβ)'(y-Xβ); expanding the above formula, we can get: min(ε'ε)=min(y'y-2y'Xβ+β'X'Xβ) Taking the derivative of β, we get: d(ε'ε) / dβ=-2X'y+2X'Xβ=0;

[0045] Solve: β = (X'X) - 1X'y;

[0046] Among them, X' represents the transpose of X, and (X'X)-1 represents the inverse matrix of X'X; finally, by calculating this formula, we can get the value of coefficient β, and thus get the fitting curve of the polynomial regression model.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] 1. Operators can find the upward space for optimizing boiler combustion adjustment based on the baseline value of the current operating conditions, thereby improving the unit's operating level and increasing economic benefits.

[0049] 2. Standardize the operations of operators, reduce the differences between different teams, and assist operators with less experience to improve their operation level.

[0050] 3. The calculation of coal supply is more accurate, which makes some key indicators such as boiler efficiency, power generation coal consumption more accurate, and cost accounting more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention.

[0052] Figure 1 Flowchart of the method for optimizing the energy efficiency of coal-fired power generation units. DETAILED DESCRIPTION

[0053] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0054] Figure 1 The flowchart of the method for optimizing the energy efficiency of a coal-fired power generation unit is as follows. In an embodiment of the present invention, a method for optimizing the energy efficiency of a coal-fired power generation unit includes:

[0055] Step S100: Obtain historical data, filter and clean the historical data, and remove invalid data;

[0056] Step S200: segmenting the data after invalid data is removed to obtain historical data of different steady-state conditions;

[0057] Step S300: For historical data of different steady-state conditions, the coal quality entering the furnace and the output enthalpy value are calculated in real time according to the online coal quality detection equipment, and the boiler efficiency is calculated using a positive balance calculation method;

[0058] Step S400: inverting relevant parameters based on boiler efficiency and determining parameter reference values ​​of the relevant parameters;

[0059] Step S500: training a regression algorithm according to parameter benchmark values ​​of relevant parameters.

[0060] The application scenarios of the above energy efficiency optimization solution are described as follows:

[0061] (1) The present invention is applied to the combustion optimization part of the boiler of a coal-fired power plant.

[0062] (2) According to the law of conservation of energy, the total heat of coal entering the furnace is equal to the total heat output. The total heat of coal entering the furnace is determined by two factors, namely the calorific value Qnet (internal energy) of coal quality and the amount of coal fed B; the total heat output includes boiler efficiency η, smoke exhaust loss Q2, chemical incomplete combustion loss Q3, mechanical incomplete combustion loss Q4, heat dissipation loss Q5, and ash physical heat loss Q6. Q1 and Q2 can be calculated through the measurement point data, and the sum of the two accounts for about 95%-97% of the total; Q3-Q6 lack effective automatic monitoring means. Due to the small proportion, they are usually estimated based on the design values ​​under different working conditions in the operation manual, or manually sampled and tested, which has certain errors.

[0063] (3) Qtotal = Qnet×B = Q1+Q2+Q3+Q4+Q5+Q6, to determine boiler efficiency η = Q1 / Qtotal, boiler efficiency is an important indicator for determining boiler combustion effects, the higher the boiler efficiency, the better the economy. Use η = Q1 / (Qnet×B) as a positive balance calculation method, and use η = Q1 / (Q1+Q2+Q3+Q4+Q5+Q6) as a reverse balance calculation method; because coal quality is complex and there is no suitable testing equipment, the traditional method is to use the reverse balance calculation method to calculate boiler efficiency.

[0064] (4) The process of coal entering the furnace, burning, releasing heat, transferring heat to water to become steam, exhausting flue gas and generating other heat losses requires a certain amount of time, and the coal feeding signal is a pre-signal. Therefore, increasing or decreasing the coal feeding amount will result in unequal input and output heat at the same time. In order to obtain accurate boiler efficiency calculation results under various operating conditions, it is necessary to wait for the coal feeding amount to remain stable, the heat exchange to be sufficient, and the operating conditions to be stable before conducting data optimization.

[0065] (5) In addition to using large indicators such as boiler efficiency to monitor boiler combustion status, there are also multiple main small indicator parameters of the boiler. Specific problem points can be found based on the small indicator parameters. Small indicator parameters are highly coupled, so it is difficult for the performance of each indicator to reach the most ideal state at the same time. The present invention uses a relatively high discrete interval of boiler efficiency η under different steady-state conditions as a benchmark value, and uses an association model to mine the benchmark value of the small indicator.

[0066] (6) The present invention uses an online coal quality detection device to calculate the quality of the coal entering the furnace in real time, and uses a positive balance to calculate the boiler efficiency η, thereby getting rid of the traditional reverse balance method of calculating the effective use of heat by the boiler. This calculation method is simple, fast, requires little computation, and has little error. Using the reverse balance calculation method to verify the positive balance calculation result can effectively calibrate the coal feed error.

[0067] In a preferred embodiment of the technical solution of the present invention, a positive balance calculation scheme is provided as follows:

[0068] (1) Screen and clean historical data to eliminate invalid data.

[0069] (2) To deal with different seasons, we need to mine the benchmark values ​​of each major parameter and divide the data into winter heating data and non-heating data in other seasons. Winter heating data requires deducting the heating network data to obtain the non-heating data.

[0070] (3) The performance of boilers under different working conditions is different. The main steam flow rate is divided into multiple working conditions. The load maintains the main steam flow rate within 1.5% of the rated working condition within 30 minutes (the time and interval can be adjusted according to the specific situation) as the steady-state working condition. The boiler efficiency η of the boiler under different steady-state working conditions is calculated.

[0071] (4) Get rid of the traditional reverse balance method to calculate boiler efficiency. Use coal quality online detection equipment to calculate the coal quality and output enthalpy in real time, and use positive balance to calculate boiler efficiency η. The calculation method is as follows:

[0072] (4.1) Calculate the total heat input to the boiler Qtotal, Qtotal = Qnet × B, Qnet is the calorific value of the raw coal, and B is the coal feed rate.

[0073] (4.2) Calculate the effective utilization heat Q1, Q1 = D1 × Δh1 + D2 × Δh2 + D3 × Δh3 + ΔQ; where D1 is the superheated steam flow, Δh1 is the difference between the superheater outlet enthalpy and the economizer inlet enthalpy, D2 is the reheated steam flow, Δh2 is the difference between the reheater outlet enthalpy and the inlet enthalpy, D3 is the reheater desuperheating water flow, Δh3 is the difference between the reheater outlet enthalpy and the reheater desuperheating water flow, and ΔQ is the sum of the enthalpy values ​​of soot blowing, heating, and blowdown. This calculation method is simple, fast, has a small amount of calculation, and a small error.

[0074] Use the reverse balance calculation method to verify the positive balance calculation result and calibrate the coal feed rate. The method is as follows:

[0075] (4.3)Qtotal heat = Q1+Q2+Q3+Q4+Q5+Q6, of which the effectively utilized heat is Q1, the smoke exhaust loss is Q2, the chemical incomplete combustion loss is Q3, the mechanical incomplete combustion loss is Q4, the heat dissipation loss is Q5, and the ash physical heat loss is Q6.

[0076] (4.4) Under steady-state conditions, the heat output of the boiler is equal to the heat input and output due to the law of conservation of energy. If Qtotalreverse≈Qtotal holds, the positive balance calculation result is correct. Under the premise that Qtotalreverse≈Qtotal holds, if Qtotalreverse / B≈Qnet, the coal feeding amount obtained by verification is correct. If Qtotalreverse / B is significantly greater than the error range of Qnet, it can be concluded that the coal feeding amount B is too small, otherwise the coal feeding amount is too large. Under the premise that Qtotalreverse≈Qtotal, the actual coal feeding amount is obtained through Qtotalreverse / Qnet to correct the original coal feeding amount.

[0077] (5) The boiler efficiency η under different steady-state conditions is discretized into intervals of 2%, and the discrete interval of the higher data peak of 2% (the time and interval can be adjusted according to the specific situation) is used as the target of operation optimization adjustment under this condition.

[0078] (6) The thermal system of coal-fired units is coupled, and various indicators are mutually constrained and cannot be completely decoupled independently. The thermal map is used to mine parameters that are strongly correlated with boiler efficiency η. The implementation method is as follows:

[0079] (6.1) Assume there are n samples and p variables. For the i-th variable, its value is x1 i, x2i, ..., xn i, and for the j-th variable, its value is y1 j, y2j, ..., ynj.

[0080] (6.2) Calculate the mean value of each variable, denoted as xi and yj respectively, which can be calculated by dividing the cumulative value of n samples by the number of samples.

[0081] (6.3) Calculate the standard deviation of each variable, denoted as σi and σj respectively, which can be calculated by (∑(xi-xi)2 / (n-1))^0.5.

[0082] (6.4) Calculate the covariance, denoted as sigma(i,j), which can be calculated using the following formula:sigma(i,j)=∑(xi-xi)(yj-yj) / (n-1)where xi, xi, yj, yj are the values ​​and averages of the i-th and j-th variables respectively.

[0083] (6.5) The Pearson correlation coefficient, denoted as ρ(i,j), can be calculated using the following formula: ρ(i,j) = sigma(i,j) / (σiσj) where sigma(i,j) is the covariance between the ith and jth variables, and σi and σj are the standard deviations of the ith and jth variables, respectively.

[0084] (6.6) The strength of the correlation between variables can be judged based on the size of the Pearson correlation coefficient. Usually, the following indicators can be used: When |ρ(i,j)|>0.8, it means that the correlation between the two variables is very strong;

[0085] When 0.6≤|ρ(i,j)|≤0.8, it means that the correlation between the two variables is relatively strong;

[0086] When 0.3≤|ρ(i,j)|≤0.6, it means that the correlation between the two variables is general;

[0087] When |ρ(i,j)|<0.3, it means that the correlation between the two variables is weak or there is no obvious correlation.

[0088] (6.7) Draw a heat map based on the calculated correlation coefficient and set the corresponding color code. The darker the color, the greater the correlation, and the lighter the color, the smaller the correlation or irrelevant, so as to obtain the parameters with stronger correlation.

[0089] (7) The association model uses an iterative method of layer-by-layer search to find the relationship between item sets in the database to form rules. The process consists of connection and pruning. There are many combinations of association parameters of boiler efficiency η under various operating conditions. Each combination is regarded as an item set. If an item set is frequent, then the subset of the item set is also frequent. Frequent item sets are mined, and strong association rules are generated from these item sets to obtain item sets that meet the requirements.

[0090] (8) Taking the adjustment target interval of the boiler efficiency η under this operating condition as the benchmark value, the model is used to obtain the parameter benchmark value that is strongly correlated with the boiler efficiency η benchmark value in the same period.

[0091] (9) According to the scatter point distribution of the benchmark values ​​of the operating parameters under different working conditions, it is determined whether to use linear regression or polynomial regression, and multiple working condition benchmark values ​​of the operating parameters are used to form a regression algorithm, which is formed as follows:

[0092] (9.1) If it is linear regression, the model is converted into a matrix form y = Xβ + ε, where y represents a vector of n rows and 1 column, X is a matrix of n rows and p + 1 columns, β represents the model coefficient, and ε represents the error term.

[0093] Our goal is to find the best β by the least squares method so that the sum of squared errors is minimized:

[0094] min(ε'ε)=min(y-Xβ)'(y-Xβ) Expand the above formula, you can get: min(ε'ε)=min(y'y-2y'Xβ+β'X'Xβ) Take the derivative of β, you get: d(ε'ε) / dβ=-2X'y+2X'Xβ=0;

[0095] The solution is: β = (X'X)-1X'y. Among them, X' represents the transpose of X, and (X'X)-1 represents the inverse matrix of X'X. Finally, by calculating this formula, we can get the value of coefficient β, and thus get the linear regression model fitting curve.

[0096] (9.2) If it is a polynomial regression, the model is fitted by a p-order polynomial equation y=β0+β1x+β2x^2+...+βpx^p, that is, the higher-order terms of the independent variable x are introduced into the fitting equation. Similarly, we convert the model into a matrix form, that is: y=Xβ+ε; where y represents a vector of n rows and 1 column, X is a matrix of n rows and p+1 columns, β represents the model coefficient, and ε represents the error term. Our goal is to find the optimal β through the least squares method so that the sum of squared errors is minimized: min(ε'ε)=min(y-Xβ)'(y-Xβ); Expanding the above formula, we can get: min(ε'ε)=min(y'y-2y'Xβ+β'X'Xβ) Taking the derivative of β, we get: d(ε'ε) / dβ=-2X'y+2X'Xβ=0;

[0097] Solve: β = (X'X) - 1X'y;

[0098] Among them, X' represents the transpose of X, and (X'X)-1 represents the inverse matrix of X'X. Finally, by calculating this formula, we can get the value of coefficient β, and thus get the fitting curve of the polynomial regression model.

[0099] Through the above fitting curve prediction, the benchmark values ​​of non-steady-state conditions or conditions that did not occur in the historical data are finally predicted.

[0100] In an example of the technical solution of the present invention, due to the complexity of coal quality, differences in operating habits of operators in each team, and other problems in a coal-fired power generation unit, the levels of boiler combustion optimization adjustment vary. The present invention mines the benchmark values ​​of the main parameter indicators of the boiler under different steady-state conditions based on historical data optimization and association models, formulates optimization standards for boiler combustion adjustment, and clarifies the optimization goals.

[0101] Calibrate the coal feeding amount data. Since the coal feeding amount data of the coal-fired power generation unit is weighed by a belt scale or calculated by current fitting, the data has a large error. The present invention uses the boiler positive balance calculation and reverse balance verification method to calibrate the coal feeding amount and reduce the data error of the coal feeding amount.

[0102] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for optimizing energy efficiency of a coal-fired power generation unit, characterized in that: The method comprises: Obtain historical data, filter and clean the historical data, and eliminate invalid data; The data after invalid data is removed is segmented to obtain historical data of different steady-state conditions; For historical data of different steady-state conditions, the coal quality entering the furnace and the output enthalpy value are calculated in real time based on the online coal quality detection equipment, and the boiler efficiency is calculated using the positive balance calculation method; Inversely analyze relevant parameters based on boiler efficiency and determine parameter reference values ​​of relevant parameters; The regression algorithm will be trained based on the parameter benchmark values ​​of the relevant parameters.

2. The method for optimizing energy efficiency of a coal-fired power generation unit according to claim 1, characterized in that: The step of dividing the data after removing invalid data to obtain historical data of different steady-state conditions includes: Query the time tag of the data and divide the data into winter heating data and non-heating data; among them, the winter heating data needs to deduct the heating network data to obtain the non-heating data; Query the main steam flow rate of the data, divide the working conditions according to the main steam flow rate, and obtain the historical data of different steady-state working conditions; The meaning of steady-state operating condition is that the load maintains the main steam flow rate oscillating within 1.5% of the rated operating condition within 30 minutes.

3. The method for optimizing energy efficiency of a coal-fired power generation unit according to claim 1, characterized in that: The steps of calculating the boiler efficiency by using a positive balance calculation method and calculating the coal quality and output enthalpy in real time according to the historical data of different steady-state conditions using the coal quality online detection equipment include: Calculate the total heat input to the boiler Qtotal, Qtotal = Qnet × B, Qnet is the calorific value of the raw coal, B is the amount of coal fed; Calculate the effective utilization heat Q1, Q1 = D1 × Δh1 + D2 × Δh2 + D3 × Δh3 + ΔQ; where D1 is the superheated steam flow rate, Δh1 is the difference between the superheater outlet enthalpy and the economizer inlet enthalpy, D2 is the reheated steam flow rate, Δh2 is the difference between the reheater outlet enthalpy and the inlet enthalpy, D3 is the reheater desuperheating water flow rate, Δh3 is the difference between the reheater outlet enthalpy and the reheater desuperheating water flow rate, ΔQ is the sum of the enthalpies of soot blowing, tracing, and blowdown; Use the reverse balance calculation method to verify the positive balance calculation results and calibrate the coal feed rate; the calculation process is: Q total reaction = Q1 + Q2 + Q3 + Q4 + Q5 + Q6, of which the effective utilization heat is Q1, the smoke loss is Q2, the chemical incomplete combustion loss is Q3, the mechanical incomplete combustion loss is Q4, the heat dissipation loss is Q5, and the ash physical heat loss is Q6; The calibration logic is: If Qtotal reverse ≈ Qtotal is established, it proves that the positive balance calculation result is correct; under the premise that Qtotal reverse ≈ Qtotal is established, if Qtotal reverse / B ≈ Qnet, then the verification result shows that the coal feeding amount is correct. If Qtotal reverse / B is obviously larger than the error range of Qnet, it can be concluded that the coal feeding amount B is too small, otherwise the coal feeding amount is too large; under the premise that Qtotal reverse ≈ Qtotal, the actual coal feeding amount is obtained through Qtotal reverse / Qnet to correct the original coal feeding.

4. The method for optimizing energy efficiency of a coal-fired power generation unit according to claim 1, characterized in that: The step of inverting relevant parameters based on boiler efficiency and determining parameter reference values ​​of relevant parameters comprises: The boiler efficiency η under different steady-state conditions is discretized into intervals of 2%, and the discrete interval of 2% of the higher data peak is used as the target interval for operation optimization adjustment under the condition; Use heat maps to mine parameters whose correlation coefficients with boiler efficiency reach preset thresholds; The association model uses an iterative method of layer-by-layer search to find the relationship between item sets in the database to form rules. The process consists of connection and pruning. There are multiple combinations of association parameters of boiler efficiency η under various working conditions. Each combination is regarded as an item set. If an item set is frequent, then the subset of the item set is also frequent. Frequent item sets are mined, and strong association rules are generated from these item sets to obtain item sets that meet the requirements. Taking the target range of boiler efficiency under this operating condition as the benchmark value, the model is used to obtain the parameter benchmark values ​​of the parameters in the same period as the boiler efficiency benchmark value.

5. The method for optimizing energy efficiency of a coal-fired power generation unit according to claim 4, characterized in that: The step of using the heat map to mine parameters whose correlation coefficient with boiler efficiency reaches a preset threshold value includes: Assume there are n samples and p variables in total; for the i-th variable, its value is x1 i, x2 i, ..., xni, for the j-th variable, its value is y1 j, y2j, ..., ynj; Calculate the mean value of each variable, recorded as xi and yj respectively, which can be calculated by dividing the cumulative value of n samples by the number of samples; Calculate the standard deviation of each variable, denoted as σi and σj, which can be calculated by (∑(xi-xi)2 / (n-1))^0.5; Calculate the covariance, denoted as sigma(i,j), which can be calculated using the following formula: sigma(i,j) = ∑(xi-xi)(yj-yj) / (n-1) where xi, xi, yj, yj are the values ​​and averages of the i-th and j-th variables respectively; The Pearson correlation coefficient, denoted as ρ(i,j), can be calculated using the following formula: ρ(i,j) = sigma(i,j) / (σiσj) where sigma(i,j) is the covariance between the ith and jth variables, σi and σj are the standard deviations of the ith and jth variables respectively; The strength of the correlation between variables can be judged based on the size of the Pearson correlation coefficient. Usually, the following indicators can be used: when |ρ(i,j)|>0.8, it means that the correlation between the two variables is very strong; When 0.6≤|ρ(i,j)|≤0.8, it means that the correlation between the two variables is relatively strong; When 0.3≤|ρ(i,j)|≤0.6, it means that the correlation between the two variables is general; When |ρ(i,j)|<0.3, it means that the correlation between the two variables is weak or there is no obvious correlation; Draw a heat map based on the calculated correlation coefficient and set the corresponding color label.

6. The method for optimizing energy efficiency of a coal-fired power generation unit according to claim 1, characterized in that: The step of training the regression algorithm according to the parameter reference values ​​of the relevant parameters comprises: If it is linear regression, the model is converted into a matrix form y=Xβ+ε, where y represents a vector of n rows and 1 column, X is a matrix of n rows and p+1 columns, β represents the model coefficient, and ε represents the error term; Our goal is to find the best β by the least squares method so that the sum of squared errors is minimized: min(ε'ε)=min(y-Xβ)'(y-Xβ) Expand the above formula, you can get: min(ε'ε)=min(y'y-2y'Xβ+β'X'Xβ) Take the derivative of β, you get: d(ε'ε) / dβ=-2X'y+2X'Xβ=0; The solution is: β = (X'X)-1X'y; where X' represents the transpose of X, and (X'X)-1 represents the inverse matrix of X'X; finally, by calculating this formula, we can get the value of the coefficient β, and thus get the linear regression model fitting curve; If it is a polynomial regression, the model is fitted by a p-order polynomial equation y=β0+β1x+β2x^2+...+βpx^p, that is, the higher-order terms of the independent variable x are introduced into the fitting equation; similarly, we convert the model into a matrix form, that is: y=Xβ+ε; where y represents a vector of n rows and 1 column, X is a matrix of n rows and p+1 columns, β represents the model coefficient, and ε represents the error term; our goal is to find the best β through the least squares method to minimize the sum of squares of errors: min(ε'ε)=min(y-Xβ)'(y-Xβ); expanding the above formula, we can get: min(ε'ε)=min(y'y-2y'Xβ+β'X'Xβ) Taking the derivative of β, we get: d(ε'ε) / dβ=-2X'y+2X'Xβ=0; Solve: β = (X'X) - 1X'y; Among them, X' represents the transpose of X, and (X'X)-1 represents the inverse matrix of X'X; finally, by calculating this formula, we can get the value of coefficient β, and thus get the fitting curve of the polynomial regression model.

Images