Thermal power unit operation optimization method, device and equipment and storage medium

Abstract

The present application relates to the technical field of operation optimization of thermal power generating units, and specifically discloses a thermal power generating unit operation optimization method, device, equipment and storage medium, which comprises determining a target optimization function of a thermal power generating unit and a target constraint condition corresponding to the target optimization function; obtaining a nonlinear mapping function, converting the target optimization function into a representation of the nonlinear mapping function, and obtaining a target mapping function; determining a current initial operation parameter of the target mapping function within the target constraint condition based on a simulated annealing algorithm; iteratively optimizing the current initial operation parameter by using a trust region interior point method to obtain a current target optimization parameter of the target mapping function; updating a current operation parameter of the thermal power generating unit based on the current target optimization parameter to obtain a current target operation parameter; and controlling the thermal power generating unit to operate based on the current target operation parameter. The present application optimizes the operation data of the thermal power generating unit, thereby improving the operation accuracy of the thermal power generating unit.

Description

Technical Field

[0001] This invention belongs to the field of thermal power generating unit operation optimization technology, specifically relating to thermal power generating unit operation optimization methods, devices, equipment, and storage media. Background Technology

[0002] A thermal power plant, or coal-fired power plant for short, is a factory that uses combustible materials (such as coal) as fuel to produce electricity. Its basic production process is as follows: when fuel is burned, it heats water to generate steam, converting the chemical energy of the fuel into heat energy. The steam pressure drives the turbine to rotate, converting the heat energy into mechanical energy. Then, the turbine drives the generator to rotate, converting the mechanical energy into electrical energy.

[0003] In the thermal power generation sector, efficient and stable unit operation is crucial for energy utilization and power supply. With the increasing demands for energy conservation, emission reduction, and refined management in the power industry, optimizing unit data during operation has become a key direction for industry development. Summary of the Invention

[0004] To address the shortcomings of existing technologies, according to one aspect of this application, a method for optimizing the operation of thermal power units is disclosed, the method comprising:

[0005] Determine the objective optimization function of the thermal power unit and the objective constraints corresponding to the objective optimization function;

[0006] Obtain the nonlinear mapping function, transform the target optimization function into a representation of the nonlinear mapping function, and obtain the target mapping function;

[0007] The current initial operating parameters of the target mapping function within the target constraints are determined based on the simulated annealing algorithm;

[0008] The current initial operating parameters are iteratively optimized using the in-trust region point method to obtain the current target optimization parameters of the target mapping function;

[0009] The current operating parameters of the thermal power unit are updated based on the current target optimization parameters to obtain the current target operating parameters;

[0010] Control the thermal power unit to operate based on the current target operating parameters;

[0011] The nonlinear mapping function is expressed based on formula (13):

[0012] minf(x), x∈R n stc(x) = 0 or c(x) ≤ 0 (13);

[0013] In the formula, min represents minimization;

[0014] f(x) is the target mapping function to be optimized;

[0015] x is the optimization variable;

[0016] R n To optimize the domain of the variables, n is the number of feature variables;

[0017] st represents the constraint condition;

[0018] c(x) = 0 is an equality constraint;

[0019] c(x)≤0 is an inequality constraint.

[0020] In some embodiments, the target optimization function includes a combustion optimization function, and determining the target optimization function of the thermal power unit and the target constraints corresponding to the target function includes:

[0021] The combustion optimization function of the thermal power unit is determined based on the combustion optimization objective of the thermal power unit and the combustion influence characteristics of the combustion optimization objective;

[0022] The combustion constraints corresponding to the combustion optimization function are determined based on the combustion operation mechanism of the thermal power unit, the combustion safety efficiency balance boundary, and historical combustion data.

[0023] The combustion optimization function is expressed based on formula (1):

[0024]

[0025] In the formula, max means maximization; m unit_coal The steam production per ton of coal is calculated from the total primary air volume (m³). wind1 Total secondary air volume (m³) wind2 Main water supply flow rate (m) w Water-to-coal ratio (a), lower calorific value of coal fed into the furnace (Q) net.ar The feature variables are obtained after normalization and training.

[0026] The combustion constraint conditions are expressed based on formulas (2), (4), and (5):

[0027]

[0028]

[0029] In the formula, O air 'Upper is a training model for flue gas oxygen content.' Oair and Lower Oair These are the upper and lower limits obtained from historical data.

[0030] In some embodiments, the objective optimization function includes a cold-end optimization function, and determining the objective optimization function of the thermal power unit and the objective constraints corresponding to the objective function includes:

[0031] The cold-end optimization function of the thermal power unit is determined based on the cold-end optimization objective and the cold-end influence characteristics of the cold-end optimization objective.

[0032] The cold-end constraint conditions of the thermal power unit are determined based on the cold-end operation mechanism, cold-end safety efficiency balance boundary and historical cold-end data.

[0033] The cold-end optimization function is expressed based on formula (6):

[0034]

[0035] In the formula To select the main steam flow rate m st Main steam temperature T st Main steam pressure P st High-temperature reheat steam temperature T rest and condenser back pressure p con The power generation per ton of steam obtained after normalization training, P pump Power consumption of the circulating pump;

[0036] The cold end constraint conditions are expressed based on formulas (7)-(12):

[0037]

[0038] I pump =n L I L +n H I H (9);

[0039] x L +x H =1,x L ∈{0,1},x H ∈{0,1} (10);

[0040] n L ≤2x L ,n L ∈{0,1,2}, (11)

[0041] n H ≤2x H ,n H ∈{0,1,2} (12).

[0042] In the formula, g represents the training model of condenser vacuum. I is the normalized circulating water feedwater temperature. pump and U pump For the circulating pump current and voltage, n L and n H I represents the number of low-speed pumps and high-speed pumps in operation, respectively. L and I H For low-speed and high-speed current of circulating water pumps, x L and x H This indicates the operating status of the high / low speed pump; 0 / 1 represents low speed / high speed. (error) con This represents the deviation between the condenser vacuum level at the previous moment and the condenser vacuum level at the current moment in the training model.

[0043] In some embodiments, determining the current initial operating parameters of the target mapping function within the target constraints based on the simulated annealing algorithm includes:

[0044] Determine the target optimization variables and iterative influence parameters of the objective optimization function;

[0045] Apply random perturbations to the target optimization variables;

[0046] The objective constraints corresponding to the objective mapping function are constrained based on the penalty function method to obtain an unconstrained optimization function.

[0047] Based on the iterative influence parameters, determine the iterative termination condition for the target optimization variable;

[0048] Obtain update criteria;

[0049] Based on the update criteria, the target optimization variables are updated iteratively;

[0050] When the update iteration of the target optimization variable satisfies the iteration termination condition, the update variable corresponding to the iteration termination condition is used as the current initial running parameter of the target mapping function;

[0051] The random perturbation is represented by formula (20):

[0052] x k+1 =x k +Δx,Δx~N(0,σ 2 (20);

[0053] In the formula x k+1 x k Let denot be the optimization variables for the (k+1)th and kth iterations, respectively, and let Δx represent the random perturbation, where Δx ~ N(0, σ). 2 ) indicates that Δx follows a mean of 0 and a variance of σ. 2 It follows a normal distribution.

[0054] The unconstrained optimization model is expressed based on formula (21):

[0055] E(x)=f(x)+λ(∑max(0,g i (x)) 2 +∑h i (x) 2 ) (twenty one);

[0056] In the formula, E(x) represents the penalty function, f(x) represents the original objective function, and λ(Σmax(0,g)) represents the objective function. i (x) 2 )+Σh i (x) 2 ) represents the penalty function constraint, where λ is the penalty factor, and max represents choosing 0 and g. i (x) 2 The larger one, g i (x) 2 Let h represent the inequality constraints of the original problem. i (x) 2 The equality constraints of the original problem are expressed.

[0057] The update criterion is expressed based on formula (22):

[0058]

[0059] In the formula, P represents the probability of accepting the new optimized solution, and E(x) k+1 E(x) represents the penalty function for the (k+1)th and kth cycles, and T is the current annealing temperature.

[0060] In some embodiments, when the target constraint is an equality constraint, the step of iteratively optimizing the current initial running parameters using the in-trust region point method to obtain the current target optimization parameters of the target mapping function includes:

[0061] Obtain the iteration variables for the point method within the trust region;

[0062] Based on the trust region sequential quadratic programming method, the target step size in the approximate objective function of the quadratic model is determined, wherein the target step size is within the trust region;

[0063] The current target optimization parameters and the iteration variables are iterated based on the parameter iteration conditions;

[0064] Define a step size descent ratio, and update the target step size according to the number of iterations based on the step size descent ratio to obtain the current updated step size;

[0065] When the current update step size meets the step size threshold, the iteration variables and optimization parameters under the current iteration number are updated in combination with the update formula, wherein the optimization parameters under the current iteration number are obtained by iterating the initial running parameters over the current iteration number;

[0066] The iteration variables and optimization parameters corresponding to the current update step size not meeting the step size threshold are used as the current target optimization parameters of the target optimization model;

[0067] in,

[0068] The quadratic model approximates the objective function based on formula (14):

[0069]

[0070] In the formula m k (p) represents a subproblem of the sequential quadratic programming method, f(x) k () represents the original problem with k iterations. This indicates that the objective function f in x k The transpose of the gradient at point p, where p is the trial step size vector, and B k Let |p|| be the approximate matrix of the Hessian matrix of the objective function f, ||p|| be the norm of vector p, and Δ k The radius of the trust region in the kth iteration;

[0071] The step size descent ratio is expressed based on formula (16):

[0072]

[0073] In the formula ρ k Let f(x) be the acceptance ratio for the k-th iteration. k )-f(x k +p k (This represents the actual decrease.) To predict the amount of decline;

[0074] The iteration conditions are expressed based on formula (17):

[0075]

[0076] In the formula gradient The norms of k, ∈1 and ∈2 are the first and second precision thresholds, respectively. max This represents the maximum number of iterations.

[0077] The update formula is based on formula (18) - formula (19), which means:

[0078]

[0079] In the formula Δk+1 Δ k The trust region radii for the (k+1)th and kth iterations are ρ. k Let η1 and η2 be the trial step for the k-th iteration, and η1 and η2 be the threshold values ​​for the acceptance ratio. The maximum radius of the trust region;

[0080]

[0081] In the formula x k+1 This is the iteration point for the (k+1)th iteration.

[0082] In some embodiments, when the objective constraints include inequality constraints, determining the objective step size in the approximate objective function of the quadratic model based on the trust region sequential quadratic programming method includes:

[0083] By introducing slack variables, the inequality constraints in the target constraints are transformed into equality constraints.

[0084] Add a logarithmic barrier function to construct a penalty function;

[0085] Based on the trust region sequential quadratic programming method, the target step size in the quadratic model approximation objective function of the penalty function is determined.

[0086] The penalty function is expressed based on formula (15):

[0087]

[0088] In the formula, min represents minimization, and μ represents the obstacle parameter. Let f(x) represent the penalty function, f(x) be the objective function of the original problem, and s be the penalty function. i x represents k The vector represents the slack variable, st represents the constraint condition, c(x) represents the constraint function, and s represents the corresponding slack variable vector.

[0089] In some embodiments, the method further includes updating the current target operating parameters at the current time based on the residual between the training values ​​and the true values ​​of the operating parameters at the previous time step.

[0090] According to another aspect of this application, a thermal power unit operation optimization device is also disclosed, the device comprising:

[0091] The target data determination module is used to determine the target optimization function of the thermal power unit and the target constraint conditions corresponding to the target optimization function.

[0092] The target mapping function determination module is used to obtain the nonlinear mapping function and then transform the target optimization function into a representation of the nonlinear mapping function to obtain the target mapping function.

[0093] The current initial operating parameter determination module is used to determine the current initial operating parameters of the target mapping function within the target constraints based on the simulated annealing algorithm;

[0094] The current target optimization parameter determination module is used to iteratively optimize the current initial running parameters using the trust region in-point method to obtain the current target optimization parameters of the target mapping function;

[0095] The current target operating parameter determination module is used to update the current operating parameters of the thermal power unit based on the current target optimization parameters to obtain the current target operating parameters;

[0096] The operation control module is used to control the thermal power unit to operate based on the current target operating parameters;

[0097] The nonlinear mapping function is expressed based on formula (13):

[0098] minf(x), x∈R n stc(x) = 0 or c(x) ≤ 0 (13);

[0099] In the formula, min represents minimization;

[0100] f(x) is the target mapping function to be optimized;

[0101] x is the optimization variable;

[0102] R n To optimize the domain of the variables, n is the number of feature variables;

[0103] st represents the constraint condition;

[0104] c(x) = 0 is an equality constraint;

[0105] c(x)≤0 is an inequality constraint.

[0106] According to another aspect of this application, an electronic device is also disclosed, the electronic device including a memory and at least one processor, the memory storing instructions; the at least one processor invokes the instructions in the memory to cause the electronic device to perform various steps of the thermal power unit operation optimization method as described in any of the preceding claims.

[0107] According to another aspect of this application, a computer-readable storage medium is also disclosed, wherein instructions are stored on the computer-readable storage medium, characterized in that, when executed by a processor, the instructions implement the various steps of the thermal power unit operation optimization method as described in any of the preceding claims.

[0108] The present invention includes, but is not limited to, the following beneficial effects: (1) This scheme improves the accuracy and economy of thermal power unit operation by optimizing the thermal power unit operation data; (2) The model mismatch problem is handled by the residual compensation mechanism; (3) This scheme integrates the trust region in-point method and simulated annealing algorithm, which significantly improves the solution efficiency and global optimality. Attached Figure Description

[0109] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0110] Figure 1 This is a flowchart of a thermal power generating unit operation optimization method according to an embodiment of this application;

[0111] Figure 2 This is another flowchart of the thermal power generating unit operation optimization method according to an embodiment of this application;

[0112] Figure 3 This is another flowchart of the thermal power generating unit operation optimization method according to an embodiment of this application;

[0113] Figure 4 This is another flowchart of the thermal power generating unit operation optimization method according to an embodiment of this application;

[0114] Figure 5 This is a structural block diagram of the thermal power generating unit operation optimization device according to an embodiment of this application;

[0115] Figure 6 This is a schematic diagram of the structure of the electronic device provided in an embodiment of the present invention. Detailed Implementation

[0116] This invention provides a method for optimizing the operation of thermal power units. The method includes determining a target optimization function for the thermal power unit and target constraints corresponding to the target optimization function; obtaining a nonlinear mapping function, transforming the target optimization function into a representation of the nonlinear mapping function to obtain the target mapping function; determining the current initial operating parameters of the target mapping function within the target constraints based on a simulated annealing algorithm; iteratively optimizing the current initial operating parameters using a trust region in-point method to obtain the current target optimization parameters of the target mapping function; updating the current operating parameters of the thermal power unit based on the current target optimization parameters to obtain the current target operating parameters; and controlling the thermal power unit to operate based on the current target operating parameters. This solution improves the accuracy of thermal power unit operation by optimizing the operating data.

[0117] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” or “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0118] For ease of understanding, the specific process of the embodiments of the present invention will be described below. Figure 1 This is a flowchart illustrating a thermal power unit operation optimization method according to an embodiment of this application. Figure 1 As shown, it includes the following steps:

[0119] S100. Determine the objective optimization function of the thermal power unit and the objective constraints corresponding to the objective optimization function.

[0120] During the operation of a thermal power unit, the target optimization function may include a combustion optimization function and a cold-end optimization function. When the target optimization function is a combustion optimization function, step S100, which determines the target optimization function of the thermal power unit and the corresponding target constraints, may be based on the combustion optimization objective of the thermal power unit and the combustion influence characteristics of the combustion optimization objective, and may also be based on the combustion operation mechanism of the thermal power unit, the combustion safety-efficiency balance boundary, and historical combustion data to determine the combustion constraints corresponding to the combustion optimization function.

[0121] Combustion optimization objectives can include, but are not limited to, maximizing boiler efficiency, steam production per ton of coal, and steam production per ton of standard coal. The combustion operation mechanism of thermal power units can refer to controlling the combustion process by adjusting the secondary air volume ratio under various operating conditions, with the core being maintaining the optimal flue gas oxygen content. The combustion safety efficiency balance boundary can refer to the hard constraints that operating parameters must meet, mainly including: upper and lower limits of flue gas oxygen content, and the physical operating range of air volume (especially secondary air volume). Historical combustion data can refer to time-series data of operating parameters collected from the power plant's SIS (System Integrated Circuit System), after screening and preprocessing. Its key dimensions include: unit load, total primary air volume, total secondary air volume, main feedwater flow rate, water-to-coal ratio, and lower heating value of coal fed into the furnace.

[0122] In this example, the function modeling is illustrated using the combustion optimization function as the objective optimization function:

[0123] Combustion optimization works by maximizing boiler efficiency through adjusting air volume (primary or secondary air). Air volume commands in thermal power plants are typically determined by the relationship between unit load and air volume function, and then corrected by flue gas oxygen content. Too low oxygen content indicates incomplete coal combustion, leading to increased losses from incomplete chemical combustion; too high oxygen content results in increased flue gas heat loss. Therefore, a flue gas oxygen content model needs to be established to optimize air volume by constraining oxygen levels.

[0124] Most coal-fired power plants typically use the inverse balance method to calculate boiler efficiency, which involves calculating various heat losses based on test data of coal, fly ash carbon content, slag carbon content, and flue gas temperature for each shift. This method heavily relies on sampling test results and has limited data, making it insufficient for establishing a data-driven boiler efficiency model. Steam production per ton of coal refers to the amount of main steam produced per ton of coal consumed. This indicator does not require complex testing procedures or heat balance calculations, can be monitored in real time, and is applicable to all operating conditions. It can replace boiler efficiency in modeling to characterize combustion performance.

[0125] Before establishing models for steam production per ton of coal and oxygen content in flue gas, it is necessary to classify the boiler operating conditions to improve the model's accuracy and applicability. The boiler operating conditions are classified based on the actual load of the unit (or other parameters that can represent the load). The number of operating conditions depends on the amount of processed data and the number of features selected for modeling. If the number of features is 3 to 6, the data volume for each operating condition must be no less than 1000 sets.

[0126] Steam production per ton of coal (m³) unit_coal and flue gas oxygen content O air The modeling employs a data-driven approach for multidimensional representation, selecting features based on the process mechanism. The total primary air volume (m³) is selected. wind1 Total secondary air volume (m³) wind2 Main water supply flow rate (m) w Water-to-coal ratio (a), lower calorific value of coal fed into the furnace (Q) net.ar As the steam production per ton of coal (m) unit_coal The characteristic variables, i.e., the influence characteristics of the combustion optimization objective, are obtained. The overline indicates normalization of the feature, converted to a value between 1 and 2, expressed in terms of total primary air volume (m). wind1 For example, its normalization formula is: f is the multinomial model obtained by regression after normalizing the processed historical data. The total primary air volume m is selected. wind1 Total secondary air volume (m³) wind2 Main water supply flow rate (m) w The ratio of water to coal, a, is used as O air The characteristic variables, i.e. The overline indicates the normalization of the features, converted to a value between 1 and 2, and g is the multinomial model obtained by regression after normalizing the processed historical data.

[0127] In summary, the combustion optimization model for coal-fired boilers under various operating conditions is shown in the following equation:

[0128]

[0129] Where max represents maximization; m unit_coal The steam production per ton of coal is calculated from the total primary air volume (m³). wind1 Total secondary air volume (m³) wind2 Main water supply flow rate (m) w Water-to-coal ratio (a), lower calorific value of coal fed into the furnace (Q) net.ar Obtained through training after normalization as feature variables; O air 'Upper is a training model for flue gas oxygen content.' Oair and Lower Oair These are the upper and lower limits obtained from historical data.

[0130] Total secondary air volume (m) wind2 The variable being optimized is m, while the others are real-time parameters, normalized by constraints. wind2 The results of the flue gas oxygen training model meet the operating conditions, and the upper and lower limits of flue gas oxygen are obtained based on the maximum and minimum values ​​of the processed historical data.

[0131] Understandably, in the operation optimization of process industries such as thermal power plants, models built using either mechanism-based or data-driven methods cannot perfectly match actual values. Therefore, the recommended values ​​obtained through optimization may not conform to the current actual operating conditions. For example, in this case, assuming the actual flue gas oxygen content is 5.2%, while the flue gas oxygen content obtained through training is 4.3%, directly using the training values ​​as upper and lower limit constraints will result in a total secondary air volume that deviates significantly from the actual situation. Fortunately, for process industries such as thermal power plants, the operation process is continuous and relatively stable. By calculating the model training values ​​at the previous moment and... The residual of the true value is used to supplement the model trained at this moment for optimization. Assuming the training value at the previous moment is 4.25%, the true oxygen content at the previous moment will not deviate too much from the true oxygen content at this moment, which is 5.2%. Assuming it is 5.17%, the residual between the two is 0.92%. The training value at this moment is 4.3%. After residual compensation, the flue gas oxygen content compensation value is 5.22%. The recommended value used for optimization is more in line with reality than the result obtained by using only the training value of 4.3%. This method is limited to sampling and optimization time intervals within the order of seconds, thus ensuring the continuity of data. Specifically, the residual formula between the flue gas oxygen content training value and the true value of flue gas oxygen content can be expressed based on formula (3), and the constraint condition for determining the residual can be expressed based on formula (2).

[0132]

[0133] Furthermore, in another example, the objective optimization function may also include a cold-end optimization function. When the objective optimization function includes a cold-end optimization function, step S100, determining the objective optimization function of the thermal power unit and the corresponding objective constraints, may specifically include: determining the cold-end optimization function of the thermal power unit based on the cold-end optimization objective and the cold-end influence characteristics of the cold-end optimization objective; and determining the cold-end constraints of the thermal power unit based on the cold-end operating mechanism, the cold-end safety efficiency balance boundary, and historical cold-end data. The cold-end optimization objective may include, but is not limited to, maximizing the net power generation of the turbine (turbine power generation - circulating water pump power consumption). The cold-end safety efficiency balance boundary may refer to the constraints that must be met during operation, mainly including: the upper and lower limits of the condenser back pressure, the inherent discrete operating states of the circulating water pump and their combination rules, and the circulating water inlet temperature range, etc. Historical cold-end operation data can refer to time-series data obtained from the power plant's SIS (System Information System) after screening and preprocessing. Key parameters include: unit load (or main steam flow), circulating water inlet temperature, condenser back pressure, circulating water pump current, main steam pressure / temperature, reheat steam temperature, power generation, etc.

[0134] In this example, the function modeling is illustrated using the cold-end optimization function as the objective optimization function:

[0135] Specifically, the principle of cold-end optimization is to maximize net power generation by adjusting the back pressure of the turbine condenser. As the condenser back pressure decreases, the power generation increases accordingly, but it is necessary to increase the power of the circulating water pump to increase the circulating water flow rate in order to reduce the turbine back pressure. Therefore, there exists an optimal condenser back pressure that maximizes the net power generation of the turbine (turbine power generation - circulating water pump power consumption).

[0136] Most coal-fired power plants use either constant-speed pumps or dual-speed pumps that switch between high and low speeds. The operating states of these pumps are discrete. For two identical dual-speed pumps, their operating states can be categorized as: 1. Low-speed single pump, 2. High-speed single pump, 3. Low-speed dual pumps, and 4. High-speed dual pumps. Due to limitations such as hydraulic balance and mechanical vibration, it is impossible to operate one low-speed pump and one high-speed pump simultaneously.

[0137] As can be seen from the above principles, to achieve cold-end optimization, it is necessary to establish a model of power generation and condenser back pressure. Before that, the operating conditions of the steam turbine must first be divided to improve the accuracy and applicability of the model. The operating conditions of the steam turbine cold end are divided according to the main steam flow rate (or other parameters that can represent the load) and the circulating water feedwater temperature. The number of operating conditions depends on the amount of processed data and the number of features selected for modeling. If the number of features is 3 to 6 sets, the data volume of each operating condition must be no less than 1000 sets.

[0138] Steam turbine power generation P genand condenser back pressure p con The modeling employs a data-driven approach for multidimensional representation, selecting features based on the process mechanism. Since turbine power generation is highly coupled with main steam flow, the impact of other features on the target might be underestimated; therefore, the target is replaced with power generation per ton of steam, P. unit_st Select the main steam flow rate m st Main steam temperature T st Main steam pressure P st High-temperature reheat steam temperature T rest and condenser back pressure p con .Right now: The overline indicates the normalization of the features, converted to a value between 1 and 2, and f is the multinomial model obtained by regression after normalizing the processed historical data.

[0139] Condenser back pressure model selection of main steam flow rate m st Circulating water supply temperature T cwin Power consumption P of circulating water pump pump As a feature, namely The overline indicates feature normalization, converted to a value between 1 and 2. g represents the multinomial model obtained from the normalized historical data after processing. The power consumption of the circulating water pump is determined by the total current I of the circulating water pump. pump get, U pump and These are the three-phase motor voltage and power factor, respectively. The current difference between two identical pumps at low speed and high speed is not significant. Therefore, there are four discrete power consumption scenarios for the four circulating water pump operating states: low-speed single pump, high-speed single pump, low-speed dual pump, and high-speed dual pump.

[0140] In summary, the cold-end optimization model of the steam turbine under various operating conditions is shown in the following equation:

[0141]

[0142] I pump =n L I L +n H I H (9);

[0143] x L +x H =1,x L ∈{0,1},x H ∈{0,1} (10);

[0144] n L ≤2x L ,n L∈{0,1,2}, (11)

[0145] n H ≤2x H ,n H ∈{0,1,2} (12).

[0146] in, To select the main steam flow rate m st Main steam temperature T st Main steam pressure P st High-temperature reheat steam temperature T rest and condenser back pressure p con The power generation per ton of steam obtained after normalization training, P pump denoted as the power consumption of the circulating pump; g represents the training model of the condenser vacuum. I is the normalized circulating water feedwater temperature. pump and U pump For the circulating pump current and voltage, n L and n H I represents the number of low-speed pumps and high-speed pumps in operation, respectively. L and I H For low-speed and high-speed current of circulating water pumps, x L and x H This indicates the operating status of the high / low speed pump; 0 / 1 represents low speed / high speed. (error) con This represents the deviation between the condenser vacuum level at the previous moment and the condenser vacuum level at the current moment in the training model. L +x H =1 indicates that the system can only select either low-speed or high-speed mode, and the two are mutually exclusive; n L ≤2x L ,n H ≤2x H This indicates the binding of the number of pumps to the mode. The calculation method and reason for adding the residual error in the constraint are similar to those in the combustion optimization process.

[0147] S102. Obtain the nonlinear mapping function. Transform the objective optimization function into a representation of the nonlinear mapping function to obtain the objective mapping function.

[0148] Specifically, the nonlinear mapping function is expressed based on formula (13):

[0149] minf(x), x∈R n stc(x) = 0 or c(x) ≤ 0 (13);

[0150] In the formula, min represents minimization;

[0151] f(x) is the target mapping function to be optimized;

[0152] x is the optimization variable;

[0153] R n To optimize the domain of the variables, n is the number of feature variables;

[0154] st represents the constraint condition;

[0155] c(x) = 0 is an equality constraint;

[0156] c(x)≤0 is an inequality constraint.

[0157] Understandably, by transforming the objective optimization function into a nonlinear mapping function, engineering problems can be transformed into mathematical problems, simplifying the solution process.

[0158] S104. Determine the current initial operating parameters of the target mapping function within the target constraints based on the simulated annealing algorithm.

[0159] In one example, Figure 2 This is another flowchart of the thermal power unit operation optimization method according to an embodiment of this application. This flowchart is an exemplary illustration of step S104. For details, please refer to [link to relevant documentation]. Figure 2 It includes the following steps:

[0160] S200, Determine the target optimization variables and iterative influence parameters of the objective optimization function.

[0161] Specifically, taking the combustion optimization function as an example, the target optimization variable is the secondary air volume, and the iterative influence parameters include the initial point x0 and the upper limit of the trust region radius. Maximum number of iterations k max Radius adjustment threshold 0 ≤ η1 < η2 < 1, initial trust region radius

[0162] S202. Apply random perturbation to the target optimization variable.

[0163] The random perturbation is expressed as shown in formula (20):

[0164] x k+1 =x k +Δx,Δx~N(0,σ 2 (20);

[0165] In the formula x k+1 x k Let denot be the optimization variables for the (k+1)th and kth iterations, respectively, and let Δx represent the random perturbation, where Δx ~ N(0, σ). 2 ) indicates that Δx follows a mean of 0 and a variance of σ. 2 It follows a normal distribution.

[0166] S204. Based on the penalty function method, the objective constraints corresponding to the objective mapping function are constrained to obtain an unconstrained optimization function.

[0167] Specifically, the unconstrained optimization model is expressed based on formula (21):

[0168] E(x)=f(x)+λ(∑max (0,g) i (x)) 2 +Σh i (x) 2 ) (twenty one);

[0169] In the formula, E(x) represents the penalty function, f(x) represents the original objective function, and λ(Σmax(0,g)) represents the objective function. i (x) 2 )+Σh i (x) 2 ) represents the penalty function constraint, where λ is the penalty factor, and max represents choosing 0 and g. i (x) 2 The larger one, g i (x) 2 Let h represent the inequality constraints of the original problem. i (x) 2 The equality constraints of the original problem are expressed.

[0170] S206. Based on the iterative influence parameters, determine the iterative termination condition for the target optimization variable.

[0171] In this example, the iteration termination condition can be that the algorithm stops computation when any of the following conditions are met: 1. The gradient norm of the objective function at the current iteration point is less than a preset threshold; 2. The trust region radius shrinks to below the minimum value; 3. The algorithm execution reaches the preset maximum number of iterations; 4. The temperature drops to the final value or the annealing iteration count exceeds the limit.

[0172] S208, Obtaining Update Criteria.

[0173] Specifically, the update criterion is expressed based on formula (22):

[0174]

[0175] In the formula, P represents the probability of accepting the new optimized solution, and E(x) k+1 E(x) represents the penalty function for the (k+1)th and kth cycles, and T is the current annealing temperature.

[0176] S210. Based on the update criterion, update and iterate the target optimization variables.

[0177] S212. When the update iteration of the target optimization variable satisfies the iteration termination condition, the updated variable corresponding to the iteration termination condition is used as the current initial running parameter of the target mapping function.

[0178] S106. Use the in-trust region point method to iteratively optimize the current initial running parameters to obtain the current target optimization parameters of the target mapping function.

[0179] Specifically, Figure 3 This is another flowchart of the thermal power unit operation optimization method according to an embodiment of this application. This flowchart is an exemplary illustration of step S106. For details, please refer to [link to relevant documentation]. Figure 3 It includes the following steps:

[0180] S300, Obtain the iterative variables for the trust region point method.

[0181] S302. Based on the trust region sequence quadratic programming method, determine the target step size in the approximate objective function of the quadratic model.

[0182] Specifically, the quadratic model approximates the objective function based on formula (14):

[0183]

[0184] In the formula m k (p) represents a subproblem of the sequential quadratic programming method, f(x) k () represents the original problem with k iterations. This indicates that the objective function f in x k The transpose of the gradient at point p, where p is the trial step size vector, and B k Let |p|| be the approximate matrix of the Hessian matrix of the objective function f, ||p|| be the norm of vector p, and Δ k The radius of the trust region in the kth iteration;

[0185] The target step size is within the trust region.

[0186] S304. Iterate the current target optimization parameters and iteration variables based on the parameter iteration conditions.

[0187] The iteration conditions are expressed based on formula (17):

[0188]

[0189] In the formula gradient The norms of k, ∈1 and ∈2 are the first and second precision thresholds, respectively. max This represents the maximum number of iterations.

[0190] S306. Define the step size descent ratio, update the target step size according to the number of iterations based on the step size descent ratio, and obtain the current updated step size.

[0191] Specifically, the step size descent ratio is expressed based on formula (16):

[0192]

[0193] In the formula ρ k Let f(x) be the acceptance ratio for the k-th iteration. k )-f(x k +p k (This represents the actual decrease.) To predict the amount of decline.

[0194] S308. When the current update step size meets the step size threshold, update the iteration variables and optimization parameters for the current iteration number in combination with the update formula.

[0195] Specifically, the update formula is based on formula (18) - formula (19):

[0196]

[0197] In the formula Δ k+1 Δ k The trust region radii for the (k+1)th and kth iterations are ρ. k Let η1 and η2 be the trial step for the k-th iteration, and η1 and η2 be the threshold values ​​for the acceptance ratio. The maximum radius of the trust region;

[0198]

[0199] In the formula x k+1 This is the iteration point for the (k+1)th iteration.

[0200] The optimization parameters for the current iteration number are obtained by iterating the initial running parameters for the current iteration number.

[0201] S310. When the current update step size does not meet the step size threshold, the corresponding iteration variable and optimization parameter are used as the current target optimization parameters of the target optimization model.

[0202] S108. Update the current operating parameters of the thermal power unit based on the current target optimization parameters to obtain the current target operating parameters.

[0203] S110, Control the thermal power unit to operate based on the current target operating parameters.

[0204] This solution improves the operational accuracy and efficiency of thermal power units by optimizing their target operating parameters.

[0205] Furthermore, in yet another example, Figure 4 This is another flowchart of the thermal power unit operation optimization method according to an embodiment of this application. This flowchart is an exemplary illustration of determining the target step size in the approximate objective function of a quadratic model based on the trust region sequence quadratic programming method when the target constraints include inequality constraints. For details, please refer to [link to relevant documentation]. Figure 4 It includes the following steps:

[0206] S400. Introduce slack variables to convert inequality constraints in the objective constraints into equality constraints.

[0207] S402. Add a logarithmic barrier function to construct a penalty function.

[0208] S404. Based on the trust region sequence quadratic programming method, determine the target step size in the quadratic model approximation objective function of the penalty function.

[0209] The penalty function is expressed based on formula (15):

[0210]

[0211] In the formula, min represents minimization, and μ represents the obstacle parameter. Let f(x) represent the penalty function, f(x) be the objective function of the original problem, and s be the penalty function. i x represents k The vector represents the slack variable, st represents the constraint condition, c(x) represents the constraint function, and s represents the corresponding slack variable vector.

[0212] Specifically, to facilitate understanding of the technology in this solution, an example is provided to illustrate the optimization of thermal power units by combining combustion optimization and cold-end optimization:

[0213] Example 1: Solving the Boiler Combustion Optimization Model

[0214] The boiler combustion optimization model is shown in equations (1) to (5). The initial point of the trust region point method is calculated using the simulated annealing algorithm. The initial point of the simulated annealing algorithm is arbitrarily set to x0 = 1.5. The initial annealing temperature T0 is set to 1000, the cooling rate α = 0.9, and the maximum number of iterations is 1000.

[0215] Apply random perturbation to variable x:

[0216] x1 = x0 + Δx, Δx ~ N(0, 0.3) 2 ) (twenty three)

[0217] Where σ gradually decreases as temperature decreases, σ k =T k / T0*σ0

[0218] By processing the constraints and introducing a penalty factor λ using the penalty function method, we can obtain the following formula:

[0219]

[0220] Based on the Metropolis criterion, determine whether to accept the new solution. At this point, cool down the temperature and re-enter the iteration, eventually obtaining the optimal initial point x0.

[0221] The boiler combustion model is solved using the trust region method. First, the initial point x0 of the total secondary air volume is set as the optimal value for the annealing algorithm, and the upper limit of the trust region radius is set. Maximum number of iterations k max =1000, radius adjustment thresholds η1=0.05, η2=0.9, initial trust region radius Δ0=0.5. By introducing slack variables to handle the inequality constraints, we obtain the subproblem:

[0222]

[0223] Solve the constrained subproblems using the trust region sequential quadratic programming method. The solution.

[0224] Calculate the step size descent ratio:

[0225]

[0226] Since ρ0 > η2 and ||p0|| = Δ0, and ρ0 > η1, we accept the new point x1 = x0 + p0 = 2, expand the trust region radius Δ1 = 2Δ0, and verify whether the result value is within the model constraints. We then inversely normalize the total secondary air volume variable to verify this. Therefore, continue iterating and repeating the above steps until the termination condition is met, and output the optimal result of steam production per ton of coal.

[0227] Example 2: Solving the cold end model of a steam turbine

[0228] The turbine cold end optimization model is shown in equations (6) to (12). First, we assume a circulating water pump operation mode that satisfies equations (11) and (12). We assume that two high-speed pumps are turned on at this time, i.e., x L =0, x H =1, n L =0, n H =2, using the established condenser pressure prediction model to predict the condenser pressure, and simultaneously calculating the power consumption of the circulating water pump, the formula is as follows:

[0229] p con =g(m st ,T cwin ,P pump )+error con (27)

[0230]

[0231] I pump =n L I L +n H I H (29)

[0232] After obtaining the predicted condenser pressure, the value is substituted into the turbine power generation prediction model to predict the slight increase in turbine power.

[0233]

[0234] Based on the predicted slight increase in turbine power and the operating power consumption of the circulating water pump, the net increase in turbine power is calculated.

[0235] At this point, determine whether all (low-speed single pump, low-speed dual pump, high-speed single pump, high-speed dual pump) circulating water pump operating modes have been tested. If not, return to the first step to rebuild an untested circulating water pump operating mode and repeat the above prediction and calculation steps. If all circulating water pump operating modes have been tested, output the optimal circulating water pump operating mode, the optimal condenser back pressure, and the net power generation.

[0236] Furthermore, Figure 5 A structural block diagram of a thermal power unit operation optimization device, such as... Figure 5 As shown, the device includes:

[0237] The target data determination module is used to determine the target optimization function of the thermal power unit and the target constraints corresponding to the target optimization function.

[0238] The target mapping function determination module is used to transform the target optimization function into a representation of the nonlinear mapping function after obtaining the nonlinear mapping function, thus obtaining the target mapping function;

[0239] The current initial operating parameter determination module is used to determine the current initial operating parameters of the target mapping function within the target constraints based on the simulated annealing algorithm;

[0240] The current target optimization parameter determination module is used to iteratively optimize the current initial running parameters using the trust region in-point method to obtain the current target optimization parameters of the target mapping function;

[0241] The current target operating parameter determination module is used to update the current operating parameters of the thermal power unit based on the current target optimization parameters to obtain the current target operating parameters;

[0242] The operation control module is used to control the thermal power unit to operate based on the current target operating parameters;

[0243] The nonlinear mapping function is expressed based on formula (13):

[0244] minf(x), x∈R n stc(x) = 0 or c(x) ≤ 0 (13);

[0245] In the formula, min represents minimization;

[0246] f(x) is the target mapping function to be optimized;

[0247] x is the optimization variable;

[0248] R n To optimize the domain of the variables, n is the number of feature variables;

[0249] st represents the constraint condition;

[0250] c(x) = 0 is an equality constraint;

[0251] c(x)≤0 is an inequality constraint.

[0252] The application of the relevant modules of the device in this example can be referred to the relevant introduction of the method principle above, and will not be repeated here.

[0253] above Figure 5 The apparatus in the embodiments of the present invention will be described in detail from the perspective of modular functional entities. The electronic device in the embodiments of the present invention will be described in detail from the perspective of hardware processing.

[0254] Figure 6 This is a schematic diagram of the structure of an electronic device 600 provided in an embodiment of the present invention. The electronic device 600 can vary significantly due to differences in configuration or performance. It may include one or more central processing units (CPUs) 610 (e.g., one or more processors) and a memory 620, and one or more storage media 630 (e.g., one or more mass storage devices) for storing application programs 633 or data 632. The memory 620 and storage media 630 can be temporary or persistent storage. The program stored in the storage media 630 may include one or more modules (not shown in the diagram), each module including a series of instruction operations on the electronic device 600. Furthermore, the processor 610 may be configured to communicate with the storage media 630 and execute the series of instruction operations in the storage media 630 on the electronic device 600.

[0255] Electronic device 600 may also include one or more power supplies 640, one or more wired or wireless network interfaces 650, one or more input / output interfaces 660, and / or one or more operating systems 631, such as Windows Server, MacOSX, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 6 The illustrated electronic device structure does not constitute a limitation on electronic devices and may include more or fewer components than illustrated, or combine certain components, or have different component arrangements.

[0256] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the steps of a thermal power unit operation optimization method.

[0257] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system, device, or unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0258] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0259] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing the operation of thermal power units, characterized in that, The method includes: Determine the objective optimization function of the thermal power unit and the objective constraints corresponding to the objective optimization function; Obtain the nonlinear mapping function, transform the target optimization function into a representation of the nonlinear mapping function, and obtain the target mapping function; The current initial operating parameters of the target mapping function within the target constraints are determined based on the simulated annealing algorithm; The current initial operating parameters are iteratively optimized using the in-trust region point method to obtain the current target optimization parameters of the target mapping function; The current operating parameters of the thermal power unit are updated based on the current target optimization parameters to obtain the current target operating parameters; Control the thermal power unit to operate based on the current target operating parameters; The nonlinear mapping function is expressed based on formula (13): （13）； In the formula, To minimize; It is the target mapping function that needs to be optimized; To optimize variables; To optimize the domain of variables, The number of feature variables; These are constraints; For equality constraints; Inequality constraints; Wherein, when the objective optimization function includes a combustion optimization function, the determination of the objective optimization function of the thermal power unit and the objective constraint conditions corresponding to the objective optimization function include: The combustion optimization function of the thermal power unit is determined based on the combustion optimization objective of the thermal power unit and the combustion influence characteristics of the combustion optimization objective; The combustion constraints corresponding to the combustion optimization function are determined based on the combustion operation mechanism of the thermal power unit, the combustion safety efficiency balance boundary, and historical combustion data. The combustion optimization function is expressed based on formula (1): （1）； In the formula, max means maximization; The steam production per ton of coal is determined by the total primary air volume. Total secondary air volume Main water supply flow rate Water-to-coal ratio Low calorific value of coal fed into the furnace The feature variables are obtained after normalization and training, and their normalization interval is [1,2]. The combustion constraint conditions are expressed based on formulas (4) and (5): (4)； （5）； In the formula, This is a training model for flue gas oxygen content. and These are the upper and lower limits obtained from historical data.

2. The thermal power unit operation optimization method according to claim 1, characterized in that, The objective optimization function includes a cold-end optimization function, and the determination of the objective optimization function of the thermal power unit and the corresponding objective constraints include: The cold-end optimization function of the thermal power unit is determined based on the cold-end optimization objective and the cold-end influence characteristics of the cold-end optimization objective. The cold-end constraint conditions of the thermal power unit are determined based on the cold-end operation mechanism, cold-end safety efficiency balance boundary and historical cold-end data. The cold-end optimization function is expressed based on formula (6): (6)； In the formula, To select the main steam flow rate Main steam temperature Main steam pressure High-temperature reheat steam temperature Condenser back pressure The power generation per ton of steam obtained after normalization training. Power consumption of the circulating pump; The cold end constraint conditions are expressed based on formulas (7)-(12): (7)； (8)； (9)； (10)； (11)； (12)； In the formula These represent the training models for condenser vacuum. The normalized circulating water feed temperature and For the circulating pump current and voltage, and These represent the number of low-speed pumps and high-speed pumps in operation, respectively. and For low-speed and high-speed current of circulating water pumps, and This indicates the operating status of the high / low speed pump; 0 / 1 represents low speed / high speed. This represents the deviation between the condenser vacuum level at the previous moment and the condenser vacuum level at the current moment in the training model.

3. The thermal power unit operation optimization method according to claim 1, characterized in that, The determination of the current initial operating parameters of the target mapping function within the target constraints based on the simulated annealing algorithm includes: Determine the target optimization variables and iterative influence parameters of the objective optimization function; Apply random perturbations to the target optimization variables; The objective constraints corresponding to the objective mapping function are constrained based on the penalty function method to obtain an unconstrained optimization function. Based on the iterative influence parameters, determine the iterative termination condition for the target optimization variable; Obtain update criteria; Based on the update criteria, the target optimization variables are updated iteratively; When the update iteration of the target optimization variable satisfies the iteration termination condition, the update variable corresponding to the iteration termination condition is used as the current initial running parameter of the target mapping function; The random perturbation is represented by formula (20): (20)； In the formula Let these represent the optimization variables for the (k+1)th and kth iterations, respectively. This indicates a random perturbation. Δx ~N(0,σ 2 ) express Δx Follows a mean of 0 and a variance of σ 2 The normal distribution; The unconstrained optimization model is expressed based on formula (21): (21)； In the formula Represents the penalty function, Represents the original objective function. Denotes the penalty function constraint, where As a penalty factor, max Indicates selecting 0 and The larger one, Let represent the inequality constraints of the original problem. Represent the equality constraints of the original problem; The update criterion is expressed based on formula (22): (22)； In the formula P This represents the probability of accepting the optimized new solution. Indicates the first k+1 The next loop and k The penalty function for the next iteration of the loop. T This is the current annealing temperature.

4. The thermal power unit operation optimization method according to claim 1, characterized in that, When the target constraint is an equality constraint, the step of iteratively optimizing the current initial running parameters using the in-trust region point method to obtain the current target optimization parameters of the target mapping function includes: Obtain the iteration variables for the point method within the trust region; Based on the trust region sequential quadratic programming method, the target step size in the approximate objective function of the quadratic model is determined, wherein the target step size is within the trust region; The current target optimization parameters and the iteration variables are iterated based on the parameter iteration conditions; Define a step size descent ratio, and update the target step size according to the number of iterations based on the step size descent ratio to obtain the current updated step size; When the current update step size meets the step size threshold, the iteration variables and optimization parameters under the current iteration number are updated in combination with the update formula, wherein the optimization parameters under the current iteration number are obtained by iterating the initial running parameters over the current iteration number; The iteration variables and optimization parameters corresponding to the current update step size not meeting the step size threshold are used as the current target optimization parameters of the target optimization model; in, The quadratic model approximates the objective function based on formula (14): (14)； In the formula, This represents a subproblem of sequential quadratic programming. express k The original problem in the next loop, Describe the objective function exist The transpose of the gradient at that point. To test the step size vector, For the objective function The approximate matrix of the Hessian matrix, For vectors norm, The radius of the trust region in the kth iteration; The step size descent ratio is expressed based on formula (16): （16）； In the formula Let the acceptance ratio be the one accepted in the k-th iteration. This represents the actual decrease. To predict the amount of decline; The iteration conditions are expressed based on formula (17): （17）； In the formula gradient norm, These are the first precision threshold and the second precision threshold, respectively. The maximum number of iterations, The update formula is based on formula (18) - formula (19), which means: （18）； In the formula These are the trust region radii for the (k+1)th and kth iterations, respectively. This is the trial step for the k-th iteration. , The threshold for determining the acceptance ratio. The maximum radius of the trust region; （19）； In the formula This is the iteration point for the (k+1)th iteration.

5. The method for optimizing the operation of thermal power units according to claim 1, characterized in that, When the objective constraints include inequality constraints, the objective step size in the approximate objective function of the quadratic model is determined based on the trust region sequential quadratic programming method, including: By introducing slack variables, the inequality constraints in the target constraints are transformed into equality constraints. Add a logarithmic barrier function to construct a penalty function; Based on the trust region sequential quadratic programming method, the target step size in the quadratic model approximation objective function of the penalty function is determined. The penalty function is expressed based on formula (15): （15）； In the formula min Indicates minimization. Indicates obstacle parameters, Represents the penalty function. Let the objective function of the original problem be... express The slack variable corresponding to the vector As constraints, For constraint function, s This is the corresponding slack variable vector.

6. The method for optimizing the operation of thermal power units according to claim 1, characterized in that, The method further includes updating the current target operating parameters at the current time based on the residual between the training values ​​and the true values ​​of the operating parameters at the previous time step.

7. A thermal power unit operation optimization device, characterized in that, The device includes: The target data determination module is used to determine the target optimization function of the thermal power unit and the target constraint conditions corresponding to the target optimization function. The target mapping function determination module is used to obtain the nonlinear mapping function and then transform the target optimization function into a representation of the nonlinear mapping function to obtain the target mapping function. The current initial operating parameter determination module is used to determine the current initial operating parameters of the target mapping function within the target constraints based on the simulated annealing algorithm; The current target optimization parameter determination module is used to iteratively optimize the current initial running parameters using the trust region in-point method to obtain the current target optimization parameters of the target mapping function; The current target operating parameter determination module is used to update the current operating parameters of the thermal power unit based on the current target optimization parameters to obtain the current target operating parameters; The operation control module is used to control the thermal power unit to operate based on the current target operating parameters; The nonlinear mapping function is expressed based on formula (13): （13）； In the formula, To minimize; It is the target mapping function that needs to be optimized; To optimize variables; To optimize the domain of variables, The number of feature variables; These are constraints; For equality constraints; Inequality constraints; When the target optimization function includes a combustion optimization function, the target data determination module is specifically used for: The combustion optimization function of the thermal power unit is determined based on the combustion optimization objective of the thermal power unit and the combustion influence characteristics of the combustion optimization objective; The combustion constraints corresponding to the combustion optimization function are determined based on the combustion operation mechanism of the thermal power unit, the combustion safety efficiency balance boundary, and historical combustion data. The combustion optimization function is expressed based on formula (1): （1）； In the formula, max means maximization; The steam production per ton of coal is determined by the total primary air volume. Total secondary air volume Main water supply flow rate Water-to-coal ratio Low calorific value of coal fed into the furnace The feature variables are obtained after normalization and training, and their normalization interval is [1,2]. The combustion constraint conditions are expressed based on formulas (4) and (5): (4)； （5）； In the formula, This is a training model for flue gas oxygen content. and These are the upper and lower limits obtained from historical data.

8. An electronic device, characterized in that, The electronic device includes a memory and at least one processor, the memory storing instructions; the at least one processor invokes the instructions in the memory to cause the electronic device to perform the steps of the thermal power unit operation optimization method as described in any one of claims 1-6.

9. A computer-readable storage medium storing instructions thereon, characterized in that, When the instructions are executed by the processor, they implement each step of the thermal power unit operation optimization method as described in any one of claims 1-6.

Images