A method for optimizing low-carbon distribution of thermal load of multi-mode heat supply of a thermal power unit group

Abstract

The application discloses a kind of heat load low-carbon distribution optimization methods of thermal power unit group multi-mode heating, belong to the technical field of combined heat and power operation optimization.The application generates multi-source variable working condition data set through Ebsilon mechanism model, constructs the multi-task learning neural network proxy model with power generation power and steam extraction flow as input, energy consumption and steam extraction parameter as output;Combined with the heat balance of the first heating station, the heat load decomposition and initial steam extraction distribution are completed, the minimum total coal consumption of the whole plant is taken as the target, the particle swarm optimization algorithm is used to optimize the constraints such as power balance, heat balance, steam extraction pressure and minimum steam intake of low-pressure cylinder, the unit steam extraction heating capacity is predicted in real time using the proxy model, and the fitness of the particle is calculated, the optimal steam extraction distribution and heating strategy are output.The application improves the optimization accuracy and calculation efficiency, adapts to multi-cylinder mode unit and multi-type first heating station, reduces coal consumption and ensures operation safety, and is suitable for online low-carbon operation scheduling of complex thermal power unit group.

Description

Technical Field

[0001] This invention belongs to the field of cogeneration operation optimization technology, specifically relating to a method for optimizing the low-carbon distribution of heat load in a multi-mode heating system of a cogeneration unit group. Background Technology

[0002] With the continuous increase in the proportion of new energy power generation such as wind power and photovoltaic power, coal-fired power units are transforming from the main power source to a regulating power source, significantly increasing their peak-shaving tasks. Combined heat and power (CHP) units are limited by their "heat-driven power generation" operating characteristics, resulting in insufficient operational flexibility. To improve peak-shaving capabilities, existing technologies have developed various heating modes, such as mid-exhaust steam extraction heating, absorption heat pump heating, low-pressure cylinder cut-off heating, and back-compression turbine exhaust steam heating, effectively expanding the CHP operating domain of the units.

[0003] In practical engineering, thermal power plants typically consist of complex, multi-mode heating systems comprised of multiple generating units and multiple primary heating stations. Currently, adjustments to the heating method rely heavily on operational experience, failing to fully consider the energy consumption characteristics of the generating units and lacking scientific decision-making tools. Existing heat load allocation methods have the following shortcomings: (1) It is difficult to accurately characterize the nonlinear relationship of the unit's coal consumption characteristics; (2) The unit’s extraction steam heating capacity is generally set to a constant value, ignoring the nonlinear effects of the dynamic changes in extraction steam pressure and extraction steam enthalpy caused by the change in extraction steam volume under variable operating conditions on the heating capacity. (3) The accuracy of coal consumption assessment is low, the actual heating effect deviates greatly from the expected result, and it is impossible to accurately calculate the total coal consumption of the whole plant under different allocation schemes.

[0004] In summary, existing technologies cannot simultaneously address multiple constraints, the dynamic characteristics of extraction steam parameters, and the nonlinear characteristics of coal consumption. There is an urgent need for a high-precision and highly adaptable method for optimizing the low-carbon allocation of heat load in thermal power unit groups. Summary of the Invention

[0005] This invention addresses the shortcomings of existing technologies by providing a low-carbon optimization method for heat load allocation in multi-mode heating of thermal power unit groups. It generates multi-source datasets through the Ebsilon mechanism model, constructs a multi-task learning neural network proxy model, and combines particle swarm optimization algorithm to achieve accurate heat load allocation. This solves the problems of low optimization accuracy, inaccurate coal consumption assessment, neglect of dynamic changes in extraction steam parameters, and poor engineering adaptability of existing methods.

[0006] The technical solution of this invention is implemented according to the following steps: Step S1: Mechanism Model Construction and Data Generation A mechanistic simulation model based on Ebsilon software was established for various types of heating units and their corresponding primary heating stations within a thermal power plant; the unit power generation capacity was set. Steam extraction flow rate Variable operating condition calculations are performed under different operating conditions to generate a multi-source dataset including main steam flow, boiler heat consumption, extraction steam pressure, extraction steam enthalpy, and low-pressure cylinder steam inlet flow.

[0007] The operating modes of the heating units include non-shutdown mid-exhaust steam extraction heating, low-pressure cylinder-shutdown single-cylinder steam extraction heating, and low-pressure cylinder-shutdown double-cylinder steam extraction heating; the primary heating stations include Station 1 equipped with an absorption heat pump and steam extraction heater, and Station 2 equipped with a back-pressure steam turbine, exhaust steam heater, and steam extraction heater; electrical load The value is taken as 30%~100% of the unit's rated load, and the extraction steam flow rate. The value ranges from 0 to the unit's maximum extraction steam capacity, and the number of operating points is no less than 2000.

[0008] Step S2: Construction of Multi-Task Learning Neural Network Agent Model Based on the dataset generated in step S1, a multi-task learning neural network agent model is constructed; the model uses the unit's power generation capacity as the benchmark. Steam extraction flow rate As input, with main steam flow rate Boiler heat consumption extraction steam pressure extraction enthalpy Low-pressure cylinder steam intake The model employs a multilayer perceptron architecture, which learns the nonlinear mapping relationship between input and multiple output parameters through shared hidden layers, thereby achieving high-precision prediction of unit energy consumption and extraction steam parameters.

[0009] Step S3: Decomposition of heat load demand and allocation of initial steam extraction volume Real-time acquisition of power grid dispatch instructions Total heat load demand of the heating network Based on the characteristics of the primary heating station equipment, the internal heat sources of the primary station are determined according to the operating priority of the heating equipment. The total heat supply is then established based on the mass balance and heat balance equations of the primary station. With total steam extraction The functional relationship considers the dynamic impact of steam extraction rate on unit steam heating capacity; based on the operating status of each unit, the total steam extraction rate is... The samples were initially allocated to each unit as the initial population for the particle swarm optimization algorithm.

[0010] Step S3.1: Heat balance calculation for Station 1 (configuration of absorption heat pump) In Station 1, the extracted steam for heating enters both the absorption heat pump and the extraction steam heater. The return water from the heating network is heated sequentially by the heat pump and the extraction steam heater. The heat balance calculation is as follows: (1) Heating capacity of heat pump : in, The coefficient of performance (COP) of the heat pump; The steam flow rate driven by the heat pump is expressed in kg / s. The enthalpy of the driving steam (i.e., the enthalpy of the extracted steam) is expressed in kJ / kg. The hydrophobic enthalpy of the driving steam, expressed in kJ / kg. The flow rate of the heating network circulating water at the first station is expressed in kg / s. Specific heat capacity of water, expressed in kJ / (kg·℃); , These are the temperatures after heating by the absorption heat pump and the return water temperature of the heating network, respectively, in °C.

[0011] (2) Heating capacity of extraction steam heater : in, The flow rate of heating steam for the heat network extraction steam heater, in kg / s; The enthalpy of the heater's hydrophobicity is expressed in kJ / kg. The temperature at the outlet of the steam extraction heater (i.e., the water supply temperature) is expressed in °C.

[0012] (3) Total heating supply of No. 1 station : (4) Derive the total steam extraction volume of No. 1 first station With heating Return water temperature extraction enthalpy The relationship is: When the heat load is small, the amount of steam driven by the heat pump is variable; when the heat load is large, the heat pump will preferentially operate at full load, and the amount of steam driven will be constant.

[0013] Step S3.2: Thermal balance calculation for No. 2 primary station (configuration of back-pressure steam turbine) In the No. 2 first station, the exhaust steam heater and the extraction steam heater are arranged in parallel, and their heat balance calculation is as follows: (1) Power generation of back pressure unit : in, The extraction steam rate into the back pressure unit is expressed in kg / s. The enthalpy of the steam entering the back pressure unit is expressed in kJ / kg. The exhaust enthalpy of the back pressure turbine is expressed in kJ / kg. The power generation efficiency of the back-pressure unit is expressed in percent.

[0014] (2) Heat supply from exhaust steam heater : in, This is the enthalpy of the steam exhaust heater, expressed in kJ / kg.

[0015] (3) Heat supply from steam extraction heater : in, The steam extraction rate that directly enters the extraction heater is expressed in kg / s. This is the enthalpy of the steam extraction heater condensate, expressed in kJ / kg.

[0016] (4) Total heating supply of No. 2 first station : Step S3.3: Correlation between total steam extraction rate and total heat load The total heat load demand of the heating network is met: Based on the extraction enthalpy predicted by the surrogate model in step S2, the total heat supply is established. With total steam extraction The functional relationship will This serves as the total amount of steam to be extracted between units. Based on the operating status of each unit, the total steam extraction volume... The samples were initially allocated to each unit as the initial population for the particle swarm optimization algorithm.

[0017] Step S4: Optimized heat load allocation based on PSO The objective function is to minimize the total coal consumption of the entire plant, and the steam extraction rate of each unit is used as the decision variable. The particle swarm optimization algorithm is used to find the optimal solution under multiple constraints. When calculating the fitness of the particles in each iteration, the neural network surrogate model trained in step S2 is called. Based on the current decision variables and the unit's electrical load, the unit's coal consumption and steam extraction enthalpy are calculated in real time to accurately calculate the actual heat supply and the objective function value.

[0018] Step S4.1: Objective Function The objective function is to minimize the total coal consumption of the entire plant, and its expression is as follows: in, Number of generating units; For the first Coal consumption of the unit, in t / h; For the first The extraction steam capacity of the unit is expressed in t / h. For the first The electrical load of the generator unit is expressed in MW.

[0019] Step S4.2: Constraints The optimization process needs to satisfy the following multiple constraints: (1) Power generation balance constraints: in, For the power grid to the first The electrical load command for the generator unit, in MW.

[0020] (2) Thermal equilibrium constraint: in, The actual heat supplied to the first station is expressed in kW. This represents the heat demand at the first station, expressed in kW.

[0021] (3) Constraints on the range of steam extraction for a single unit: in, , The first The minimum and maximum values ​​of the steam extraction capacity of the unit are determined by the unit's thermal-electric operating domain.

[0022] (4) Flow balance constraint: (5) Lower limit constraint of extraction steam pressure: in, The minimum required extraction steam pressure for the primary heating station is expressed in MPa.

[0023] (6) Minimum steam intake constraint for low-pressure cylinder: in, This is the minimum steam intake for the low-pressure cylinder, expressed in kg / s, to ensure cooling and safe operation of the low-pressure cylinder.

[0024] (7) Heating equipment capacity constraints: in, , , , These are the maximum steam intake capacity designed for the heat pump, the maximum steam intake capacity designed for the back pressure turbine, the maximum steam intake capacity of the No. 1 primary station extraction heater, and the maximum steam intake capacity of the No. 2 primary station extraction heater, all in kg / s.

[0025] Step S4.3: Particle fitness calculation (1) Calling the neural network proxy model: The current electrical load and steam extraction volume of each unit are input into the neural network proxy model to obtain the corresponding steam extraction enthalpy and boiler heat consumption.

[0026] (2) Calculate the mixing enthalpy of the heating main pipe : in, For the first The extraction steam enthalpy of the unit is expressed in kJ / kg.

[0027] (3) Calculate the actual heat supply and steam demand of the first station: Mixed enthalpy of the mother tube As the first-stage steam enthalpy, it enters the No. 1 and No. 2 first-stage stations respectively. Combining the heat exchange calculation method of each heat exchanger in the first-stage stations in step S3, the actual total heat supply of the first-stage stations is calculated. .

[0028] (4) Calculate the constraint penalty term: For all constraints, calculate the penalty for constraint violation, including violations of the lower limit of extraction steam pressure, imbalance of extraction steam flow, insufficient steam intake to the low-pressure cylinder, and exceeding equipment capacity limits. The penalty is 0 when a constraint is satisfied, and a penalty value is assigned according to preset rules when a constraint is violated. The sum of the penalty values ​​for all constraint violations is denoted as Penalty, and its expression is: in, As a penalty factor, To constrain violations.

[0029] (5) Calculate the final fitness value: A fitness function is constructed based on the coal consumption of each unit and the penalty value. The fitness value of the particle is calculated using the following formula: in, Let i be the coal consumption of the i-th unit. This is the sum of the penalty values ​​for violating all constraints corresponding to this particle.

[0030] Step S4.4 Particle swarm iteration update After calculating the fitness of all particles, update the individual optimal position of each particle. with the global optimal position Then, based on the particle swarm velocity update formula and position update formula, the velocity and position of all particles are updated to complete one iteration.

[0031] Step S4.5 Iteration Termination Judgment Determine whether the preset maximum number of iterations has been reached. If not, return to the particle fitness calculation step to continue iterating; if the maximum number of iterations has been reached or the convergence condition is met, terminate the iteration.

[0032] Step S5: Optimize the solution output and execution After the particle swarm optimization algorithm iterates to the maximum number of iterations or converges, it outputs the global optimal solution, including the steam consumption of each primary heating station, the optimal steam extraction allocation scheme of each unit, and the low-pressure cylinder operation mode of the unit. This solution is then issued as a heating strategy instruction to each unit and the primary station for execution.

[0033] Beneficial effects Improved optimization accuracy: By embedding the dynamic changes in extraction steam pressure and enthalpy, the heating capacity per unit extraction steam is predicted in real time through a neural network proxy model. The nonlinear impact of extraction steam volume on steam quality and heating capacity is quantified, solving the problem of inaccurate heating capacity assessment by traditional methods. At the same time, by combining the heat balance equations of multiple first stations, the heat load and extraction steam volume are accurately mapped, and the optimization results are more in line with actual engineering needs.

[0034] Optimize computational efficiency: A multi-task learning neural network surrogate model is adopted to replace the iterative calculation of the mechanistic model, which greatly shortens the calculation time of a single round of optimization and meets the real-time requirements of online optimization scheduling of thermal load.

[0035] Reduce operating coal consumption: With the goal of minimizing the total coal consumption of the entire plant, and taking into account various constraints such as power generation balance, heat balance, and equipment capacity, heat load allocation is completed to improve the operating economy of the cogeneration unit group and achieve low-carbon optimized operation.

[0036] It adapts to multiple scenarios: covering various unit operation modes such as no cylinder cut-off, single cylinder cut-off, and double cylinder cut-off, as well as two types of complex heating first station structures: No. 1 first station with absorption heat pump and No. 2 first station with back pressure unit. It is suitable for complex thermal power unit group heating systems with multiple units and multiple first stations.

[0037] Ensuring operational safety: Setting constraints such as the lower limit of extraction steam pressure and the minimum steam intake of the low-pressure cylinder, combined with equipment capacity constraints, ensures the safety and stability of the unit and heating equipment during optimized operation. Attached Figure Description

[0038] Figure 1 This is an overall flowchart of the low-carbon heat load distribution optimization method of the present invention; Figure 2 This is a schematic diagram of the structure of the first heating station of the present invention; Figure 3 This is a structural diagram of the multi-task learning neural network agent model of the present invention; Figure 4 This is a flowchart of the heat load optimization allocation based on the particle swarm algorithm of the present invention. Detailed Implementation

[0039] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings. These specific embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.

[0040] Taking a thermal power plant as an example, the plant is equipped with two heating units (Unit 1 and Unit 2) and two primary heating stations. Primary heating station 1 prioritizes the recovery of waste heat using absorption heat pumps, with any shortfall supplemented by extraction steam heaters. Primary heating station 2 is part of a long-distance heat transmission network and is equipped with a back-pressure steam turbine generator unit connected in parallel with the extraction steam heaters. During the initial and final stages of the heating season, steam extracted from Unit 1 supplies steam to primary heating station 1, and steam extracted from Unit 2 supplies steam to primary heating station 2. During the middle stage of the heating season, both units supply steam together. The structural diagram of the plant's primary heating stations is shown below. Figure 2 .

[0041] S1 Mechanism Model Construction and Data Generation First, mechanistic simulation models of the first power station of Unit 1+1 and Unit 2+2 were established in Ebsilon software, and the accuracy of the models was verified using actual power plant operating data. For Unit 1, three operating modes were considered: no cylinder extraction for heating (normal), single-cylinder extraction for heating from the low-pressure cylinder, and double-cylinder extraction for heating from the low-pressure cylinder, simulating its power generation... For the unit's rated load of 30%~100%, the extraction steam flow rate The system covers multiple operating conditions ranging from 0 to the unit's maximum extraction steam capacity, with no fewer than 2000 operating point values. Operating data for each condition is recorded, including the unit's power generation. Steam extraction flow rate Main steam flow rate Low-pressure cylinder steam intake Boiler heat consumption extraction steam pressure extraction enthalpy This forms a multi-source dataset for Unit 1. Similarly, for Unit 2, we simulate its variable operating conditions under extraction steam heating, low-pressure cylinder cut-off heating, and coupled back-compressor exhaust steam heating modes, generating a dataset for Unit 2 containing the same parameter dimensions.

[0042] S2 Multi-Task Learning Neural Network Agent Model Construction like Figure 3 As shown, a multi-task learning neural network agent model is constructed, employing a multi-task multilayer perceptron (MLP) architecture. The generator's power output is used as the reference. Steam extraction flow rate As two nodes in the input layer, the nonlinear feature extraction between input parameters and multiple output parameters is completed through multiple shared hidden layers. The model is set with five independent output branches, which predict the main steam flow rate. Low-pressure cylinder steam intake Boiler heat consumption extraction steam pressure extraction enthalpy The model is trained using the multi-source dataset generated in step S1, ensuring that the model's prediction accuracy meets the following criteria: Mean Absolute Percentage Error (MAPE) less than 0.5%, and a coefficient of determination... Approaching 1, it enables high-precision and rapid prediction of unit energy consumption and extraction steam parameters.

[0043] S3 Heat Load Demand Decomposition and Initial Steam Extraction Allocation The power load commands of the two generating units are obtained in real time from the power grid dispatching system. , The heat load demand of the No. 1 primary heating station is obtained from the heating network dispatch system. Heat load demand of No. 2 station .

[0044] First, establish the thermal balance relationship for each primary station: For the No. 1 heating station, a cascade heating method using absorption heat pumps and extraction steam heaters is adopted, with the heat pump's heating capacity... satisfy: in, The coefficient of performance (COP) of a heat pump. Steam flow rate (kg / s) for heat pump drive. To drive the steam enthalpy (i.e., extraction steam enthalpy, kJ / kg). To drive the hydrophobic enthalpy of vapor (kJ / kg). The circulating water flow rate of the No. 1 primary heating network is (kg / s). is the specific heat capacity of water (kJ / (kg・℃)). Temperature (°C) after heating by the heat pump. The return water temperature of the heating network is (°C).

[0045] Heating capacity of extraction steam heater satisfy: in, The steam flow rate (kg / s) for the extraction steam heater. The enthalpy of hydrophobicity of the heater (kJ / kg). The outlet water supply temperature of the steam extraction heater is (°C).

[0046] Total heating supply for Station 1: Derivation of the total steam extraction volume of Station No. 1 With heating Return water temperature extraction enthalpy The relationship is: When the heat load demand is low, the amount of steam driven by the heat pump As a variable; when the heat load demand is large, the heat pump will preferentially operate at full load. It is a constant.

[0047] For the No. 2 heating station, the back-pressure unit's power generation capacity satisfy: in, The extraction steam rate (kg / s) entering the back pressure unit. The enthalpy of the steam entering the back pressure compressor (kJ / kg). The exhaust enthalpy (kJ / kg) of the back pressure turbine. The back-pressure unit's power generation efficiency (%).

[0048] Exhaust steam heater heat supply satisfy: in, The enthalpy of the steam exhaust heater (kJ / kg).

[0049] Steam extraction heat supply directly enters the extraction heater satisfy: in, The extraction steam rate (kg / s) directly enters the extraction steam heater. The enthalpy of the steam extraction heater is enthalpy (kJ / kg).

[0050] Total heating supply of Station No. 2: Derivation of the relationship between the total heat supply and total steam extraction of Station No. 2: When the back-pressure compressor is running at full load, its steam intake is constant. When the back-pressure compressor does work, it converts some energy into electrical energy, resulting in a decrease in its exhaust steam heating capacity. Therefore, under the same heat load conditions, the amount of steam required for exhaust steam heating by the back-pressure compressor is greater than that of the direct heating mode. The extra steam is used to compensate for the heat lost due to the work done.

[0051] The two primary heat stations supply heat to users in different regions, and the total system heat load demand is met: And the actual heat supply of the first station is greater than or equal to the heat demand of the first station. ( Based on the extraction enthalpy predicted by the surrogate model in step S2, the total heat supply is established. With total steam extraction The functional relationship will This serves as the total amount of steam to be extracted between units. Based on the operating status of each unit, the total steam extraction volume... The samples were initially assigned to Unit 1 and Unit 2 as the initial population for the particle swarm optimization algorithm.

[0052] S4 Heat Load Optimization Allocation Based on PSO The Particle Swarm Optimization (PSO) algorithm is used to solve the optimal steam extraction allocation scheme for Unit 1 and Unit 2. , The overall algorithm flow and particle fitness calculation logic are as follows: Figure 4 As shown, the specific steps are as follows: Algorithm parameter initialization Algorithm parameter settings: Population size 50-100, maximum number of iterations 100-200, learning factor , Inertial weight The total steam extraction rate determined in step S3. Based on this, initialize the position and velocity of the particle swarm, where the particle position corresponds to the decision variable ( , (i) refers to the steam extraction allocation scheme for the two generating units.

[0053] S4.1 Objective Function The objective function, taking the minimum total coal consumption of the entire plant as the objective, is expressed as: in, Number of generating units (in this embodiment) ), For the first Coal consumption per unit For the first Steam extraction volume of the unit. For the first Unit electrical load; unit generating capacity With extraction steam volume Main steam flow rate The coupling relationship can be expressed as: This is implicit in the prediction results of the neural network surrogate model.

[0054] S4.2: Constraints The decision variable is the steam extraction rate of Unit 1. Unit 2 steam extraction volume The optimization process satisfies the following constraints: (1) Power generation balance constraints: (2) Thermal equilibrium constraint: (3) Constraints on the range of steam extraction for a single unit: (4) Flow balance constraint: (5) Lower limit constraint of extraction steam pressure: (6) Minimum steam intake constraint for low-pressure cylinder: (7) Heating equipment capacity constraints: including the maximum steam inlet capacity of the heat pump, the maximum steam inlet capacity of the back pressure unit, and the maximum steam inlet capacity of each heater: S4.3 Particle Fitness Calculation like Figure 4 As shown, in each iteration of the PSO algorithm, for each group ( , The particle performs the following fitness calculation steps: (1) Calling the neural network proxy model: ( , Input the multi-task learning neural network proxy model trained in step S2, and output the main steam flow rate of Unit 1. Boiler heat consumption extraction enthalpy extraction steam pressure Low-pressure cylinder steam intake ;Will( , Input model, output main steam flow rate of unit 2 Boiler heat consumption extraction enthalpy extraction steam pressure Low-pressure cylinder steam intake .

[0055] (2) Calculate the mixing enthalpy of the main heating pipe: Based on the extraction steam rate and extraction steam enthalpy of the two units, calculate the mixing enthalpy of the heating main heating pipe using the following formula: The mixed enthalpy of the main pipe is used as the steam enthalpy of the two heating stations for subsequent heat balance verification.

[0056] (3) Calculate the actual heat supply and steam demand of the first station: Using the mixed enthalpy of the main pipe as input, calculate the actual heat supply of the two first stations according to the heat balance equations of No. 1 (heat pump + extraction steam heater) and No. 2 (back pressure turbine + extraction steam heater) established in step S3. , At the same time, determine the amount of steam required for the first station to meet the heat load demand. , The calculation process follows the principle of prioritizing full-load operation of heat pumps and back pressure units, and verifies the capacity constraints of each heating device.

[0057] (4) Calculate constraint penalty terms: For all constraints, calculate constraint violation penalty terms, including violations of the lower limit of extraction steam pressure, imbalance of extraction steam flow, insufficient steam inlet to the low-pressure cylinder, and over-limit of equipment capacity. When a constraint is satisfied, the penalty term is 0. When a constraint is violated, a penalty value is assigned according to the preset rules. The sum of the violation penalty values ​​of all constraints is denoted as Penalty, and its expression is: in, As a penalty factor, To constrain the amount of violation. To ensure effective punishment for violations, this punishment factor is set to 10. 5 .

[0058] (5) Calculate the final fitness value: Construct a fitness function based on the coal consumption of each unit and the penalty value, and calculate the fitness value of the particle using the following formula: in, , The coal consumption of Unit 1 and Unit 2 are respectively (based on boiler heat consumption). , (converted) This represents the sum of the constraint violation penalties for the particle.

[0059] S4.4 Particle Swarm Iterative Update After calculating the fitness of all particles, update the individual optimal position of each particle. with the global optimal position Then, based on the particle swarm velocity update formula and position update formula, the velocity and position of all particles are updated to complete one iteration.

[0060] S4.5 Iteration Termination Judgment Determine whether the preset maximum number of iterations has been reached. If not, return to the particle fitness calculation step to continue iterating; if the maximum number of iterations has been reached or the convergence condition is met, terminate the iteration.

[0061] S5 Output Optimization Allocation Scheme When the particle swarm optimization algorithm reaches its maximum number of iterations or meets the convergence condition, it outputs the global optimal solution, including the optimal steam extraction allocation scheme for Unit 1 and Unit 2. , Based on the optimal extraction steam volume corresponding to the low-pressure cylinder steam intake, the operating mode of the unit's low-pressure cylinder (no cylinder cutoff, single cylinder cutoff, double cylinder cutoff) is determined. Combined with the internal heating strategy of the primary heating station, the steam consumption of the heat pump and extraction heater at primary station No. 1, and the steam consumption of the back pressure turbine, exhaust heater, and extraction heater at primary station No. 2 are determined. The above optimal heating extraction steam allocation, unit operating mode, and steam consumption of heating equipment are used as heat load allocation instructions and issued to each unit and the primary heating station for execution.

Claims

1. A method for optimizing the low-carbon distribution of heat load in a multi-mode heating system of a thermal power unit group, characterized in that, Includes the following steps: S1. Mechanism Model Construction and Data Generation: For each heating unit in the thermal power plant, a mechanism simulation model is constructed based on Ebsilon software to simulate various operating modes and variable operating conditions, and generate a multi-source dataset including unit power generation, extraction steam flow, main steam flow, boiler heat consumption, extraction steam pressure, extraction steam enthalpy and low-pressure cylinder steam inlet. S2. Proxy Model Construction: Based on the multi-source dataset generated in step S1, a multi-task learning neural network proxy model is constructed with the unit's power generation and extraction steam flow as inputs and the main steam flow, boiler heat consumption, extraction steam pressure, extraction steam enthalpy and low-pressure cylinder steam inlet as outputs. S3. Heat load decomposition and initial allocation: Real-time acquisition of power grid load instructions and total heat load demand of the heating network, combined with the heat balance equation of the first heating station, to convert the total heat load demand into total steam extraction demand, and initially allocate it to each unit as the initial value for optimization. S4. Heat load optimization allocation: With the goal of minimizing the total coal consumption of the entire plant, and the steam extraction rate of each unit as the decision variable, the particle swarm optimization algorithm is used for optimization. When calculating the fitness of the particles in each iteration, the neural network surrogate model constructed in step S2 is called to obtain the unit coal consumption and steam extraction enthalpy under the current decision variable in real time, and to accurately calculate the actual heat supply and the objective function value. S5. Output the optimal allocation scheme: After the iteration is completed, output the optimal steam extraction allocation scheme of each unit that minimizes the total coal consumption of the whole plant, and issue it for execution.

2. The method according to claim 1, characterized in that, In step S1, the multiple operating modes include a non-switching cylinder extraction steam heating mode, a low-pressure cylinder-switching single-cylinder extraction steam heating mode, and a low-pressure cylinder-switching double-cylinder extraction steam heating mode; the variable operating conditions include at least the power generation power varying between 30% and 100% of the rated load, and the extraction steam flow varying between 0 and the unit's maximum extraction steam capacity.

3. The method according to claim 1, characterized in that, In step S3, the heat balance equation combined with the primary heating station specifically includes: For a primary heating station equipped with an absorption heat pump, a functional relationship between its total heat supply and total steam extraction is established based on the heat pump coefficient of performance, heating network circulating water parameters, extraction steam enthalpy and condensate enthalpy. and / or For a heating station equipped with a back-pressure steam turbine, a functional relationship between its total heat supply and total steam extraction is established based on the power generation of the back-pressure unit, the heat exchange of the exhaust steam heater and the extraction steam heater.

4. The method according to claim 3, characterized in that, When calculating the total steam extraction volume required for the primary heating station, the extraction enthalpy value used is a dynamic value predicted by the neural network surrogate model in step S2.

5. The method according to claim 1, characterized in that, In step S4, the expression for the objective function is: in, Number of generating units For the first Coal consumption of the unit For the first Steam extraction rate of the unit. For the first Electrical load of the generator set.

6. The method according to claim 1 or 5, characterized in that, The optimization process in step S4 needs to satisfy the following constraints: power generation balance constraint, heat balance constraint, single unit steam extraction range constraint, flow balance constraint, lower limit of steam extraction pressure constraint, minimum steam inlet of low-pressure cylinder constraint, and heating equipment capacity constraint.

7. The method according to claim 1, characterized in that, In step S4, calculating the fitness of a particle specifically includes the following steps: S4.3.1: Input the current electrical load and steam extraction volume of each unit into the neural network proxy model to obtain the corresponding steam extraction enthalpy and boiler heat consumption; S4.3.2: Calculate the mixed enthalpy of the heating header based on the extraction steam volume and extraction steam enthalpy of each unit; S4.3.3: Using the aforementioned mixed enthalpy value as the steam enthalpy of the first heating station, calculate the actual heat supply based on the heat balance equation of the first heating station; S4.3.4: Calculate the constraint violation penalty for all constraints; S4.3.5: Calculate the fitness value composed of unit coal consumption and constraint violation penalty terms.

8. The method according to claim 7, characterized in that, In step S4.3.2, the mixing enthalpy value of the heating main pipe The calculation formula is: in, For the first Steam extraction rate of the unit. For the first The extraction steam enthalpy of the Taiwanese generator unit.

9. The method according to claim 7, characterized in that, In step S4.3.4, penalty terms for constraint violations are calculated, including violations of the lower limit of extraction steam pressure, imbalance of extraction steam flow, insufficient steam intake to the low-pressure cylinder, and exceeding equipment capacity limits. When a constraint is satisfied, the penalty term is 0; when a constraint is violated, a penalty value is assigned according to preset rules. The sum of the penalty values ​​for all constraint violations is denoted as Penalty, and its expression is: ,in, As a penalty factor, To constrain the quantity of violations; In step S4.3.5, a fitness function is constructed based on the coal consumption of each unit and the penalty value, and the fitness value of the particle is calculated using the following formula: in, Let i be the coal consumption of the i-th unit. This is the sum of the penalty values ​​for violating all constraints corresponding to this particle.

10. The method according to claim 1, characterized in that, In step S5, the output optimal allocation scheme also includes: the optimal steam consumption of each heating equipment in each heating station, and the corresponding operating mode of the low-pressure cylinder of each unit.

Images