Electric-gas integrated energy system rapid decision generation method, system and equipment based on neural network and medium

Abstract

The invention provides an electricity-gas integrated energy system rapid decision generation method, system, equipment and medium based on a neural network, and relates to the technical field of electricity-gas integrated energy system operation control, and the method comprises the steps: collecting natural gas system data, gas unit safety domain data and power system data of each time period; after normalization processing, constructing a multi-module neural network training set and a power system data set; constructing and training a feedback-compensation multi-module neural network to obtain a security domain prediction model; constructing a penalty function neural network training set; constructing and training a penalty function neural network to obtain an optimal decision generation model; and cooperatively operating the two models to generate an optimal scheduling decision of the power system. According to the method, economic dispatching can be optimized while safe operation of the pipe network is guaranteed, the optimal dispatching decision of the power system is generated, the decision generation efficiency is improved, and good safety, stability and adaptability are achieved.

Description

Background Art

[0002] With the rapid development of renewable energy, the proportion of renewable energy such as wind power and photovoltaic power in the power system has continued to increase. However, due to the strong uncertainty and volatility of these renewable energy sources, the power system is facing increasing challenges.

[0003] In order to effectively cope with the power mismatch caused by the fluctuation of renewable energy, gas-fired units have become an important regulating resource in the power system due to their flexible regulation capabilities and low pollution emissions. With the increase in the installed capacity of gas-fired units, the coupling relationship between the power system and the natural gas system is becoming increasingly close. In the process of participating in the operation and dispatch of the power system, gas-fired units can effectively balance the uncertainty of renewable energy, but their dynamic regulation will also affect the pressure and flow of the natural gas pipeline network, bringing certain safety hazards. Therefore, the traditional independent dispatching mode can no longer give full play to the complementary advantages of the power and natural gas systems, and it is urgent to carry out the coordinated optimization of the power-gas system. Through the study of the coordinated decision-making of the power-gas integrated energy system, the absorption capacity of renewable energy can be improved, while ensuring the safe, stable and economic operation of the power system, which is of great significance to promoting the transformation of the energy system to a clean and low-carbon direction.

[0004] However, since most gas-fired power plants do not store fuel on site, the natural gas used for power generation is extracted from the natural gas pipeline in real time, and the uncertain extraction of gas units may cause fluctuations in gas pressure at the nodes of the natural gas pipeline network, and even violate the safety constraints of the natural gas pipeline network. In the absence of natural gas supply, the rapid ramping of gas units may cause the inlet pipeline pressure to fall below the minimum threshold, causing the gas unit to shut down and cause safety accidents. Therefore, it is necessary to study the safety domain of pipeline operation in the electric-gas integrated energy system (that is, the output range of gas units that does not violate the natural gas safety constraints, also known as the safety domain of gas units).

[0005] In the existing technology, many studies that consider the coordinated decision-making of the electricity-gas integrated energy system use model-driven methods to model and solve the optimal strategy. It is necessary to establish accurate mathematical models of the physical principles, chemical processes, electrical characteristics, etc. of the electricity-gas integrated energy system to describe the behavior and characteristics of the system, such as the energy flow characteristics of the system modeled by the power flow equation and the natural gas partial differential equation. However, these methods require the use of finite difference and other technologies to discretize the partial differential equations, which greatly increases the complexity of the system solution. In addition, with the increasing complexity of modern multi-energy systems, it is difficult to accurately obtain the system's topological structure and related parameters of the network. Therefore, when facing complex and changeable actual operating conditions, the accuracy and adaptability of the existing model-driven methods may be limited. Summary of the invention

[0006] In view of the technical problem that the existing technology uses model-driven methods to model and solve the optimal strategy for coordinated decision-making of electricity-gas integrated energy systems, and it is difficult to solve the topological structure and network-related parameters of complex systems, the invention provides a method, system, equipment and medium for rapid decision-making of electricity-gas integrated energy systems based on neural networks, which can optimize economic dispatch while ensuring the safe operation of the pipeline network, generate optimal dispatch decisions for the power system, improve the efficiency of decision-making, and have good safety, stability and adaptability.

[0007] In a first aspect, the present invention provides a method for generating rapid decisions for an electric-gas integrated energy system based on a neural network, the steps comprising: S1. Collect natural gas system data, gas unit safety domain data and power system data of the same electric-gas integrated energy system at different time periods, and construct a multi-module neural network training set and power system data set after normalization; S2. construct a feedback-compensation multi-module neural network and train it using the multi-module neural network training set to obtain a safety domain prediction model, the input of the safety domain prediction model is the natural gas system data, and the output is the safety domain prediction value of the gas unit; S3. Input the natural gas system data into the safety domain prediction model, calculate the actual output boundary value of the gas unit according to the output safety domain prediction value of the gas unit and the inherent output boundary value of the gas unit, splice the actual output boundary value with the power system data, and construct a penalty function neural network training set; Among them, the inherent output boundary value of the gas unit is the upper and lower output boundary value set when the gas unit is designed and manufactured; S4. construct a penalty function neural network and train it using the penalty function neural network training set to obtain an optimal decision generation model; The input of the power system optimal decision generation model is the actual output boundary value of the gas-fired unit and the power system data, and the output is the dispatching strategy of the gas-fired unit and the coal-fired unit output; S5. Input the natural gas system data to be decided into the safety domain prediction model, use its output to calculate the actual processing boundary value of the gas group, and then input it together with the power system data into the optimal decision model to generate the optimal scheduling decision for the power system.

[0008] It should be further explained that, in step S1, a multi-module neural network training set is constructed based on the normalized natural gas system data and gas unit safety domain data, and a power system data set is constructed based on the normalized power system data; Among them, the natural gas system data includes the gas source pressure, compressor boost pressure, and natural gas load in the natural gas system; The safety domain data of the gas unit includes the upper and lower output limits for safe operation of the gas unit; Power system data includes power load and wind power output data in the power system.

[0009] It should be further explained that in step S1, the original data of any natural gas system data or power system data , and its normalized formula is:

[0010] In the formula, Represents the normalized value; and Respectively The maximum and minimum values ​​of the data type in the collected samples.

[0011] It should be further explained that in step S2, the feedback-compensation multi-module neural network includes a prediction module, a feedback module and a compensation module, wherein: The prediction module includes three single-layer perceptrons with activation functions and one multi-layer perceptron. After the gas source pressure, natural gas load and compressor boost in the natural gas system data are input into the three single-layer perceptrons respectively, the single-layer perceptron extracts features, and then the multi-layer perceptron is used to obtain the preliminary estimated value of the safety domain of the gas unit. The expression of the prediction module is:

[0012] In the formula, It is the set of upper limit outputs in the preliminary estimated value of the safety region of the gas unit; It is the set of lower limit outputs in the preliminary estimated value of the safety region of the gas unit; It is a function for calculating the preliminary estimate of the safety region of the gas unit; Represents the model parameters of the four perceptrons in the prediction module, including the weights and biases of each layer of neurons; is the gas source pressure; For natural gas load; is the compressor boost; The feedback module includes a multi-layer perceptron. The feedback module receives the preliminary estimated value of the safety domain of the gas unit obtained by the prediction module and the corresponding natural gas system data, extracts data features through the multi-layer perceptron, and generates the upper and lower limit correction values ​​of the safe output of the gas unit. The expression of the feedback module is:

[0013] In the formula, It is the correction value of the upper limit of safe output of the gas unit; It is the correction value of the lower limit of safe output of the gas unit; It is a function for calculating the safety output correction value of the gas unit; Represents the model parameters of the multilayer perceptron in the feedback module, including the weights and biases of each layer of neurons; The compensation module includes a multi-layer perceptron. The compensation module is used to capture the fluctuations in the safety domain of the gas unit caused by the overall changes in the natural gas system data. It receives the total gas source pressure data, the total natural gas load data, and the total compressor boost data, and generates the upper and lower limit compensations for the safe output of the gas unit to compensate for the fluctuations in the safety domain of the gas unit caused by the changes in the natural gas system input. The expression of the compensation module is:

[0014] In the formula, It is the compensation for the upper limit of safe output of the gas unit; It is the compensation for the upper limit of safe output of the gas unit; It is a function for calculating the safe output compensation of the gas unit; The model parameters of the multilayer perceptron in the compensation module, including the weights and biases of each layer of neurons; is the total gas source pressure; is the total natural gas load; is the total compressor boost; By combining the prediction module, feedback module and compensation module to form a feedback-compensation multi-module neural network, the calculation of the safety domain prediction value of the gas unit is realized. The expression of the safety domain prediction value of the gas unit is:

[0015] In the formula, It is a function for calculating the predicted value of the safety region of the gas unit; It is the upper limit output in the safety range prediction value of the gas unit; It is the lower limit output in the safety domain prediction value of the gas unit.

[0016] It should be further explained that the feedback-compensation multi-module neural network is trained by multi-module neural data sets. , the training steps are: S201. Initializing the parameters of the feedback-compensation multi-module neural network using the Xavier method, including initializing the weights and biases of each layer of neurons; S202. A feedback-compensation multi-module neural network loss function is constructed using mean square error MSE. The feedback-compensation multi-module neural network loss function is specifically:

[0017]

[0018]

[0019]

[0020] In the formula, It represents the model prediction error calculated by MSE, which constitutes the main component of the loss function; secondly; It represents the error between the output of the prediction module calculated by MSE and the true value, which aims to reduce the correction amount from the feedback module and the compensation amount of the compensation module, thereby improving the overall accuracy; is a regularization term, which uses the Manhattan norm to limit the size of the compensation; Indicates the data batch size; and represents the scale factor; and denote the Manhattan norm and the Euclidean norm respectively; and Represents a set of predicted values ​​of the safety domain of the gas unit; and Represents a set of preliminary estimated values ​​of the safety domain of the gas unit; and The expression corresponds to the true value of the output of the feedback-compensation multi-module neural network; S203. Calculate the loss function based on the multi-module neural network training set and the predicted value of the safety domain of the gas unit. Use the chain rule and automatic differentiation to calculate the gradient of the loss function to the network parameters. Then use the Adam optimizer to update the network parameters. Then, make predictions based on the new network parameters and calculate the loss function based on the true values. Repeat the iterative update until the preset training rounds are reached to obtain the gas unit safety domain prediction model.

[0021] It should be further explained that in step S3, the calculation formula for the actual output boundary of the gas unit is:

[0022]

[0023] In the formula, It is the upper limit output of the actual output boundary value of the gas unit; It is the lower limit output of the actual output boundary value of the gas unit; It is the upper limit output in the safety range prediction value of the gas unit; It is the lower limit output in the safety range prediction value of the gas unit; It is the upper limit output of the inherent output boundary value of the gas unit; It is the lower limit output of the inherent output boundary value of the gas unit.

[0024] It should be further explained that in step S1, the power system data set is represented as ， in is the power load, Provide power for wind turbines; In step S3, the penalty function neural network training set is expressed as .

[0025] It should be further explained that step S4 specifically includes: establishing an economic dispatch model of the power system and a penalty function neural network, based on the objectives and constraints defined by the economic dispatch model of the power system, using the penalty function neural network training set to train the penalty function neural network to approximate the optimal solution in a data-driven manner, and finally forming a decision generation model.

[0026] It should be further explained that the economic dispatch model of the power system includes the output constraints of gas-fired units, the output constraints of coal-fired units and the power balance constraints of source and load. The expression of the economic dispatch model of the power system is:

[0027]

[0028]

[0029] In the formula, For gas units At the moment contribution; For coal-fired units At the moment contribution; For wind turbines At the moment contribution; For power load At the moment size; For gas units At the moment The upper limit output within the actual output boundary value; For gas units At the moment The lower limit output of the actual output boundary value; and For coal-fired units The upper and lower boundaries of the output; , , and They represent the sets corresponding to coal-fired units, gas-fired units, wind turbine units and power loads respectively; The objective function is to minimize the power generation cost of coal-fired units. The objective function is;

[0030] In the formula, , , are the coefficients of the quadratic term and the linear term and the constant term corresponding to the cost function of the coal-fired unit; is the total scheduling period.

[0031] It should be further explained that in step S4, the expression of the penalty function neural network is:

[0032] In the formula, The output of the gas unit; The output of coal-fired units; A function for calculating power system dispatch decisions; Represents the model parameters of the penalty function neural network, including the weights and biases of each layer of neurons; The penalty function neural network loss function expression is:

[0033]

[0034]

[0035]

[0036]

[0037] In the formula, is the weight factor of the loss term of the power balance constraint; is the weight factor of the loss term of the gas unit output constraint; It is the weight factor of the loss term of the coal-fired unit output constraint.

[0038] It should be further explained that in step S4, the training process of the penalty function neural network is: The penalty function neural network training set is input into the penalty function neural network, and then the loss function value is calculated. The network parameters are updated through the Adam optimizer, and then recalculated according to the new parameters. The iterative update is repeated until the preset training rounds are reached to obtain the optimal decision generation model of the power system.

[0039] In a second aspect, the present invention provides a neural network-based electricity-gas integrated energy system rapid decision-making generation system, which is used to implement the above-mentioned neural network-based electricity-gas integrated energy system rapid decision-making generation method, including: The data collection and processing module is used to collect the natural gas system data, gas unit safety domain data and power system data of the same electric-gas integrated energy system at different time periods, and construct a multi-module neural network training set and power system data set after normalization processing; A feedback-compensation multi-module neural network construction and training module is used to construct a feedback-compensation multi-module neural network and train it using a multi-module neural network training set to obtain a safety domain prediction model; A gas unit actual output boundary value calculation module, used to calculate the gas unit actual output boundary value; A penalty function neural network training set construction module is used to splice the actual output boundary value of the gas unit with the corresponding power system data to construct a penalty function neural network training set; The penalty function neural network construction and training module is used to construct the penalty function neural network and train it using the penalty function neural network training set to obtain the optimal decision generation model; The collaborative operation module is used to input the natural gas system data to be decided into the safety domain prediction model, use its output to calculate the actual processing boundary value of the gas group, and then input it into the optimal decision model together with the power system data to generate the optimal scheduling decision for the power system.

[0040] In a third aspect, the present invention provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor is used to implement the steps of the above-mentioned method for rapid decision-making generation of an electric-gas integrated energy system based on a neural network when executing the computer program.

[0041] In a fourth aspect, the present invention provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned method for rapid decision-making generation of an electric-gas integrated energy system based on a neural network.

[0042] The beneficial effects of the present invention are: 1. The present invention collects natural gas system data, gas unit safety domain data and power system data, normalizes them and constructs a multi-module neural network data set and a power system data set, constructs and trains a feedback-compensation multi-module neural network to obtain a gas unit safety domain prediction model, constructs a penalty function neural network training set according to the output of the gas unit safety domain prediction model and the inherent output boundary value of the gas unit combined with the corresponding power system data, trains the penalty function neural network to obtain an optimal decision generation model for the power system, and finally operates the gas unit safety domain prediction model and the power system optimal decision generation model in a coordinated manner to generate an optimal dispatching decision for the power system. The optimal dispatching decision for the power system that fully considers economic factors can be proposed on the premise of considering the actual output boundary value of the gas unit and satisfying the safe operation of the system, reasonably allocates power resources, reduces the operating cost of the power system, and improves the economy of the electric-gas integrated energy system.

[0043] 2. In the process of generating the optimal dispatching decision of the power system, the present invention constructs a feedback compensation multi-module neural network and a penalty function neural network to process the economic dispatching problem of the power system. It does not need to model the entire system, does not rely on complex system models, and reduces the time cost of model establishment and solution. Among them, by constructing and training the feedback-compensation multi-module neural network, the operating safety domain of the gas unit can be accurately predicted in real time to provide protection for the safe operation of the pipeline network; by constructing and training the penalty function neural network, the optimal dispatching problem of the power system is integrated into the network training, and the trained penalty function neural network can provide the optimal dispatching strategy in real time according to the operating safety domain of the gas unit provided by the feedback-compensation multi-module neural network; the coordinated operation of the gas unit safety domain prediction model and the power system optimal decision generation model can effectively avoid the impact of inaccurate model parameters and overly complex models on the dispatching problem of the electric-gas integrated energy system, and at the same time greatly shorten the problem solving time, improve the efficiency of decision generation, and meet the real-time requirements.

[0044] 3. The data sources of the present invention are natural gas system data, gas unit safety domain data and power system data, covering gas source pressure, compressor boost, natural gas load, gas unit safe operation output boundary, power load, wind power output and other information. Through the normalization and comprehensive utilization of these data, the mutual relationship and influence between different subsystems in the electric-gas integrated energy system are fully considered, which can more comprehensively guarantee the safe operation of the pipeline network, improve the adaptability of the decision-making method, and maintain stability under complex and changeable actual operating conditions. Among them, a gas unit safety domain prediction model is constructed and trained using a multi-module neural network training set, so that the prediction of the safe operation of the gas unit is more accurate, which avoids the safety problems that may be caused by improper operation of the gas unit from the source, and further ensures the safe and stable operation of the entire system.

[0045] 4. The rapid decision-making method of the present invention is data-driven. Through the collection, processing and training of a large amount of historical data, it can adapt to the electric-gas integrated energy system under different working conditions and operating conditions. With the continuous accumulation of system operation data, the model can be continuously optimized and updated to further improve the accuracy and adaptability of prediction and decision-making, and has strong versatility and scalability. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solution of the present invention, the accompanying drawings required for use in the description will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0047] Figure 1 It is a flow chart of a method for quickly making decisions for an electric-gas integrated energy system based on a neural network in one embodiment of the present invention.

[0048] Figure 2 It is a schematic diagram of the structure of a feedback-compensation multi-module neural network in one embodiment of the present invention.

[0049] Figure 3 It is a schematic diagram of the structure of a penalty function neural network in one embodiment of the present invention.

[0050] Figure 4 It is a schematic block diagram of a rapid decision-making system for an electric-gas integrated energy system based on a neural network in one embodiment of the present invention.

[0051] Figure 5 It is a schematic diagram of the hardware structure of an electronic device in one embodiment of the present invention. DETAILED DESCRIPTION

[0052] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical scheme of the present invention will be clearly and completely described below in conjunction with the drawings in this specific embodiment. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] The neural network-based rapid decision-making method for an electric-gas integrated energy system involved in this application is mainly aimed at the field of operation control technology of an electric-gas integrated energy system. The technical solution includes collecting natural gas system data, gas unit safety domain data and power system data in various time periods, constructing a multi-module neural network training set and a power system data set after normalization processing; constructing and training a feedback-compensation multi-module neural network to obtain a gas unit safety domain prediction model; constructing a penalty function neural network training set; constructing and training a penalty function neural network to obtain an optimal decision-making generation model for the power system; and synergistically running the two models to generate the optimal scheduling decision for the power system. Compared with the prior art, the present invention collects natural gas system data, gas unit safety domain data and power system data, normalizes them and constructs a multi-module neural network data set and a power system data set, constructs and trains a feedback-compensation multi-module neural network to obtain a gas unit safety domain prediction model, constructs a penalty function neural network training set according to the output of the gas unit safety domain prediction model and the inherent output boundary value of the gas unit combined with the corresponding power system data, trains the penalty function neural network to obtain the power system optimal decision generation model, and finally operates the gas unit safety domain prediction model and the power system optimal decision generation model in a coordinated manner to generate the power system optimal dispatching decision, which can propose the power system optimal dispatching decision that fully considers economic factors under the premise of considering the actual output boundary value of the gas unit and satisfying the safe operation of the system, reasonably allocate power resources, reduce the operating cost of the power system, and improve the economy of the electric-gas integrated energy system; a model-free method is used to generate the optimal dispatching decision of the power system, and the feedback compensation multi-module neural network and the penalty function neural network are constructed to process the economic dispatch problem of the power system, which does not rely on complex system models and reduces the time cost of model establishment and solution. Among them, by constructing and training a feedback-compensation multi-module neural network, the operating safety domain of the gas unit can be accurately predicted in real time to provide protection for the safe operation of the pipeline network. By constructing and training a penalty function neural network, the optimization scheduling problem of the power system is integrated into the network training. The trained penalty function neural network can provide the optimal scheduling strategy in real time according to the operating safety domain of the gas unit provided by the feedback-compensation multi-module neural network; the coordinated operation of the gas unit safety domain prediction model and the power system optimal decision generation model can effectively avoid the impact of inaccurate model parameters and overly complex models on the scheduling problem of the electric-gas integrated energy system, and at the same time greatly shorten the problem solving time, improve the efficiency of decision generation, and meet the real-time requirements. The data sources of the present invention are natural gas system data, gas unit safety domain data and power system data, covering gas source pressure, compressor boost, natural gas load, gas unit safe operation output boundary, power load, wind power output and other information.By normalizing and comprehensively utilizing these data, the interrelationships and influences between different subsystems in the electricity-gas integrated energy system are fully considered, which can more comprehensively guarantee the safe operation of the pipeline network, improve the adaptability of the decision-making method, and maintain stability under complex and changeable actual operating conditions. Among them, a gas unit safety domain prediction model is constructed, and a multi-module neural network training set is used for training, so that the prediction of the safe operation of the gas unit is more accurate, and the safety problems that may be caused by improper operation of the gas unit are avoided from the source, further ensuring the safe and stable operation of the entire system. The rapid decision-making method of the present invention is based on data-driven, and can adapt to the electricity-gas integrated energy system under different working conditions and operating conditions through the collection, processing and training of a large amount of historical data; with the continuous accumulation of system operation data, the model can be continuously optimized and updated, further improving the accuracy and adaptability of predictions and decisions, and has strong versatility and extensibility.

[0054] The neural network-based rapid decision-making method for an integrated electricity-gas energy system involved in this application is mainly aimed at the technical problem that the existing technology uses a model-driven method to model and solve the optimal strategy for collaborative decision-making of an integrated electricity-gas energy system, and it is difficult to solve the complex topology structure and network-related parameters of the complex system.

[0055] The following is a detailed description of the neural network-based electric-gas integrated energy system rapid decision-making method involved in the present application. For the purpose of illustration rather than limitation, specific details such as specific system structures and technologies are proposed to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application can also be implemented in other embodiments without these specific details.

[0056] In the neural network-based electric-gas integrated energy system rapid decision-making generation method involved in the present application, the term "including" used indicates the existence of the described features, entities, steps, operations, elements and / or components, but does not exclude the existence or addition of one or more other features, entities, steps, operations, elements, components and / or their collections. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0057] In order to clearly describe the technical solution of the present application, the words "first", "second" and the like are used to distinguish the same or similar items with substantially the same functions and effects. Those skilled in the art can understand that the words "first", "second" and the like do not limit the quantity and execution order, and the words "first", "second" and the like do not necessarily limit the difference.

[0058] The phrases such as "one embodiment" or "some embodiments" described in the present application mean that the specific features, structures or characteristics described in the embodiment are included in one or more embodiments of the present application. Therefore, the phrases such as "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments" etc. that appear in different places in the present application do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways.

[0059] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] The neural network-based electricity-gas integrated energy system rapid decision-making generation method provided in the embodiment of the present invention is executed by a computer device, and accordingly, the neural network-based electricity-gas integrated energy system rapid decision-making generation system runs in the computer device.

[0061] Figure 1 The flowchart of a method for generating rapid decisions for an electric-gas integrated energy system based on a neural network according to an embodiment of the present invention is shown in FIG. Figure 1 The execution subject can be a neural network-based electric-gas integrated energy system rapid decision-making generation system. According to different needs, the order of the steps in the flow chart can be changed, and some can be omitted.

[0062] like Figure 1 As shown, the neural network-based electric-gas integrated energy system rapid decision-making generation method includes: Step S1, collect natural gas system data, gas unit safety domain data and power system data of the same electric-gas integrated energy system in various time periods, collect natural gas system data, gas unit safety domain data and power system data of the same electric-gas integrated energy system in various time periods, and construct a multi-module neural network training set and power system data set after normalization.

[0063] Collecting and normalizing data on the natural gas system, gas unit safety domain, and power system can provide an accurate and standardized data basis for subsequent model training; constructing corresponding training sets based on different data can help to train the model in a targeted manner, so that the model can better learn the data characteristics and laws of each system and improve the accuracy of model prediction and decision-making.

[0064] In some specific embodiments, a multi-module neural network training set is constructed based on normalized natural gas system data and gas unit safety domain data, and a power system data set is constructed based on normalized power system data; Among them, the natural gas system data includes the gas source pressure, compressor boost pressure, and natural gas load in the natural gas system; The safety domain data of the gas unit includes the upper and lower output limits for safe operation of the gas unit; Power system data includes power load and wind power output data in the power system.

[0065] In some specific embodiments, the power system data set is represented as ,in is the power load, Powering wind turbines.

[0066] In some specific embodiments, for any raw data of natural gas system data or power system data , and its normalized formula is:

[0067] In the formula, Represents the normalized value; and Respectively The maximum and minimum values ​​of the data type in the collected samples.

[0068] Step S2, construct a feedback-compensation multi-module neural network and use the multi-module neural network training set to train, to obtain a safety domain prediction model, the input of the safety domain prediction model is the natural gas system data, and the output is the safety domain prediction value of the gas unit; the structure of the feedback-compensation multi-module neural network is as follows Figure 2 shown.

[0069] In some specific embodiments, the feedback-compensation multi-module neural network includes a prediction module, a feedback module and a compensation module, wherein: The prediction module includes three single-layer perceptrons with activation functions and one multi-layer perceptron. After the gas source pressure, natural gas load and compressor boost in the natural gas system data are input into the three single-layer perceptrons respectively, the single-layer perceptron extracts features, and then the multi-layer perceptron is used to obtain the preliminary estimated value of the safety domain of the gas unit. The feedback module includes a multi-layer perceptron. The feedback module receives the preliminary estimated value of the safety domain of the gas unit obtained by the prediction module and the corresponding natural gas system data, extracts data features through the multi-layer perceptron, and generates the upper and lower limit correction values ​​of the safe output of the gas unit. The compensation module includes a multi-layer perceptron, which is used to capture the fluctuations in the safety domain of the gas unit caused by the overall changes in the natural gas system data. It receives the total gas source pressure data, the total natural gas load data, and the total compressor boost data, and generates the upper and lower limit compensations for the safe output of the gas unit, which is used to compensate for the fluctuations in the safety domain of the gas unit caused by the changes in the natural gas system input. By combining the prediction module, feedback module and compensation module to form a feedback-compensation multi-module neural network, the prediction value of the safety domain of the gas unit can be calculated. The prediction module, feedback module and compensation module work together to more accurately capture the complex relationship between the natural gas system data and the safety domain of the gas unit. The feedback module can correct the upper and lower limits of the safe output according to the preliminary estimate and input data. The compensation module can capture the fluctuations caused by the overall change of the system data, improve the accuracy of the safety domain prediction of the gas unit, and provide a reliable basis for subsequent decision-making.

[0070] The specific structure and functions of each module of the feedback-compensation multi-module neural network enable the network to predict and correct the safety domain of the gas unit from different angles. The modules have clear division of labor and cooperate with each other, which improves the network's ability to handle complex data relationships. Compared with a single-structure neural network, it can more accurately predict the safety domain of the gas unit and enhance the reliability and adaptability of the model.

[0071] Step S3, inputting the natural gas system data into the safety domain prediction model, calculating the actual output boundary value of the gas unit according to the output safety domain prediction value of the gas unit and the inherent output boundary value of the gas unit, splicing the actual output boundary value with the power system data, and constructing a penalty function neural network training set; Among them, the inherent output boundary value of the gas unit is the upper and lower output boundary values ​​set when the gas unit is designed and manufactured.

[0072] The actual output boundary value is calculated through the safety domain prediction value of the gas unit and the inherent output boundary value, and a penalty function neural network training set is constructed. This can comprehensively consider the prediction results and the inherent capabilities of the unit itself, so that the decisions generated by the subsequent training penalty function neural network are more in line with the actual operating conditions, ensuring that the gas unit operates within a safe and reasonable output range and ensuring the safety and stability of the system.

[0073] In some specific embodiments, the actual output boundary calculation formula of the gas unit is:

[0074]

[0075] In the formula, It is the upper limit output of the actual output boundary value of the gas unit; It is the lower limit output of the actual output boundary value of the gas unit; It is the upper limit output in the safety range prediction value of the gas unit; It is the lower limit output in the safety range prediction value of the gas unit; It is the upper limit output of the inherent output boundary value of the gas unit; It is the lower limit output of the inherent output boundary value of the gas unit.

[0076] The calculation formula for the actual output boundary of the gas unit comprehensively considers the predicted value of the safety domain and the inherent output boundary value of the gas unit, ensuring that the calculated actual output boundary is based on the predicted result and conforms to the capacity limitation of the unit itself, providing an accurate basis for the subsequent determination of the actual operating output range of the unit, and ensuring the safety and feasibility of the operation of the gas unit.

[0077] In some specific embodiments, the penalty function neural network training set is expressed as .

[0078] The representation of the penalty function neural network training set facilitates data management, processing and model training. During the model training process, a clear data structure helps the neural network to accurately learn data features and relationships, improve training efficiency and model performance, and enable the model to be better applied to actual power system decision-making.

[0079] Step S4, construct a penalty function neural network and train it using the penalty function neural network training set to obtain an optimal decision generation model. The structure of the penalty function neural network is as follows: Figure 3 As shown; The input of the power system optimal decision generation model is the actual output boundary value of the gas-fired unit, the power load, and the wind power output, and the output is the dispatching strategy for the output of the gas-fired unit and the coal-fired unit.

[0080] An economic dispatch model of the power system is established, and a penalty function neural network is constructed and trained to solve the model to obtain the optimal decision generation model, which takes into account the output constraints of gas-fired units and coal-fired units and the source-load power balance constraints, with the goal of minimizing the power generation cost of coal-fired units; the penalty function neural network can effectively solve the complex model by reasonably setting the loss function and weight factor, and the obtained optimal decision generation model can realize the economic dispatch of the power system, reduce the power generation cost, and improve the economic benefits of the system.

[0081] In some specific embodiments, the power system economic dispatch model includes gas-fired unit output constraints, coal-fired unit output constraints and source-load power balance constraints, and the expression of the power system economic dispatch model is:

[0082]

[0083]

[0084] In the formula, For gas units At the moment contribution; For coal-fired units At the moment contribution; For wind turbines At the moment contribution; For power load At the moment size; For gas units At the moment The upper limit output within the actual output boundary value; For gas units At the moment The lower limit output of the actual output boundary value; and For coal-fired units The upper and lower boundaries of the output; , , and They represent the sets corresponding to coal-fired units, gas-fired units, wind turbine units and power loads respectively; The objective function is to minimize the power generation cost of coal-fired units. The objective function is;

[0085] In the formula, , , are the coefficients of the quadratic term and the linear term and the constant term corresponding to the cost function of the coal-fired unit; is the total scheduling period.

[0086] The constraints and objective functions of the power system economic dispatch model take into account the output limits of gas-fired and coal-fired units and the source-load power balance, with the goal of minimizing the power generation cost of coal-fired units. These constraints and objectives meet the actual operation requirements of the power system. By optimizing the model, the rational allocation of power system resources can be achieved, energy utilization efficiency can be improved, power generation costs can be reduced, and the economic and stable operation of the power system can be guaranteed.

[0087] In some specific embodiments, the penalty function neural network is expressed as: ; In the formula, The output of the gas unit; The output of coal-fired units; A function for calculating power system dispatch decisions; Represents the model parameters of the penalty function neural network, including the weights and biases of each layer of neurons; The penalty function neural network loss function expression is:

[0088]

[0089]

[0090]

[0091]

[0092] In the formula, is the weight factor of the loss term of the power balance constraint; is the weight factor of the loss term of the gas unit output constraint; It is the weight factor of the loss term of the coal-fired unit output constraint.

[0093] By setting a reasonable loss function and comprehensively considering multiple factors such as power balance constraints, gas-fired unit and coal-fired unit output constraints, the neural network can be effectively guided to learn the optimal solution that meets the actual operation requirements; the weight factor can adjust the importance of each constraint according to the actual situation, improve the flexibility and adaptability of the model, and enable the penalty function neural network to better solve the economic dispatch model of the power system.

[0094] In some specific embodiments, the training process of the penalty function neural network is: The penalty function neural network training set is input into the penalty function neural network, and then the loss function value is calculated. The network parameters are updated through the Adam optimizer, and then recalculated according to the new parameters. The iterative update is repeated until the preset training rounds are reached to obtain the optimal decision generation model of the power system.

[0095] The Adam optimizer is used for iterative updates, which can continuously adjust the network parameters and gradually reduce the loss function value, thereby making the model closer to the optimal solution; the preset training rounds ensure the controllability of the training process, ensuring that the model achieves good performance within a reasonable time and computing resources, improving the efficiency and quality of model training, and obtaining a more reliable optimal decision-making generation model for the power system.

[0096] Step S5, input the natural gas system data to be decided into the safety domain prediction model, use its output to calculate the actual processing boundary value of the gas group, and then input it together with the power system data into the optimal decision model to generate the optimal scheduling decision for the power system.

[0097] The two models are run in sequence to generate the optimal dispatching decision for the power system, combining the safety domain prediction of the gas-fired units with the economic dispatching decision of the power system. This not only ensures the safe operation of the gas-fired units, but also realizes the economic optimization dispatching of the power system, comprehensively improving the operating performance and benefits of the electricity-gas integrated energy system.

[0098] In a specific embodiment, a method for generating rapid decisions for an electric-gas integrated energy system based on a neural network includes: Step S1, collecting natural gas system data, gas unit safety domain data and power system data of the same electric-gas integrated energy system in different time periods, performing normalization processing, constructing a multi-module neural network training set based on the normalized natural gas system data and gas unit safety domain data, and constructing a power system data set based on the normalized power system data; Among them, the natural gas system data includes the gas source pressure, compressor boost pressure, and natural gas load in the natural gas system; The safety domain data of the gas unit includes the upper and lower output limits for safe operation of the gas unit; Power system data includes power load and wind power output data in the power system; For the raw data of any natural gas system data or power system data , and its normalized formula is:

[0099] In the formula, Represents the normalized value; and Respectively The maximum and minimum values ​​of the data of the corresponding type in the collected samples; Step S2, constructing a feedback-compensation multi-module neural network, using the multi-module neural network training set to train the feedback-compensation multi-module neural network parameters, and obtaining a gas unit safety domain prediction model, the input of the gas unit safety domain prediction model is the natural gas system data, and the output is the gas unit safety domain prediction value; The feedback-compensation multi-module neural network includes a prediction module, a feedback module and a compensation module, wherein: The prediction module includes three single-layer perceptrons with activation functions and one multi-layer perceptron. The single-layer perceptron for processing the gas source pressure contains 5 neurons and uses the ReLU activation function. The initial value of the weight matrix is ​​set by the Xavier initialization method, the initialization range is [-0.5, 0.5], and the initial value of the bias vector is set to 0.1. The single-layer perceptron for processing natural gas load has 8 neurons, uses the ReLU activation function, the weight matrix is ​​initialized in the range of [-0.5, 0.5], and the initial value of the bias vector is set to 0.2; The single-layer perceptron that processes the compressor boost contains 6 neurons, uses the ReLU activation function, the initial range of the weight matrix is ​​[-0.5, 0.5], and the bias vector is 0.15; The multilayer perceptron has two hidden layers, the first hidden layer contains 10 neurons, the second hidden layer contains 8 neurons, and the output layer has 2 neurons, which correspond to the upper and lower limits of the preliminary estimated value of the safety domain of the gas unit. The activation function of the hidden layer is ReLU, and the output layer uses a linear activation function. The initial value of the weight matrix is ​​determined by Xavier initialization, and the value range is [-0.5, 0.5]. The initial value of the bias vector is set to [0.1, 0.3]. After the gas source pressure, natural gas load and compressor boost in the natural gas system data are input into the corresponding single-layer perceptron, the single-layer perceptron extracts features, and then the multi-layer perceptron is used to obtain the preliminary estimate of the safety domain of the gas unit. The expression of the prediction module is:

[0100] In the formula, It is the set of upper limit outputs in the preliminary estimated value of the safety region of the gas unit; It is the set of lower limit outputs in the preliminary estimated value of the safety region of the gas unit; It is a function for calculating the preliminary estimate of the safety region of the gas unit; Represents the model parameters of the four perceptrons in the prediction module, including the weights and biases of each layer of neurons; is the gas source pressure; For natural gas load; is the compressor boost; The feedback module includes a multi-layer perceptron with three hidden layers, and the number of neurons in each layer is 12, 10, and 8 respectively. The output layer has two neurons, which are used to generate the upper and lower limit correction values ​​of the safe output of the gas unit. The hidden layer activation function uses ReLU, and the output layer is a linear activation function. The initial value of the weight matrix is ​​obtained by Xavier initialization, ranging from [-0.5, 0.5], and the initial value of the bias vector is [0.1, 0.3]. The feedback module accepts the preliminary estimate of the safety domain of the gas unit obtained by the prediction module and the corresponding natural gas system data, extracts data features through the multi-layer perceptron, and generates the upper and lower limit correction values ​​of the safe output of the gas unit. The expression of the feedback module is:

[0101] In the formula, It is the correction value of the upper limit of safe output of the gas unit; It is the correction value of the lower limit of safe output of the gas unit; It is a function for calculating the safety output correction value of the gas unit; Represents the model parameters of the multilayer perceptron in the feedback module, including the weights and biases of each layer of neurons; The compensation module includes a multi-layer perceptron with 2 hidden layers. The first hidden layer contains 10 neurons, the second hidden layer contains 6 neurons, and the output layer has 2 neurons, corresponding to the upper and lower limit compensation of the safe output of the gas unit; the hidden layer activation function is ReLU, and the output layer is a linear activation function; the initial value of the weight matrix is ​​initialized by Xavier, the value range is [-0.5, 0.5], and the initial value of the bias vector is [0.1, 0.2]; the compensation module is used to capture the safety domain fluctuations of the gas unit caused by the overall changes in the natural gas system data, accept the total gas source pressure data, the total natural gas load data and the total compressor boost data, and generate the upper and lower limit compensation of the safe output of the gas unit, which is used to compensate for the safety domain fluctuations of the gas unit caused by the changes in the natural gas system input. The expression of the compensation module is:

[0102] In the formula, It is the compensation for the upper limit of safe output of the gas unit; It is the compensation for the upper limit of safe output of the gas unit; It is a function for calculating the safe output compensation of the gas unit; The model parameters of the multilayer perceptron in the compensation module, including the weights and biases of each layer of neurons; is the total gas source pressure; is the total natural gas load; is the total compressor boost; By combining the prediction module, feedback module and compensation module to form a feedback-compensation multi-module neural network, the calculation of the safety domain prediction value of the gas unit is realized. The expression of the safety domain prediction value of the gas unit is:

[0103] In the formula, It is a function for calculating the predicted value of the safety region of the gas unit; It is the upper limit output in the safety range prediction value of the gas unit; It is the lower limit output in the safety range prediction value of the gas unit; Training feedback-compensation multi-module neural networks with a multi-module neural dataset , the data batch size is set to 32, the learning rate is set to 0.001, the Adam optimizer is used for parameter update, the preset training rounds are 500 times, and the training steps are: S201. Initializing the parameters of the feedback-compensation multi-module neural network using the Xavier method, including initializing the weights and biases of each layer of neurons; S202. A feedback-compensation multi-module neural network loss function is constructed using mean square error MSE. The feedback-compensation multi-module neural network loss function is specifically:

[0104]

[0105]

[0106]

[0107] In the formula, It represents the model prediction error calculated by MSE, which constitutes the main component of the loss function; secondly; It represents the error between the output of the prediction module calculated by MSE and the true value, which aims to reduce the correction amount from the feedback module and the compensation amount of the compensation module, thereby improving the overall accuracy; is a regularization term, which uses the Manhattan norm to limit the size of the compensation; Indicates the data batch size; and represents the scale factor; and denote the Manhattan norm and the Euclidean norm respectively; and Represents a set of predicted values ​​of the safety domain of the gas unit; and Represents a set of preliminary estimated values ​​of the safety domain of the gas unit; and The expression corresponds to the true value of the output of the feedback-compensation multi-module neural network; S203. Calculate the loss function based on the multi-module neural network training set and the predicted value of the safety domain of the gas unit, calculate the gradient of the loss function to the network parameters using the chain rule and automatic differentiation, then update the network parameters using the Adam optimizer, and then calculate the loss function based on the prediction and true value of the new network parameters, repeat the iterative update until the preset training round is reached, and obtain the gas unit safety domain prediction model; During the training process, the model performance is monitored by the validation set loss function value. If the validation set loss does not decrease for 20 consecutive rounds, the training is terminated early to prevent overfitting. Step S3, inputting the natural gas system data corresponding to each group of power system data in the power system data set into the safety domain prediction model, outputting the corresponding gas unit safety domain prediction value, and then calculating the actual output boundary value of the gas unit according to the gas unit safety domain prediction value and the inherent output boundary value of the gas unit, splicing the actual output boundary value of the gas unit with the corresponding power system data, and constructing a penalty function neural network training set; Among them, the inherent output boundary value of the gas unit is the upper and lower output boundary value set when the gas unit is designed and manufactured; The calculation formula for the actual output boundary of the gas unit is:

[0108]

[0109] In the formula, It is the upper limit output of the actual output boundary value of the gas unit; It is the lower limit output of the actual output boundary value of the gas unit; It is the upper limit output in the safety range prediction value of the gas unit; It is the lower limit output in the safety range prediction value of the gas unit; It is the upper limit output of the inherent output boundary value of the gas unit; It is the lower limit output of the inherent output boundary value of the gas unit; Step S4, constructing a penalty function neural network and training it using a penalty function neural network training set to obtain an optimal decision generation model; The input of the power system optimal decision generation model is the actual output boundary value of the gas-fired unit and the power system data, and the output is the dispatching strategy of the gas-fired unit and the coal-fired unit output; The economic dispatch model of the power system includes the output constraints of gas-fired units, the output constraints of coal-fired units and the power balance constraints of source and load. The expression of the economic dispatch model of the power system is:

[0110]

[0111]

[0112] In the formula, For gas units At the moment contribution; For coal-fired units At the moment contribution; For wind turbines At the moment contribution; For power load At the moment size; For gas units At the moment The upper limit output within the actual output boundary value; For gas units At the moment The lower limit output of the actual output boundary value; and For coal-fired units The upper and lower boundaries of the output; , , and They represent the sets corresponding to coal-fired units, gas-fired units, wind turbine units and power loads respectively; The objective function is to minimize the power generation cost of coal-fired units. The objective function is;

[0113] In the formula, , , are the coefficients of the quadratic term and the linear term and the constant term corresponding to the cost function of the coal-fired unit; is the total scheduling period; The expression of penalty function neural network is:

[0114] In the formula, The output of the gas unit; The output of coal-fired units; A function for calculating power system dispatch decisions; Represents the model parameters of the penalty function neural network, including the weights and biases of each layer of neurons; The penalty function neural network loss function expression is:

[0115]

[0116]

[0117]

[0118]

[0119] In the formula, is the weight factor of the loss term of the power balance constraint; is the weight factor of the loss term of the gas unit output constraint; is the weight factor of the loss term of the coal-fired unit output constraint; The training process of penalty function neural network is: The penalty function neural network training set is input into the penalty function neural network, and then the loss function value is calculated, the network parameters are updated through the Adam optimizer, and then the network parameters are recalculated according to the new parameters, and the iterative update is repeated until the preset training rounds are reached to obtain the optimal decision generation model of the power system; Step S5, input the natural gas system data to be decided into the safety domain prediction model, use its output to calculate the actual processing boundary value of the gas group, and then input it together with the power system data into the optimal decision model to generate the optimal scheduling decision for the power system.

[0120] The following is an embodiment of the neural network-based rapid decision-making system for an electric-gas integrated energy system provided by the embodiments of the present disclosure. The active load reduction optimization system and the neural network-based rapid decision-making method for an electric-gas integrated energy system of the above-mentioned embodiments belong to the same inventive concept. For details not fully described in the embodiment of the neural network-based rapid decision-making system for an electric-gas integrated energy system, reference can be made to the above-mentioned embodiment of the neural network-based rapid decision-making method for an electric-gas integrated energy system.

[0121] A mobile terminal implementing various embodiments of the present invention will now be described with reference to the accompanying drawings. In the subsequent description, suffixes such as "module", "component" or "unit" used to represent elements are used only to facilitate the description of the embodiments of the present invention and have no specific meaning in themselves. Therefore, "module" and "component" can be used interchangeably.

[0122] like Figure 4 As shown, the neural network-based electric-gas integrated energy system rapid decision-making generation system includes: The data collection and processing module is used to collect the natural gas system data, gas unit safety domain data and power system data of the same electric-gas integrated energy system at different time periods, and construct a multi-module neural network training set and power system data set after normalization processing; A feedback-compensation multi-module neural network construction and training module is used to construct a feedback-compensation multi-module neural network and train it using a multi-module neural network training set to obtain a safety domain prediction model; A gas unit actual output boundary value calculation module, used to calculate the gas unit actual output boundary value; A penalty function neural network training set construction module is used to splice the actual output boundary value of the gas unit with the corresponding power system data to construct a penalty function neural network training set; The penalty function neural network construction and training module is used to construct the penalty function neural network and train it using the penalty function neural network training set to obtain the optimal decision generation model; The collaborative operation module is used to input the natural gas system data to be decided into the safety domain prediction model, use its output to calculate the actual processing boundary value of the gas group, and then input it into the optimal decision model together with the power system data to generate the optimal scheduling decision for the power system.

[0123] The electric-gas integrated energy system rapid decision-making generation system of this embodiment is used to implement a neural network-based electric-gas integrated energy system rapid decision-making generation method, and the steps include: S1. Collect natural gas system data, gas unit safety domain data and power system data of the same electric-gas integrated energy system at different time periods, and construct a multi-module neural network training set and power system data set after normalization; S2. construct a feedback-compensation multi-module neural network and train it using the multi-module neural network training set to obtain a safety domain prediction model, the input of the safety domain prediction model is the natural gas system data, and the output is the safety domain prediction value of the gas unit; S3. Input the natural gas system data into the safety domain prediction model, calculate the actual output boundary value of the gas unit according to the output safety domain prediction value of the gas unit and the inherent output boundary value of the gas unit, splice the actual output boundary value with the power system data, and construct a penalty function neural network training set; Among them, the inherent output boundary value of the gas unit is the upper and lower output boundary value set when the gas unit is designed and manufactured; S4. construct a penalty function neural network and train it using the penalty function neural network training set to obtain an optimal decision generation model; The input of the power system optimal decision generation model is the actual output boundary value of the gas-fired unit and the power system data, and the output is the dispatching strategy of the gas-fired unit and the coal-fired unit output; S5. Input the natural gas system data to be decided into the safety domain prediction model, use its output to calculate the actual processing boundary value of the gas group, and then input it together with the power system data into the optimal decision model to generate the optimal scheduling decision for the power system.

[0124] The present application also provides an electronic device for implementing various embodiments of the present invention. The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor.

[0125] Those skilled in the art will appreciate that the electronic device structure involved in the embodiments of the present invention does not constitute a limitation on the electronic device, and the electronic device may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0126] Figure 5 A schematic diagram of the hardware structure of an electronic device for implementing various embodiments of the present invention.

[0127] The electronic device includes but is not limited to components such as a processor and a memory. Those skilled in the art will appreciate that the electronic device structure involved in the embodiments of the present invention does not constitute a limitation on the electronic device, and the electronic device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0128] In the embodiments of the present invention, electronic devices include but are not limited to laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices can also represent various forms of mobile devices and other similar computing devices. The components shown herein, their connections and relationships, and their functions are only examples and are not intended to limit the implementation of the embodiments of the present application described and / or required herein.

[0129] In the embodiment of the present application, the processor can be implemented by using at least one of an application-specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a processor, a controller, a microcontroller, a microprocessor, and an electronic unit designed to perform the functions described herein. In some cases, such an implementation can be implemented in a controller. For software implementation, implementations such as processes or functions can be implemented with separate software modules that allow execution of at least one function or operation. The software code can be implemented by a software application (or program) written in any appropriate programming language, and the software code can be stored in a memory and executed by a controller.

[0130] In addition, the electronic device includes some functional modules not shown, which will not be described in detail here.

[0131] Those skilled in the art will appreciate that the various aspects of the electronic device provided by the present application may be implemented as a system, method or program product. Therefore, the various aspects of the present disclosure may be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or an implementation combining hardware and software, which may be collectively referred to as a "circuit", "module" or "system" herein.

[0132] The present application also provides a storage medium, in which a program product capable of implementing a method for generating rapid decisions for an electric-gas integrated energy system based on a neural network is stored. In some possible implementations, various aspects of the present disclosure may also be implemented in the form of a program product, which includes a program code, and when the program product is run on a terminal device, the program code is used to cause the terminal device to execute the steps described in the above “Exemplary Method” section of this specification according to various exemplary implementations of the present disclosure.

[0133] The storage medium can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0134] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for rapid decision making of an electric-gas integrated energy system based on a neural network, characterized in that the steps include: S1. Collect natural gas system data, gas unit safety domain data and power system data of the same electric-gas integrated energy system at different time periods, and construct a multi-module neural network training set and power system data set after normalization; S2. Construct a feedback-compensation multi-module neural network and train it using the multi-module neural network training set to obtain a safety domain prediction model; S3. Input the natural gas system data into the safety domain prediction model, calculate the actual output boundary value of the gas unit according to the output safety domain prediction value of the gas unit and the inherent output boundary value of the gas unit, splice the actual output boundary value with the power system data, and construct a penalty function neural network training set; S4. construct a penalty function neural network and train it using the penalty function neural network training set to obtain an optimal decision generation model; S5. Input the natural gas system data to be decided into the safety domain prediction model, use its output to calculate the actual processing boundary value of the gas group, and then input it together with the power system data into the optimal decision model to generate the optimal scheduling decision for the power system.

2. The method for rapid decision making of an electric-gas integrated energy system according to claim 1, characterized in that: In step S1, a multi-module neural network training set is constructed based on the normalized natural gas system data and gas unit safety domain data, and a power system data set is constructed based on the normalized power system data; Among them, the natural gas system data includes the gas source pressure, compressor boost pressure, and natural gas load in the natural gas system; The safety domain data of the gas unit includes the upper and lower output limits for safe operation of the gas unit; Power system data includes power load and wind power output data in the power system.

3. The method for rapid decision making of an electric-gas integrated energy system according to claim 1, characterized in that: In step S2, the feedback-compensation multi-module neural network includes a prediction module, a feedback module and a compensation module, wherein: The prediction module includes three single-layer perceptrons with activation functions and one multi-layer perceptron. After the gas source pressure, natural gas load and compressor boost in the natural gas system data are input into the three single-layer perceptrons respectively, the single-layer perceptron extracts features, and then the multi-layer perceptron is used to obtain the preliminary estimated value of the safety domain of the gas unit. The feedback module includes a multi-layer perceptron. The feedback module receives the preliminary estimated value of the safety domain of the gas unit obtained by the prediction module and the corresponding natural gas system data, extracts data features through the multi-layer perceptron, and generates the upper and lower limit correction values ​​of the safe output of the gas unit. The compensation module includes a multi-layer perceptron, which is used to capture the fluctuations in the safety domain of the gas unit caused by the overall changes in the natural gas system data. It receives the total gas source pressure data, the total natural gas load data, and the total compressor boost data, and generates the upper and lower limit compensations for the safe output of the gas unit, which is used to compensate for the fluctuations in the safety domain of the gas unit caused by the changes in the natural gas system input. By combining the prediction module, feedback module and compensation module, a feedback-compensation multi-module neural network is constructed to realize the calculation of the safety domain prediction value of the gas unit.

4. The method for rapid decision making of an electric-gas integrated energy system according to claim 1, characterized in that: In step S3, the actual output boundary calculation formula of the gas unit is: In the formula, It is the upper limit output of the actual output boundary value of the gas unit; It is the lower limit output of the actual output boundary value of the gas unit; It is the upper limit output in the safety range prediction value of the gas unit; It is the lower limit output in the safety range prediction value of the gas unit; It is the upper limit output of the inherent output boundary value of the gas unit; It is the lower limit output of the inherent output boundary value of the gas unit.

5. The method for rapid decision making of an electric-gas integrated energy system according to claim 4, characterized in that: In step S1, the power system data set is represented as ,in is the power load, Provide power for wind turbines; In step S3, the penalty function neural network training set is expressed as .

6. The method for rapid decision making of an electric-gas integrated energy system according to claim 1, characterized in that: Step S4 specifically includes: establishing a power system economic dispatch model and a penalty function neural network, based on the objectives and constraints defined by the power system economic dispatch model, using a penalty function neural network training set to train the penalty function neural network to approach the optimal solution in a data-driven manner, and finally forming a decision generation model; The economic dispatch model of the power system includes the output constraints of gas-fired units, the output constraints of coal-fired units and the power balance constraints of source and load. The expression of the economic dispatch model of the power system is: In the formula, For gas units At the moment contribution; For coal-fired units At the moment contribution; For wind turbines At the moment contribution; For power load At the moment size; For gas units At the moment The upper limit output within the actual output boundary value; For gas units At the moment The lower limit output of the actual output boundary value; and For coal-fired units The upper and lower boundaries of the output; , , and They represent the sets corresponding to coal-fired units, gas-fired units, wind turbine units and power loads respectively; The objective function is to minimize the power generation cost of coal-fired units. The objective function is; In the formula, , , are the coefficients of the quadratic term and the linear term and the constant term corresponding to the cost function of the coal-fired unit; is the total scheduling period.

7. The method for rapid decision making of an electric-gas integrated energy system according to claim 6, characterized in that: In step S4, the expression of the penalty function neural network is: ； In the formula, The output of the gas unit; The output of coal-fired units; A function for calculating power system dispatch decisions; Represents the model parameters of the penalty function neural network, including the weights and biases of each layer of neurons; The penalty function neural network loss function expression is: In the formula, is the weight factor of the loss term of the power balance constraint; is the weight factor of the loss term of the gas unit output constraint; It is the weight factor of the loss term of the coal-fired unit output constraint.

8. A neural network-based rapid decision-making system for an electric-gas integrated energy system, characterized in that: A method for realizing a rapid decision-making process for an electric-gas integrated energy system as claimed in any one of claims 1 to 7, comprising: The data collection and processing module is used to collect the natural gas system data, gas unit safety domain data and power system data of the same electric-gas integrated energy system at different time periods, and construct a multi-module neural network training set and power system data set after normalization processing; A feedback-compensation multi-module neural network construction and training module is used to construct a feedback-compensation multi-module neural network and train it using a multi-module neural network training set to obtain a safety domain prediction model; A gas unit actual output boundary value calculation module, used to calculate the gas unit actual output boundary value; A penalty function neural network training set construction module is used to splice the actual output boundary value of the gas unit with the corresponding power system data to construct a penalty function neural network training set; The penalty function neural network construction and training module is used to construct the penalty function neural network and train it using the penalty function neural network training set to obtain the optimal decision generation model; The collaborative operation module is used to input the natural gas system data to be decided into the safety domain prediction model, use its output to calculate the actual processing boundary value of the gas group, and then input it into the optimal decision model together with the power system data to generate the optimal scheduling decision for the power system.

9. An electronic device, characterized in that: It includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor is used to implement the steps of the method for rapid decision making of an electric-gas integrated energy system as described in any one of claims 1-7 when executing the computer program.

10. A storage medium, characterized in that: A computer program is stored on the storage medium, and when the computer program is executed by the processor, the steps of the method for rapid decision making of an electric-gas integrated energy system as described in any one of claims 1-7 are implemented.

Images