Multi-heat-source complementary joint scheduling optimization method based on load prediction

Abstract

The invention relates to the technical field of comprehensive energy systems, in particular to a load prediction-based multi-heat-source complementation joint scheduling optimization method, which comprises the following steps of: acquiring historical heat load data, cost-related data and heat source-related data of a multi-energy complementation system, and constructing a heat load data set based on the preprocessed historical heat load data; constructing an MLP model, and inputting a thermal load data set to obtain a thermal load predicted value; according to the thermal load predicted value, the cost-related data and the heat source-related data, by taking economic operation cost minimization, carbon dioxide emission minimization and renewable energy utilization rate maximization as targets, a multi-target optimization function is constructed, and balance constraint conditions are defined; solving the multi-objective optimization function by using an NSGA-III algorithm to obtain a day-ahead regulation and control strategy; the multi-energy complementary system dynamically adjusts the control task according to the day-ahead regulation and control strategy, multi-energy complementary joint scheduling is carried out, multi-energy is combined, energy waste and carbon emission are reduced, and multi-energy complementary collaborative scheduling is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of integrated energy systems, and mainly to a multi-heat source complementary joint scheduling optimization method based on load forecasting. Background Art

[0002] In the face of changes in the energy structure, the traditional energy system is no longer sufficient to support the needs of multi-energy development planning. The traditional energy system only faces a single energy system, such as electricity, gas, heat (cold), artificially splitting the resource optimization allocation of each energy system and reducing the overall energy utilization rate; the operation mode of this single energy system can no longer meet the needs of current and future multi-energy development planning; in response to this phenomenon, the concept of coordinated planning of integrated energy systems is proposed, that is, to organically couple multiple energy systems to give full play to the complementary characteristics and synergistic effects of different energy forms; however, existing technologies do not take into account the maximization of energy system resource utilization and lack the ability to coordinate the complementary scheduling of multiple energy sources; The Chinese patent with publication number CN115342419A discloses an intelligent control method for a multi-energy complementary heating system, including: calculating the average daily outdoor temperature in the first N days of heating; determining the start and stop of the gas boiler and the heat pump; after N days of heating, collecting the operating data of the gas boiler, heat pump, circulating pump and blower in the previous N days, and calculating the efficiency of the gas boiler, heat pump, circulating pump and blower; periodically generating a control plan; including: predicting the heating demand in the next cycle; establishing a heating cost model for the multi-energy complementary heating system; setting the heating demand of the multi-energy complementary heating system to be equal to the heating demand in the next cycle, and the optimization target is the heating cost model to obtain the heating demand. The heat cost is the lowest, an optimization algorithm is established; the optimization algorithm is executed; the load of the gas boiler, heat pump and circulating pump obtained is used as the control result of the next cycle", but the method relies on the data of the previous N days of heating for prediction and formulation, and lacks sufficient real-time and flexibility to cope with rapid changes or emergencies in the actual heating process; in addition, the method mainly focuses on the start and stop of gas boilers and heat pumps during the control process, as well as the subsequent calculation of the efficiency of gas boilers, heat pumps, circulating pumps and blowers, but this control method is relatively single, and does not include multiple energy forms in the scope of comprehensive control, which limits the overall energy utilization efficiency and flexibility of the system. Summary of the invention

[0003] In order to solve the above problems existing in the prior art, the present application provides a multi-heat source complementary joint scheduling optimization method based on load forecasting.

[0004] The technical solution of this application is as follows: A multi-heat source complementary joint scheduling optimization method based on load forecasting, the method comprising: Acquire historical heat load data, heat source related data and cost related data of the multi-energy complementary system, preprocess the historical heat load data, and construct a heat load data set; construct a multi-layer perceptron MLP model, use the heat load data set as input data, and obtain a heat load prediction value; According to the predicted heat load value, heat source related data and cost related data, a multi-objective optimization function is constructed with the goal of minimizing economic operation cost, minimizing carbon dioxide emissions and maximizing renewable energy utilization, and the corresponding balance constraints are defined; the non-dominated sorting genetic NSGA-III algorithm is used to solve the multi-objective optimization function and generate a day-ahead control strategy; The multi-energy complementary system dynamically adjusts the control tasks according to the day-ahead control strategy to perform multi-energy complementary joint scheduling.

[0005] Preferably, the historical heat load data includes the primary side water supply temperature, return water temperature and flow rate, secondary side water supply temperature and return water temperature and air temperature of the thermal power station within 24 hours before the current moment, and the air temperature forecast value within the next 24 hours from the current moment; The heat sources of the multi-energy complementary system include ground source heat pumps, gas boilers, waste heat boilers, heat exchangers and gas turbines, and the heat source related data include gas turbine power, ground source heat pump power, heat exchanger efficiency, heat generation power of heat exchangers, heating performance coefficient of ground source heat pumps, gas turbine thermal power ratio, heat generation power of waste heat boilers, heat generation power of gas boilers, waste heat boiler efficiency, gas turbine efficiency and gas boiler efficiency; The cost-related data include electricity purchase costs, natural gas purchase costs, electricity purchase prices, electricity purchase volume, natural gas prices, electricity purchase power, gas turbine carbon dioxide emissions, gas boiler carbon dioxide emissions, volume of natural gas consumed by gas turbines, volume of natural gas consumed by gas boilers, electricity consumed by ground source heat pump heating and carbon dioxide emission coefficient.

[0006] Preferably, a multi-layer perceptron MLP model is constructed, wherein the multi-layer perceptron MLP model includes an input layer, a hidden layer, and an output layer, and is trained by forward propagation and back propagation, wherein: The forward propagation is specifically that the input data passes through the input layer and is transmitted to the hidden layer. Each neuron in the hidden layer performs weighted summation on the input data and performs nonlinear transformation through the activation function, which can be expressed as follows: , ; ; In the formula, Indicates the total number of hidden layers; Indicates hidden layer neurons; Indicates hidden layer neurons; Indicates The connection in the hidden layer Neuron to The weights of the neurons; Indicates The hidden layer The bias of each neuron; Indicates The hidden layer The output of a neuron; represents the activation function of the hidden layer; represents the input of the hidden layer; Represents the index value of the hidden layer; The output of the hidden layer is passed to the output layer, and the output layer outputs the heat load prediction value through the activation function , expressed as: ; In the formula, represents the activation function of the output layer; represents the output layer neurons; Indicates The connection in the hidden layer Neuron to The weights of the neurons; Indicates The connection in the hidden layer The bias of each neuron; The gradient of the loss function to the weights and biases of each layer is calculated by back propagation, and the weights and biases are updated using the gradient descent method, which can be expressed as: ; ; In the formula, represents the learning rate; represents the loss function; represents the weight gradient; represents the bias gradient.

[0007] Preferably, the balance constraint conditions include an electric load balance constraint, a thermal load balance constraint, a gas turbine and waste heat boiler electric heat conversion balance constraint, and a gas boiler and heat exchange device heat generation balance constraint, wherein: The electric load balance constraint is expressed as: ; In the formula, express The power purchased by the power grid during the period; express Gas turbine power at the moment; express Purchase electricity at any time; express Ground source heat pump power at all times; The heat load balance constraint is expressed as: ; In the formula, Indicates the efficiency of the heat exchange device; express Heat generation power of heat exchange device during time period; Indicates the heating performance coefficient of the ground source heat pump; express The amount of electricity consumed by the ground source heat pump at any given moment; express The predicted value of heat load required by users at all times; The balance constraint between the gas turbine and the waste heat boiler is expressed as follows: ; In the formula, represents the heat-to-power ratio of the gas turbine; express Heat generation power of waste heat boiler at every moment; The balance constraint of the heat generation power of the gas boiler and the heat exchange device is expressed as follows: ; In the formula, express The heating power of gas boiler at all times; Indicates the efficiency of the waste heat boiler.

[0008] Preferably, a multi-objective optimization function is constructed with the goals of minimizing economic operation costs, minimizing carbon dioxide emissions, and maximizing renewable energy utilization, specifically: The objective function is established with the goal of minimizing the economic operation cost of the multi-energy complementary system, which can be expressed as follows: ; In the formula, represents the total operating cost; Indicates the cost of purchasing electricity; It represents the cost of purchasing natural gas; express The electricity purchase price at the time; represents the price of natural gas; represents the gas turbine efficiency; Indicates gas boiler efficiency; Indicates a time interval; Indicates the time range; The objective function is established with the goal of minimizing the carbon dioxide emissions of the multi-energy complementary system, which can be expressed as follows: ; In the formula, represents the carbon dioxide emissions of the multi-energy complementary system; represents the CO2 emissions from gas turbines; Indicates the carbon dioxide emissions of gas boilers; represents the carbon dioxide emission factor; express The volume of natural gas consumed by the gas turbine at any given moment; express The volume of natural gas consumed by the gas boiler at any given moment; The objective function is established with the goal of maximizing the utilization rate of renewable energy in the multi-energy complementary system, which can be expressed as follows: ; In the formula, It represents the utilization rate of renewable energy in multi-energy complementary system; The multi-objective optimization function is expressed as follows: .

[0009] Preferably, the non-dominated sorting genetic NSGA-III algorithm is used to solve the multi-objective optimization function and generate a day-ahead control strategy, specifically: Initialize the first generation parent population, which includes individuals, each of which represents a solution vector of a set of multi-objective optimization functions; For the first generation of parent population , calculate the objective function value; perform fast non-dominated sorting based on the objective function value of the first-generation parent population, and generate the first-generation child population through selection, crossover and mutation operations ; Starting from the second generation, the parent population is merged with the child population to generate a new parent population; Calculate the objective function value of the new parent population and perform fast non-dominated sorting. According to the Pareto dominance relationship, stratify the individuals in the new parent population after fast non-dominated sorting. Specifically, start from level 0, check whether each individual is Pareto dominated by any individual in the current level. If the individual is not Pareto dominated by any individual in the current level, assign the individual to a new level; gradually assign each individual until all individuals have corresponding levels; assign a level value to each individual based on the level number of the individual; Starting from level 0, all individuals at each level are added to the next generation population until the next generation population reaches Individual or more individuals; if the last added level causes the next generation population to exceed Each individual in the next generation population is normalized, and the vertical distance and crowding degree of each individual to the reference point are calculated. The individuals are sorted according to the distance from the individual to the reference point and the crowding degree of the reference point. The individuals whose vertical distance and crowding degree are within the preset range are selected to join the next generation population, so that the next generation population has exactly individuals; the next generation population is regarded as the parent population, and selection, crossover and mutation operations are performed to generate the corresponding offspring population; Repeat the process of merging the parent population with the corresponding child population to generate the next generation population until the preset number of evolutionary generations is reached and the Pareto optimal solution set is generated; The Pareto optimal solution set is the solution set of the multi-objective optimization function, and the objective function value corresponding to each solution in the solution set is calculated; using The group decision method determines the weight of each target in the multi-target and obtains the target weight set; according to the target weight set and the objective function value corresponding to each solution in the solution set, the fuzzy comprehensive evaluation algorithm is used to calculate the comprehensive evaluation score of each solution in the solution set, and the day-ahead control strategy is generated based on the solution with the highest comprehensive evaluation score.

[0010] Preferably, use The group decision method determines the weight of each target among multiple targets and obtains the target weight set, which is as follows: The importance of each target in the multi-target is set according to user needs, and the importance is sorted from high to low to obtain the target sequence relationship; the importance of adjacent targets is calculated according to the target sequence relationship , expressed as: , ; In the formula, Indicates target weight; Indicates The index value of the target; Indicates the number of targets; According to the importance Calculate the weight value of each target and get the target weight set, which is expressed as: ; ; ; In the formula, represents the target weight set; Indicates The index value of the target weight.

[0011] Preferably, according to the target weight set and the target function value corresponding to each solution in the solution set, a fuzzy comprehensive evaluation algorithm is used to calculate the comprehensive evaluation score of each solution in the solution set, specifically: Define the evaluation level and the corresponding score level according to the objective function value corresponding to each solution in the solution set; Calculate the performance of any solution in the solution set on each objective and map the performance to the evaluation level, expressed as: ; ; ; In the formula, Indicates the total number of evaluation levels; It means that any solution in the solution set is The goal is to The degree of membership of the evaluation level; It means that any solution in the solution set is The goal was rated The frequency corresponding to each evaluation level; It means that any solution in the solution set is Fuzzy evaluation matrix on the target; Represents the fuzzy evaluation matrix of any solution in the solution set on multiple objectives; According to the fuzzy evaluation matrix and the target weight set Calculate and obtain the comprehensive evaluation result of any solution in the solution set on multiple objectives , expressed as: ; In the formula, represents the fuzzy composite operator; Represents a transpose operation; Indicates that any solution in the solution set is in the target Comprehensive evaluation results on The comprehensive evaluation results and score evaluation level Perform weighted average to obtain a comprehensive evaluation score , expressed as: ; ; In the formula, Indicates the score of the grade of the score assessment; Indicates The score of the score judgment level.

[0012] Preferably, the multi-energy complementary system dynamically adjusts the control tasks according to the day-ahead control strategy, and the control tasks include configuring the power of each heat source in the multi-energy complementary system every hour of the day, regulating the hourly power purchase from the power grid and the gas turbine power supply scheduling and allocation, and regulating the heating scheduling and allocation of ground source heat pumps, gas boilers, and waste heat boilers.

[0013] Preferably, the method further comprises: acquiring real-time data of the multi-energy complementary system, including real-time heat load data, heat source related data and cost related data, updating the heat load prediction value and the solution set of the multi-objective optimization function according to the real-time data, adjusting the day-ahead control strategy in real time based on the updated solution set of the multi-objective optimization function, and executing the control task; Monitor the adjustment of the day-ahead control strategy and the execution of the control tasks in real time, and issue an early warning if there is no adjustment or execution.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1) The present invention provides a multi-heat source complementary joint scheduling optimization method based on load forecasting. By acquiring historical heat load data and cost-related data, a multi-layer perceptron MLP model is constructed to predict heat load, thereby improving the accuracy and reliability of heat load prediction. 2) The present invention provides a multi-heat source complementary joint scheduling optimization method based on load forecasting, which uses the non-dominated sorting genetic algorithm NSGA-III to solve the multi-objective optimization function, thereby enhancing the optimization capability; the multi-objective optimization function improves the balance between economic benefits, environmental protection performance and energy utilization, while meeting user needs and minimizing operating costs and environmental impact; 3) The present invention provides a multi-heat source complementary joint scheduling optimization method based on load forecasting. Based on the real-time operation data of the multi-energy complementary system and the day-ahead control strategy, the multi-energy complementary system dynamically adjusts the control tasks, thereby enhancing the utilization rate of renewable energy and promoting the sustainable development of the multi-energy complementary system. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a method flow chart of an embodiment of the present invention; Figure 2 It is a flow chart of the NSGA-III algorithm of an embodiment of the present invention. DETAILED DESCRIPTION

[0016] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0017] The present invention provides the following technical solution: a multi-heat source complementary joint scheduling optimization method based on load forecasting.

[0018] Example 1 See also Figure 1 This embodiment provides a multi-heat source complementary joint scheduling optimization method based on load forecasting, and the specific steps include: S1. Obtain historical heat load data, heat source related data and cost related data of the multi-energy complementary system, pre-process the historical heat load data, and construct a heat load data set; construct a multi-layer perceptron MLP model, use the heat load data set as input data, and obtain a heat load prediction value; S11, preprocessing includes data cleaning, data standardization and correlation coefficient analysis, wherein the data cleaning includes processing missing values ​​and outliers, and the data standardization specifically unifies the data format; S111, the correlation coefficient analysis specifically extracts the characteristic value of the historical heat load data, and analyzes the correlation between the historical heat load data and the characteristic value of the historical heat load data by using the Pearson correlation coefficient. It is expressed as: , , ; In the formula, Indicates Characteristic values ​​of historical heat load data; The total number of eigenvalues ​​representing historical heat load data; Indicates Heat load data; Indicates the total number of heat load data; Indicates the average value of the heat load characteristic value; Indicates the average value of heat load data; Preset a correlation coefficient threshold range, select historical heat load data and corresponding characteristic values ​​with correlation coefficients within the correlation coefficient threshold range, and construct a heat load data set; S112, the historical heat load data includes the primary side water supply temperature, return water temperature and flow rate, secondary side water supply temperature and return water temperature and air temperature of the thermal power station within the 24 hours before the current time, and the temperature forecast value within the next 24 hours from the current time; The heat sources of the multi-energy complementary system include ground source heat pumps, gas boilers, waste heat boilers, heat exchangers and gas turbines, and the heat source related data include gas turbine power, ground source heat pump power, heat exchanger efficiency, heat generation power of heat exchangers, heating performance coefficient of ground source heat pumps, gas turbine thermal power ratio, heat generation power of waste heat boilers, heat generation power of gas boilers, waste heat boiler efficiency, gas turbine efficiency and gas boiler efficiency; The cost-related data include electricity purchase costs, natural gas purchase costs, electricity purchase prices, electricity purchase volume, natural gas prices, electricity purchase power, gas turbine carbon dioxide emissions, gas boiler carbon dioxide emissions, gas turbine volume of natural gas consumption, gas boiler volume of natural gas consumption, ground source heat pump heating power consumption and carbon dioxide emission coefficient; S12, constructing a multi-layer perceptron MLP model, wherein the multi-layer perceptron MLP model includes an input layer, a hidden layer, and an output layer, and is trained by forward propagation and back propagation; The forward propagation is specifically that the input data passes through the input layer and is transmitted to the hidden layer. Each neuron in the hidden layer performs weighted summation on the input data and performs nonlinear transformation through the activation function, which can be expressed as follows: , ; ; In the formula, Represents the total number of hidden layers; represents hidden layer neurons; represents hidden layer neurons; Indicates The connection in the hidden layer Neuron to The weights of the neurons; Indicates The hidden layer The bias of each neuron; Indicates The hidden layer The output of a neuron; represents the activation function of the hidden layer; Represents the input of the activation function; Represents the index value of the hidden layer; The output of the hidden layer is passed to the output layer, and the output layer outputs the heat load prediction value through the activation function , expressed as: ; In the formula, represents the activation function of the output layer; represents the output layer neurons; Indicates The connection in the hidden layer Neuron to The weights of the neurons; Indicates The connection in the hidden layer The bias of each neuron; The gradient of the loss function to the weights and biases of each layer is calculated by back propagation, and the weights and biases are updated using the gradient descent method, which can be expressed as: ; ; In the formula, represents the learning rate; represents the loss function; represents the weight gradient; represents the bias gradient; S2. Based on the predicted heat load value, heat source related data and cost related data, a multi-objective optimization function is constructed with the goal of minimizing economic operation cost, minimizing carbon dioxide emissions and maximizing renewable energy utilization, and the corresponding balance constraints are defined; S21, the balance constraints include electric load balance constraints, thermal load balance constraints, electric-heat conversion balance constraints between gas turbine and waste heat boiler, and heat generation balance constraints between gas boiler and heat exchange device; S211, the electric load balance constraint is expressed as: ; In the formula, express The power purchased by the power grid during the period; express Gas turbine power at the moment; express Purchase electricity at any time; express Ground source heat pump power at all times; S212, the heat load balance constraint is expressed as: ; In the formula, Indicates the efficiency of the heat exchange device; express Heat generation power of steam hot water heat exchange device during the time period; Indicates the heating performance coefficient of the ground source heat pump; express The amount of electricity consumed by the ground source heat pump at any given moment; express The predicted value of heat load required by users at all times; S213. The balance constraint between the gas turbine and the waste heat boiler is expressed as follows: ; In the formula, represents the heat-to-power ratio of the gas turbine; express Heat generation power of waste heat boiler at every moment; The balance constraint of the heat generation power of the gas boiler and the heat exchange device is expressed as follows: ; In the formula, express The heating power of gas boiler at all times; Indicates the efficiency of the waste heat boiler; S22, constructing a multi-objective optimization function; S221. The objective function is established with the goal of minimizing the economic operation cost of the multi-energy complementary system, which is expressed as follows: ; In the formula, represents the total operating cost; Indicates the cost of purchasing electricity; It represents the cost of purchasing natural gas; express The electricity purchase price at the time; represents the price of natural gas; represents the gas turbine efficiency; Indicates gas boiler efficiency; Indicates a time interval; Indicates the time range; S222. Establish an objective function with the goal of minimizing carbon dioxide emissions from the multi-energy complementary system, expressed as follows: ; In the formula, represents the carbon dioxide emissions of the multi-energy complementary system; represents the CO2 emissions from gas turbines; Indicates the carbon dioxide emissions of gas boilers; represents the carbon dioxide emission factor; express The volume of natural gas consumed by the gas turbine at any given moment; express The volume of natural gas consumed by the gas boiler at any given moment; S223. Establish an objective function with the goal of maximizing the utilization rate of renewable energy in the multi-energy complementary system, which is expressed as follows: ; In the formula, It represents the utilization rate of renewable energy in multi-energy complementary system; The multi-objective optimization function is expressed as follows: ; S3. Please refer to Figure 2 , the non-dominated sorting genetic NSGA-III algorithm is used to solve the multi-objective optimization function and generate the day-ahead control strategy; Initialize the first generation parent population, which includes individuals, each of which represents a solution vector of a set of multi-objective optimization functions; For the first generation of parent population , calculate the objective function value; perform fast non-dominated sorting based on the objective function value of the first-generation parent population, and generate the first-generation child population through selection, crossover and mutation operations ; Starting from the second generation, the parent population is merged with the child population to generate a new parent population; Calculate the objective function value of the new parent population and perform fast non-dominated sorting. According to the Pareto dominance relationship, stratify the individuals in the new parent population after fast non-dominated sorting. Specifically, start from level 0, check whether each individual is Pareto dominated by any individual in the current level. If the individual is not Pareto dominated by any individual in the current level, assign the individual to a new level; gradually assign each individual until all individuals have corresponding levels; assign a level value to each individual based on the level number of the individual; Starting from level 0, all individuals at each level are added to the next generation population until the next generation population reaches Individual or more individuals; if the last added level causes the next generation population to exceed Each individual in the next generation population is normalized, and the vertical distance and crowding degree of each individual to the reference point are calculated. The individuals are sorted according to the distance from the individual to the reference point and the crowding degree of the reference point. The individuals whose vertical distance and crowding degree are within the preset range are selected to join the next generation population, so that the next generation population has exactly individuals; the next generation population is regarded as the parent population, and selection, crossover and mutation operations are performed to generate the corresponding offspring population; Repeat the process of merging the parent population with the corresponding child population to generate the next generation population until the preset number of evolutionary generations is reached and the Pareto optimal solution set is generated; The Pareto optimal solution set is the solution set of the multi-objective optimization function, and the objective function value corresponding to each solution in the solution set is calculated; using The group decision method determines the weight of each target in the multi-target. According to the target weight set and the objective function value corresponding to each solution in the solution set, the fuzzy comprehensive evaluation algorithm is used to calculate the comprehensive evaluation score of each solution in the solution set, and the solution with the highest comprehensive evaluation score is selected to generate the day-ahead control strategy. S4. Utilization The group decision method determines the weight of each target in the multi-target, sets the importance of each target in the multi-target according to user needs, sorts the importance from high to low, and obtains the target sequence relationship; calculates the importance of adjacent targets based on the target sequence relationship , expressed as: , ; In the formula, Indicates target weight; Indicates The index value of the target; Indicates the number of targets; According to the importance Calculate the weight value of each target and obtain the target weight set, which can be expressed as: ; ; ; In the formula, represents the target weight set; Indicates The index value of the target weight; S5. Calculate the comprehensive evaluation score of each solution in the solution set using a fuzzy comprehensive evaluation algorithm according to the target weight set and the target function value corresponding to each solution in the solution set; Define the evaluation level and the corresponding score level according to the objective function value corresponding to each solution in the solution set; Calculate the performance of any solution in the solution set on each objective and map the performance to the evaluation level, expressed as: ; ; ; In the formula, Indicates the total number of evaluation levels; It means that any solution in the solution set is The goal is to The degree of membership of the evaluation level; It means that any solution in the solution set is The goal was rated The frequency corresponding to each evaluation level; It means that any solution in the solution set is Fuzzy evaluation matrix on the target; Represents the fuzzy evaluation matrix of any solution in the solution set on multiple objectives; According to the fuzzy evaluation matrix and the target weight set Calculate and obtain the comprehensive evaluation result of any solution in the solution set on multiple objectives , expressed as: ; In the formula, represents the fuzzy composite operator; Represents a transpose operation; Indicates that any solution in the solution set is in the target The comprehensive evaluation result on the above; wherein the fuzzy composite operator in this embodiment is a weighted average comprehensive operator; The comprehensive evaluation results and score evaluation level Perform weighted average to obtain a comprehensive evaluation score , expressed as: ; ; In the formula, Indicates the score of the grade of the score assessment; Indicates The score of the score judgment level.

[0019] S6. The multi-energy complementary system dynamically adjusts the control tasks according to the day-ahead control strategy and performs multi-energy complementary joint dispatching; the control tasks include configuring the power of each heat source in the multi-energy complementary system every hour of the day, regulating the hourly power purchase from the power grid, the dispatching and allocation of gas turbine power supply, and regulating the dispatching and allocation of heat supply from ground source heat pumps, gas boilers, and waste heat boilers; S7, acquiring real-time data of the multi-energy complementary system, including real-time heat load data, heat source related data and cost related data, updating the heat load forecast value and the solution set of the multi-objective optimization function according to the real-time data, adjusting the day-ahead control strategy in real time based on the updated solution set of the multi-objective optimization function, and executing the control task; Monitor the adjustment of the day-ahead control strategy and the execution of the control tasks in real time, and issue an early warning if there is no adjustment or execution.

[0020] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A multi-heat source complementary joint scheduling optimization method based on load forecasting, characterized in that: The method comprises: Acquire historical heat load data, heat source related data and cost related data of the multi-energy complementary system, preprocess the historical heat load data, and construct a heat load data set; construct a multi-layer perceptron MLP model, use the heat load data set as input data, and obtain a heat load prediction value; According to the predicted heat load value, heat source related data and cost related data, a multi-objective optimization function is constructed with the goal of minimizing economic operation cost, minimizing carbon dioxide emissions and maximizing renewable energy utilization, and the corresponding balance constraints are defined; the non-dominated sorting genetic NSGA-III algorithm is used to solve the multi-objective optimization function and generate a day-ahead control strategy; The multi-energy complementary system dynamically adjusts the control tasks according to the day-ahead control strategy to perform multi-energy complementary joint scheduling.

2. The multi-heat source complementary joint scheduling optimization method based on load forecasting according to claim 1 is characterized in that: The historical heat load data includes the primary side water supply temperature, return water temperature and flow rate, secondary side water supply temperature, return water temperature and air temperature of the thermal power station within the 24 hours before the current time, and the temperature forecast value within the next 24 hours from the current time; The heat sources of the multi-energy complementary system include ground source heat pumps, gas boilers, waste heat boilers, heat exchangers and gas turbines, and the heat source related data include gas turbine power, ground source heat pump power, heat exchanger efficiency, heat generation power of heat exchangers, heating performance coefficient of ground source heat pumps, gas turbine thermal power ratio, heat generation power of waste heat boilers, heat generation power of gas boilers, waste heat boiler efficiency, gas turbine efficiency and gas boiler efficiency; The cost-related data include electricity purchase costs, natural gas purchase costs, electricity purchase prices, electricity purchase volume, natural gas prices, electricity purchase power, gas turbine carbon dioxide emissions, gas boiler carbon dioxide emissions, volume of natural gas consumed by gas turbines, volume of natural gas consumed by gas boilers, electricity consumed by ground source heat pump heating and carbon dioxide emission coefficient.

3. The multi-heat source complementary joint scheduling optimization method based on load forecasting according to claim 2 is characterized in that: A multi-layer perceptron MLP model is constructed, wherein the multi-layer perceptron MLP model includes an input layer, a hidden layer, and an output layer, and is trained through forward propagation and back propagation, wherein: The forward propagation is specifically that the input data passes through the input layer and is transmitted to the hidden layer. Each neuron in the hidden layer performs weighted summation on the input data and performs nonlinear transformation through the activation function, which can be expressed as follows: ， ； ； In the formula, Indicates the total number of hidden layers; Indicates hidden layer neurons; Indicates hidden layer neurons; Indicates The connection in the hidden layer Neuron to The weights of the neurons; Indicates The hidden layer The bias of each neuron; Indicates The hidden layer The output of a neuron; represents the activation function of the hidden layer; represents the input of the hidden layer; Represents the index value of the hidden layer; The output of the hidden layer is passed to the output layer, and the output layer outputs the heat load prediction value through the activation function , expressed as: ； In the formula, represents the activation function of the output layer; represents the output layer neurons; Indicates The connection in the hidden layer Neuron to The weights of the neurons; Indicates The connection in the hidden layer The bias of each neuron; The gradient of the loss function to the weights and biases of each layer is calculated by back propagation, and the weights and biases are updated using the gradient descent method, which can be expressed as: ； ； In the formula, represents the learning rate; represents the loss function; represents the weight gradient; represents the bias gradient.

4. The method for optimizing the multi-heat source complementary joint scheduling based on load forecasting according to claim 3 is characterized in that: The balance constraints include the electric load balance constraint, the heat load balance constraint, the electric-heat conversion balance constraint between the gas turbine and the waste heat boiler, and the heat generation balance constraint between the gas boiler and the heat exchanger, among which: The electric load balance constraint is expressed as: ； In the formula, express The power purchased by the power grid during the period; express Gas turbine power at the moment; express Purchase electricity at any time; express Ground source heat pump power at all times; The heat load balance constraint is expressed as: ； In the formula, Indicates the efficiency of the heat exchange device; express Heat generation power of heat exchange device during time period; Indicates the heating performance coefficient of the ground source heat pump; express The amount of electricity consumed by the ground source heat pump at any given moment; express The predicted value of heat load required by users at all times; The balance constraint between the gas turbine and the waste heat boiler is expressed as follows: ； In the formula, represents the heat-to-power ratio of the gas turbine; express Heat generation power of waste heat boiler at every moment; The balance constraint of the heat generation power of the gas boiler and the heat exchange device is expressed as follows: ； In the formula, express The heating power of gas boiler at all times; Indicates the efficiency of the waste heat boiler.

5. The method for optimizing the multi-heat source complementary joint scheduling based on load forecasting according to claim 4 is characterized in that: With the goals of minimizing economic operation costs, minimizing carbon dioxide emissions and maximizing renewable energy utilization, a multi-objective optimization function is constructed, specifically: The objective function is established with the goal of minimizing the economic operation cost of the multi-energy complementary system, which can be expressed as follows: ； In the formula, represents the total operating cost; Indicates the cost of purchasing electricity; It represents the cost of purchasing natural gas; express The electricity purchase price at the time; represents the price of natural gas; represents the gas turbine efficiency; Indicates gas boiler efficiency; Indicates a time interval; Indicates the time range; The objective function is established with the goal of minimizing the carbon dioxide emissions of the multi-energy complementary system, which can be expressed as follows: ； In the formula, represents the carbon dioxide emissions of the multi-energy complementary system; represents the CO2 emissions from gas turbines; Indicates the carbon dioxide emissions of gas boilers; represents the carbon dioxide emission factor; express The volume of natural gas consumed by the gas turbine at any given moment; express The volume of natural gas consumed by the gas boiler at any given moment; The objective function is established with the goal of maximizing the utilization rate of renewable energy in the multi-energy complementary system, which can be expressed as follows: ； In the formula, It represents the utilization rate of renewable energy in multi-energy complementary system; The multi-objective optimization function is expressed as follows: 。 6. The method for optimizing the multi-heat source complementary joint scheduling based on load forecasting according to claim 5 is characterized in that: The non-dominated sorting genetic NSGA-III algorithm is used to solve the multi-objective optimization function and generate the day-ahead control strategy, which is as follows: Initialize the first generation parent population, which includes individuals, each of which represents a solution vector of a set of multi-objective optimization functions; For the first generation of parent population , calculate the objective function value; perform fast non-dominated sorting based on the objective function value of the first-generation parent population, and generate the first-generation child population through selection, crossover and mutation operations ; Starting from the second generation, the parent population is merged with the child population to generate a new parent population; Calculate the objective function value of the new parent population and perform fast non-dominated sorting. According to the Pareto dominance relationship, stratify the individuals in the new parent population after fast non-dominated sorting. Specifically, start from level 0 and check whether each individual is Pareto dominated by any individual in the current level. If the individual is not Pareto dominated by any individual in the current level, assign the individual to a new level. Assign each individual step by step until all individuals have corresponding levels. Assign a level value to each individual based on the level number of the individual; Starting from level 0, all individuals at each level are added to the next generation population until the next generation population reaches Individuals or more individuals; if the last added level causes the next generation population to exceed Each individual in the next generation population is normalized, and the vertical distance and crowding degree of each individual to the reference point are calculated. The individuals are sorted according to the distance from the individual to the reference point and the crowding degree of the reference point. The individuals whose vertical distance and crowding degree are within the preset range are selected to join the next generation population, so that the next generation population has exactly individuals; the next generation population is regarded as the parent population, and selection, crossover and mutation operations are performed to generate the corresponding offspring population; Repeat the process of merging the parent population with the corresponding child population to generate the next generation population until the preset number of evolutionary generations is reached and the Pareto optimal solution set is generated; The Pareto optimal solution set is the solution set of the multi-objective optimization function, and the objective function value corresponding to each solution in the solution set is calculated; using The group decision method determines the weight of each target among multiple targets and obtains the target weight set; According to the target weight set and the objective function value corresponding to each solution in the solution set, the fuzzy comprehensive evaluation algorithm is used to calculate the comprehensive evaluation score of each solution in the solution set, and the day-ahead control strategy is generated based on the solution with the highest comprehensive evaluation score.

7. The method for optimizing the multi-heat source complementary joint scheduling based on load forecasting according to claim 6 is characterized in that: use The group decision method determines the weight of each target among multiple targets and obtains the target weight set, which is as follows: The importance of each target in the multi-target is set according to user needs, and the importance is sorted from high to low to obtain the target sequence relationship; the importance of adjacent targets is calculated according to the target sequence relationship , expressed as: ， ； In the formula, Indicates target weight; Indicates The index value of the target; Indicates the number of targets; According to the importance Calculate the weight value of each target and get the target weight set, which is expressed as: ； ； ； In the formula, represents the target weight set; Indicates The index value of the target weight.

8. The method for optimizing the multi-heat source complementary joint scheduling based on load forecasting according to claim 7 is characterized in that: According to the target weight set and the target function value corresponding to each solution in the solution set, the fuzzy comprehensive evaluation algorithm is used to calculate the comprehensive evaluation score of each solution in the solution set, which is as follows: Define the evaluation level and the corresponding score level according to the objective function value corresponding to each solution in the solution set; Calculate the performance of any solution in the solution set on each objective and map the performance to the evaluation level, expressed as: ； ； ； In the formula, Indicates the total number of evaluation levels; It means that any solution in the solution set is The goal is to The degree of membership of the evaluation level; It means that any solution in the solution set is The goal was rated The frequency corresponding to each evaluation level; It means that any solution in the solution set is Fuzzy evaluation matrix on the target; Represents the fuzzy evaluation matrix of any solution in the solution set on multiple objectives; According to the fuzzy evaluation matrix and the target weight set Calculate and obtain the comprehensive evaluation result of any solution in the solution set on multiple objectives , expressed as: ； In the formula, represents the fuzzy composite operator; Represents a transpose operation; Indicates that any solution in the solution set is in the target Comprehensive evaluation results on The comprehensive evaluation results and score evaluation level Perform weighted average to obtain a comprehensive evaluation score , expressed as: ； ； In the formula, Indicates the score of the grade of the score assessment; Indicates The score of the score judgment level.

9. The method for optimizing the multi-heat source complementary joint scheduling based on load forecasting according to claim 8 is characterized in that: The multi-energy complementary system dynamically adjusts the control tasks according to the day-ahead control strategy. The control tasks include configuring the power of each heat source in the multi-energy complementary system every hour of the day, regulating the hourly power purchase from the power grid, gas turbine power supply scheduling and allocation, and regulating the heating scheduling and allocation of ground source heat pumps, gas boilers, and waste heat boilers.

10. The method for optimizing the multi-heat source complementary joint scheduling based on load forecasting according to claim 9 is characterized in that: The method further comprises: Acquire real-time data of the multi-energy complementary system, including real-time heat load data, heat source related data and cost related data, update the heat load forecast value and the solution set of the multi-objective optimization function according to the real-time data, adjust the day-ahead control strategy in real time based on the updated solution set of the multi-objective optimization function, and execute the control task; Monitor the adjustment of the day-ahead control strategy and the execution of the control tasks in real time, and issue an early warning if there is no adjustment or execution.

Images