A method for optimal scheduling of an electric-thermal coupled system with uncertain renewable resources based on constraint learning

Abstract

A kind of optimization scheduling method of electric-thermal coupling system containing uncertainty renewable resource based on constraint learning, including constructing electric-thermal coupling system structure model, considering the uncertainty of renewable energy power generation power simultaneously;Establish optimization scheduling problem model containing objective function and constraint condition, adopt opportunity constraint to process renewable energy uncertainty, constraint condition involves power system, heat network and electric-thermal coupling and multiple constraints;Train neural network model, for predicting the constraint violation of transmission line under different electric-thermal supply and demand combinations;Based on neural network output, through iterative redundant constraint screening algorithm, necessary transmission constraint is dynamically identified and added, redundant constraint is reduced, to reduce the scale of optimization problem and improve solving efficiency.The present application effectively deals with the influence of renewable energy uncertainty on the system, simplifies the optimization scheduling model and significantly improves the solving speed, providing an efficient and feasible technical solution for the lightweight energy management of electric-thermal coupling system.

Description

Technical Field

[0001] This invention relates to renewable energy and power system optimization scheduling technology, and in particular to an optimization scheduling method for an uncertain renewable energy electrothermal coupling system based on constraint learning. Background Technology

[0002] With the increasing penetration of renewable energy sources such as wind and solar power into the power grid, the uncertainty of their power generation poses a significant challenge to the optimal scheduling of the power system. However, traditional optimal scheduling methods for coupled electro-thermal systems often employ model-driven approaches, neglecting the dynamic coupling of the heat network: the dynamic characteristics of the heat network are not effectively coordinated with the power grid, affecting overall energy efficiency. Furthermore, because the power generation of renewable energy is significantly affected by natural conditions (such as solar intensity and wind speed), its output power exhibits considerable uncertainty and volatility, making accurate prediction difficult. This introduces more complex variables and constraints into the optimal scheduling problem of the power system, significantly increasing the difficulty and time consumption of the solution. In practical applications, it is difficult to respond quickly and accurately to sudden changes in electricity demand. Traditional optimization methods rely on predefined models and assumptions, which cannot adapt to the requirements of modern energy systems with high dynamism and volatility. Therefore, it is urgent to reconstruct and simplify the existing power grid energy management problem to achieve lightweight and efficient management of the power system while ensuring accuracy.

[0003] To address these challenges, advanced artificial intelligence algorithms, such as deep learning, have seen rapid development in recent years and have been widely applied. They have proven their ability to handle high-dimensional, time-varying, and nonlinear data. Deep learning, by simulating a structure and function similar to the human brain's neural networks, learns features by abstracting them layer by layer through a multi-layered structure. It can automatically adjust internal parameters based on training errors and fit complex input-output relationships, thus providing guidance for processing new data. Therefore, deep learning possesses powerful data mining and feature extraction capabilities, making it suitable for solving complex optimization scheduling problems in multidimensional heterogeneous energy systems. However, the high dimensionality and depth of neural networks introduce a large number of integer variables into the optimization problem due to their encoded mixed-integer constraints, causing the curse of dimensionality and scalability issues, making it difficult to solve the optimization problem. Furthermore, some researchers hope to reduce the size of the optimization problem while maintaining accuracy by preprocessing the optimization problem to identify and remove redundant constraints and variables. The drawback is that this method fails to consider application scenarios of energy systems with uncertain renewable energy access, and still faces problems such as excessively large model size and low solution efficiency when dealing with high-dimensional complex energy systems. Although some existing studies have reduced the size of the optimization scheduling model through pre-classification and identification and elimination of redundant security constraints, the performance is limited, and it fails to incorporate the impact of current uncertainties in renewable energy generation.

[0004] The proportion of renewable energy connected to the grid is increasing, but its power generation is significantly affected by natural conditions, exhibiting considerable uncertainty and volatility, making accurate prediction difficult. This introduces more complex variables and constraints into the optimization and scheduling problem of integrated energy systems, greatly increasing the difficulty and time consumption of the solution. In practical applications, it is difficult to respond quickly and accurately to sudden changes in electricity demand. However, traditional optimization methods rely on predefined models and assumptions, which cannot meet the requirements of modern energy systems with high dynamism and volatility. Therefore, it is urgent to reconstruct and simplify the existing energy management problem of integrated energy systems, so as to achieve lightweight and efficient management of the system while ensuring accuracy.

[0005] Furthermore, existing research incorporates opportunity constraints to cover the impact of uncertainties in renewable energy sources within the system. However, the introduction of opportunity constraints increases the complexity of solving the optimization scheduling problem. Existing optimization problem preprocessing methods primarily identify and remove redundant constraints and variables in the optimization problem beforehand, thereby reducing the problem size and improving the solution speed while maintaining accuracy. However, these methods fail to consider application scenarios of energy systems with uncertain renewable energy access, and still face problems such as excessively large model size and low solution efficiency when dealing with high-dimensional complex energy systems. For integrated energy systems including thermal systems, the heat network model is also affected by renewable energy injected into the grid, and some existing heat networks do not consider these factors, resulting in significant errors.

[0006] It should be noted that the information disclosed in the background section above is only for understanding the background of this application, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0007] The main objective of this invention is to overcome the deficiencies in the aforementioned background technology and provide an optimization scheduling method for electrothermal coupling systems with uncertain renewable resources based on constraint learning. While ensuring scheduling accuracy, this method effectively improves the optimization scheduling solution efficiency of electrothermal coupling systems with a high proportion of uncertain renewable energy, thereby achieving lightweight, online, and efficient energy management.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] An optimal scheduling method for an electrothermal coupling system of renewable resources with uncertainties, based on constraint learning, includes the following steps:

[0010] S1. Construct a structural model of the electrothermal coupling system, including a power generation / heat generation unit model, a heat network model, and a transmission network model, and consider the uncertainty of renewable energy power generation.

[0011] S2. Establish an optimization scheduling problem model for the electrothermal coupling system, including the objective function and constraints, wherein the constraints include power system constraints, heating network constraints and electrothermal coupling constraints, and use opportunity constraints to handle the uncertainty of renewable energy.

[0012] S3. Train a neural network model to predict the constraint violations of transmission lines under different combinations of power and heat supply and demand.

[0013] S4. Based on the output of the neural network model, an iterative redundancy constraint filtering algorithm is used to dynamically identify and add necessary transmission constraints, reduce redundant constraints, thereby reducing the scale of the optimization problem and improving the solution efficiency.

[0014] A computer program product includes a computer program that, when executed by a processor, implements the constraint learning-based optimization scheduling method for an uncertain renewable resource electrothermal coupling system.

[0015] The present invention has the following beneficial effects:

[0016] This invention proposes an optimal scheduling method for electrothermal coupled systems with uncertainties in renewable resources based on constraint learning. It designs a redundancy constraint reduction scheme based on neural networks, which can realize lightweight energy management of electrothermal coupled systems under the condition of uncertainty in renewable energy generation, improve the identification efficiency of redundancy constraints, reduce the scale of power system optimal scheduling problems, and thus improve the solution speed.

[0017] To address the uncertainties in renewable energy generation, this invention employs opportunity constraints to encode these uncertainties, thereby encompassing the impact of renewable energy generation prediction errors on the power and thermal systems and reducing the error between the model and the actual situation. Simultaneously, it utilizes generation prediction error data samples to reconstruct and linearize opportunity constraints, simplifying the optimization scheduling problem model and reducing the solution difficulty. Regarding redundancy constraint identification, this invention designs a deep learning-based variable preprocessing method and trains a neural network, Vio_MLP, to predict line power flow constraint violations under different power and heat supply and demand conditions. Compared to traditional iterative redundancy constraint reduction methods, this neural network can quickly complete predictions, especially for large-scale, complex, high-dimensional integrated energy systems with numerous lines. It can further improve the efficiency of redundancy constraint identification while maintaining solution accuracy, reducing the scale of the power system optimization scheduling problem and accelerating the solution speed. Furthermore, this invention applies the redundancy constraint reduction method to the uncertainty scenarios of high-dimensional, multi-heterogeneous energy systems, not only improving the performance in handling complex constraints but also expanding the application scope of constraint reduction methods, providing a new solution for lightweight power system management, and enriching research on power system energy management.

[0018] Other beneficial effects of the embodiments of the present invention will be further described below. Attached Figure Description

[0019] Figure 1 This is a flowchart of the overall process of the present invention, which is a method for optimizing the scheduling of an electrothermal coupling system of renewable resources with uncertainties based on constraint learning.

[0020] Figure 2 This is a flowchart of the redundancy constraint screening iteration based on a neural network according to an embodiment of the present invention. Detailed Implementation

[0021] The embodiments of the present invention will be described in detail below. It should be emphasized that the following description is merely exemplary and not intended to limit the scope and application of the present invention.

[0022] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of embodiments of the present invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0023] This invention proposes an optimal scheduling method for electrothermal coupled systems with uncertainties in renewable resources based on constraint learning. It designs a redundancy constraint reduction scheme based on neural networks, aiming to achieve lightweight energy management of electrothermal coupled systems considering the uncertainties of renewable energy generation, improving the efficiency of redundancy constraint identification, reducing the scale of the power system optimal scheduling problem, and increasing the solution speed. This invention applies the redundancy constraint reduction method to uncertain scenarios in high-dimensional, multi-heterogeneous energy systems, improving the performance in handling complex constraints, expanding the application scope of constraint reduction methods, and enriching research on power system energy management.

[0024] See Figure 1 This invention provides an optimal scheduling method for an electrothermal coupling system of renewable resources with uncertainties based on constraint learning, comprising the following steps:

[0025] Step S1: Construct a structural model of the electrothermal coupling system, including a power generation / heat generation unit model, a heat network model, and a transmission network model, and consider the uncertainty of renewable energy power generation.

[0026] In some embodiments, the construction of the power generation / heat generation unit model in step S1 includes: modeling the deviation between the predicted and actual values ​​of renewable energy power generation, representing it as a random variable with mean and covariance characteristics; allocating the total mismatch of renewable energy power generation proportionally to each conventional generator unit through participation factors to maintain system power balance; performing convex combination modeling on the feasible region of electro-thermal coupling of the cogeneration unit to describe the correlation between its power generation and heating power; and modeling the electric boiler including the linear conversion relationship between its electric power consumption and thermal power output, and considering its conversion efficiency.

[0027] In some embodiments, the construction of the heat network model in step S1 includes: using the generalized phasor method to transform the dynamic characteristics of the heat network pipelines, the node temperature mixing equation, and the temperature continuity equation from the time domain to the frequency domain; simulating the temperature decay and phase delay effects during pipeline transmission through an exponential decay term; establishing the heat power balance equation of the heat network nodes, considering the heat source injection, heat load demand, and the heat balance relationship of pipeline inflow and outflow; and converting the historical heat load data of the day before the scheduling date into phasor form through Fourier transform as the initial state input of the heat network.

[0028] In some embodiments, the construction of the transmission network model in step S1 includes: using a DC power flow approximation model to quickly calculate the line power flow distribution through the power transmission distribution factor matrix; the line power flow is expressed as a linear combination of power generation, load power, and renewable energy power generation deviation; only the uncertainty of renewable energy power generation is explicitly modeled, and the load power is regarded as a deterministic variable.

[0029] Step S2: Establish an optimization scheduling problem model for the electrothermal coupling system, including the objective function and constraints. The constraints include power system constraints, heating network constraints, and electrothermal coupling constraints, and opportunity constraints are used to handle the uncertainty of renewable energy.

[0030] In some embodiments, establishing the optimization scheduling problem model in step S2 includes: constructing a multi-objective function that includes minimizing the total operating cost of day-ahead scheduling and minimizing the adjustment amount of generator units in real-time scheduling, and transforming the multi-objective function into a single-objective optimization problem through a linear weighted normalization method; introducing opportunity constraints into the constraints to handle the uncertainty of renewable energy, including opportunity constraints on the upper and lower limits of generator output and opportunity constraints on line transmission power; transforming the opportunity constraints into a deterministic equivalent form through first-order Taylor expansion and inverse cumulative distribution function; introducing heat balance opportunity constraints on the heating network side to ensure that the heat load demand is met under a given confidence level, and handling the impact of uncertainty through robust margin calculation.

[0031] Step S3: Train a neural network model to predict the constraint violations of transmission lines under different combinations of power and heat supply and demand.

[0032] In some embodiments, training the neural network model in step S3 includes: constructing a multilayer perceptron model with the power injected into the grid nodes as input and the degree of violation of transmission constraints of each line as output; in the calculation of the degree of violation of line constraints, combining the parameters after the reconstruction of the opportunity constraints, including the line power flow term, the offset term and the margin term; using the modified linear unit as the activation function of the hidden layer, and learning the weight matrix and bias parameters through training to establish a nonlinear mapping relationship from the generation dispatch scheme to the line constraint violation situation.

[0033] Step S4: Based on the output of the neural network model, an iterative redundancy constraint filtering algorithm is used to dynamically identify and add necessary transmission constraints, reduce redundant constraints, thereby reducing the scale of the optimization problem and improving the solution efficiency.

[0034] In some embodiments, the redundancy constraint screening algorithm in step S4 includes: firstly solving a relaxed optimization problem without transmission constraints to obtain the generator output scheme; inputting the output scheme into a trained neural network model to predict the degree of constraint violation for each line; determining which transmission constraints are violated based on whether the degree of violation exceeds a threshold; selecting the transmission constraints corresponding to several lines with the highest degree of violation and adding them to the optimization model; iteratively solving the optimization problem and adding necessary constraints until all transmission constraints are satisfied, and outputting the final optimized scheduling scheme.

[0035] In some embodiments, the redundant constraint screening algorithm replaces the traditional iterative calculation process with neural network prediction, dynamically identifies the most critical violated constraints, and gradually constructs the minimum necessary constraint set, thereby significantly reducing the complexity of the problem while ensuring the accuracy of the solution.

[0036] In some embodiments, the method further includes: using a mathematical programming solver to solve the optimization problem in each iteration, and dynamically updating the constraint set during the iteration process, so as to finally obtain an optimized scheduling scheme that satisfies all safety constraints.

[0037] The following further describes specific embodiments of the present invention and examples of its algorithm implementation.

[0038] See Figure 1 and Figure 2 The proposed optimal scheduling method for electro-thermal coupled systems includes a neural network-based redundancy constraint reduction approach. The main contributions are as follows: Opportunity constraints are used to cover the impact of renewable energy generation prediction errors on the electro-thermal coupled system, and these constraints are reconstructed and linearized using generation prediction error data samples, simplifying the optimal scheduling problem model. A deep learning-based variable preprocessing method is designed, training a Vio_MLP neural network to predict line power flow constraint violations under different electro-thermal supply and demand conditions, improving the efficiency of redundant constraint identification, reducing the scale of the power system optimal scheduling problem, and increasing the solution speed. The redundancy constraint reduction method is applied to uncertain scenarios in high-dimensional, multi-heterogeneous energy systems, improving the performance in handling complex constraints, expanding the application scope of the constraint reduction method, and providing a new solution for lightweight power system management.

[0039] Specifically, the method of the present invention includes the following steps:

[0040] (1) Modeling of the structure and composition of the electrothermal coupling system:

[0041] (1-1) Power generation / heat generation units:

[0042] This invention uses wind power as a representative of renewable energy sources, with Represents the set of wind power nodes. (Used) express middle Predicted node power This represents the actual value, and the deviation between the two is... The mean is covariance matrix The relevant expressions are as follows:

[0043]

[0044]

[0045] in, The corresponding predicted operating point is assumed to be the balanced operating point that satisfies the nodal power balance equation. This indicates that the actual renewable energy power generation is greater than the predicted value. Conversely. , These represent the total power mismatch caused by volatility and its standard deviation, respectively, in vector form. It is a by A row vector with n elements, where each element has a value of 1.

[0046] Except for the wind turbines, the power for the electrothermal coupling system is supplied by controllable thermal power generating units, including CHP units and non-CHP coal-fired generating units. The set of all coal-fired generating nodes is as follows: Due to the uncertainties of wind power, for any node in the power system, in order to maintain power balance, wind power prediction errors require controllable generating nodes in the system to adjust their generating power to absorb them. Total generating power mismatch. By participation factor This power is allocated to each conventional generator. Therefore, the power generation expression for each thermal power generator node is as follows. The definition is:

[0047]

[0048]

[0049] in , These correspond to thermal power units under conditions of no uncertain power generation prediction error. And in the case of prediction error The output of a conventional generator, This is the upper limit of the output of thermal power units.

[0050] A combined heat and power unit (CHP) is the electrical and thermal coupling point in an integrated energy system, providing the system with thermal energy and a portion of electrical energy. Its feasible region is expressed mathematically as follows:

[0051]

[0052]

[0053]

[0054] in, and These refer to the power generation and heating capacity of the CHP unit, respectively. The first in the feasible region One vertex, For convex combination coefficients, Let be the total number of vertices in the feasible region. This is the set of all node numbers in the CHP unit.

[0055] This invention also employs an electric boiler (EB) to assist the CHP unit in heating. The electrical power required by the electric boiler is provided by thermal power generating units in the power grid, and the relationship between its heating power and electrical power consumption is as follows:

[0056]

[0057] in, and Electric boiler units Power generation and heating capacity, For boiler efficiency, This is a set of node numbers for all electric boiler units.

[0058] (1-2) Hot Network:

[0059] The heating system operates in a constant flow mode. Its structure includes a heat source, heat load, supply / return water pipes, and booster pumps, flow control valves, and other accessories on the branches. The heat source is provided by a combined heat and power (CHP) unit and an electric boiler. All water supply pipes in the heating network are numbered according to a system named... The set, This is a set of node numbers in the heating network. These are sets of numbers for heat source nodes and heat load nodes in the heating network. Indicates the number of the heating network node; They are nodes A collection of connected heat sources and heat loads. These are the inflow and outflow nodes, respectively. The set of branch numbers.

[0060] Due to the dynamic characteristics of heating networks, their mathematical models require accurate description using partial differential equations. However, commercial solvers struggle to handle optimization scheduling problems in the form of partial differential equations. Furthermore, the mainstream approximation method, the nodal method, introduces significant errors. Therefore, this invention employs the generalized phasor method to establish a heating system model with smaller errors and solvability. This includes the frequency domain equations of the dynamic characteristics of the heating network pipes, the frequency domain expression of the temperature mixing equations at the heating network nodes after processing using the generalized phasor method (GPM), and the frequency domain form of the heating network temperature continuity equation after GPM processing.

[0061]

[0062] in, The specific heat capacity of water, The density of water; These are the numbers in the heating network. The cross-sectional area of ​​the water supply pipe, the mass flow rate of the water, and the heat dissipation coefficient of the pipe; The difference between water temperature and ambient temperature, i.e., relative temperature; The tables represent the temperature vectors at both ends of the pipe. This is the fundamental frequency. These are heat source nodes. Heat load nodes The phasor of thermal power; It is a node Temperature phasor.

[0063] (1-3) Transmission network:

[0064] use Let represent the sets of nodes and branches in a DC transmission network, respectively. coal-fired power generating units Wind turbine node set satisfy The set of all load nodes is Since load curves are relatively predictable, this paper does not consider load uncertainties. The power flow expressions for each line in the power system are as follows:

[0065]

[0066] Where the matrix It is a matrix of power transfer distribution factors (PTDFs), expressed as follows:

[0067]

[0068] in and These represent the susceptance matrix of the transmission line and the susceptance matrix of the nodes, respectively. The zero row and zero column correspond to the row and column of the reference node.

[0069] (2) Model for the optimal scheduling problem of an electro-thermal coupled system:

[0070] (2-1) Objective function:

[0071] The objective functions are divided into two categories: day-ahead dispatch and real-time dispatch. Day-ahead dispatch refers to planning the operation of the power system, typically one or several days in advance, based on historical operating data and predicted load demand and generation resources. This includes the switching status and power output of each generating unit, to meet specific operational objectives. The optimization objective of the day-ahead dispatch problem constructed in this method is to minimize the total system operating cost, including the cost of coal-fired CHP units and non-CHP coal-fired generating units. Its mathematical expression is as follows:

[0072]

[0073] in, yes At that moment, the The cost function of non-CHP coal-fired power generating units at each node is typically modeled as a quadratic function, where It is a coefficient, which can be obtained from the measurement data of the heat rate curve. This is the cost function for a coal-fired CHP unit.

[0074] The day-ahead dispatch results provide a reference for the daily operation plan of power system units. However, in actual engineering practice, the daily renewable energy generation situation often fluctuates due to changes in weather conditions, etc. Therefore, real-time optimized dispatch of the power system is also required to quickly respond to system disturbances. Because adjusting the upward / downward reserve capacity of generator units incurs additional operating costs, real-time optimized dispatch not only needs to meet the goal of minimizing the system's power generation cost, but also needs to include the minimum day-ahead and intraday generator unit output deviation in the optimization scope. The corresponding mathematical expression is as follows:

[0075]

[0076] in, Let these represent the objectives of minimizing total power generation cost and minimizing generator unit adjustment power, respectively, and give the variables... Add the superscript 's' to indicate real-time output.

[0077] To reduce the difficulty of solving the problem, this invention uses a linear weighting method to normalize the two objectives. Both objectives have the same weight coefficient of 0.5, representing the same level of importance.

[0078]

[0079] in , These represent the maximum real-time system power generation cost and the maximum day-to-day generator output deviation, respectively. After processing, the objective function... It becomes a dimensionless number.

[0080] (2-2) Constraints:

[0081] The constraints of the optimal scheduling problem of electrothermal coupled systems include power system constraints, heating network constraints, and electrothermal coupling constraints.

[0082] 1) For power systems:

[0083] Each node in a power system needs to satisfy power balance constraints:

[0084]

[0085] Furthermore, the output of generator units in the power system must be within constraints, and the output variation of thermal power units between two adjacent time periods must be subject to ramping constraints. These represent the minimum and maximum output of the thermal power unit, and the downward and upward ramp power limits of the thermal power unit.

[0086]

[0087]

[0088] Meanwhile, this method uses chance constraints to represent the impact of day-ahead forecasting errors in renewable energy generation on the optimal scheduling problem, including upper and lower limits for generation capacity and transmission capacity constraints. However, introducing chance constraints can also make the optimal scheduling problem difficult to solve. Therefore, this method performs a deterministic reconstruction of the relevant chance constraints, making them easily processable by existing optimization solvers.

[0089] The deterministic reconstruction in this paper is based on the following statement: uncertainty The impact is simulated using a first-order Taylor expansion near the predicted operating point. When the prediction error is relatively small, the prediction can be approximated as accurate. Generally, the true distribution of uncertain fluctuations is unknown, but some distribution characteristics can be obtained through historical datasets, for example... At time, its mean is The covariance matrix is Generally known: .definition for The set of possible distributions, Distributions The probability distribution function and inverse cumulative distribution function are given. The chance constraint expression is as follows:

[0090]

[0091]

[0092]

[0093] in, , Representing the distributions respectively exist At that moment, and The inverse cumulative distribution function at that location. It is an identity matrix. Vector. It is a A row vector with n elements, where each element has a value of 1. The impact of uncertainty, which leads to a reduction in available power generation capacity, is called the uncertainty margin. Alternatives. Let represent the upper and lower limits of generator capacity constraints, respectively, and let be the transmission constraints of the transmission line. The probabilities of violation are respectively... , Opportunity constraints allow energy systems to operate with a given probability of violation, providing an intuitive way to limit the risks posed by uncertainty. For connecting nodes The power transmission limit of the branch.

[0094] 2) Thermal system constraints:

[0095] In addition to the established heat network flow constraints, the thermal system also needs to consider inverse Fourier transform constraints of the heating plan, initial condition constraints, and upper and lower operating limit constraints, etc.

[0096]

[0097]

[0098]

[0099] Solving the heating network scheduling problem requires determining the initial state of the heating system. Due to thermal inertia, in addition to considering the predicted heat load data for the scheduling day, historical heat load data and historical heating temperature data from all heat sources also need to be considered. The time lag of a large-scale heating system is approximately several hours; therefore, this invention selects historical data from the day before the scheduling date as a reference. When modeling the heating network, this invention uses the General Phase Method proposed in other literature to represent each variable; therefore, the original heat load data also needs to be transformed into phasor form through Fourier transform. Represents the scheduling time interval of the electrothermal coupling system ,but Indicates the first One scheduling moment; For heat load nodes In the The power value at each scheduling moment. and These represent the time of the scheduling day and the day before the scheduling, respectively. , These are nodes respectively. Upper and lower temperature limits.

[0100] Furthermore, due to the coupling between the power system and the thermal system, the uncertainty of the injected power from wind turbines can also affect the heating network. This invention introduces opportunity constraints into the thermal balance constraints to address this issue at a certain confidence level. The following measures will ensure that the heat load demand is met:

[0101]

[0102] The thermoelectric conversion coefficient of the CHP unit is: , As the system's baseline capacity, This is the standard deviation of the equivalent disturbance generated on the CHP power output after the wind power prediction error is linearized. Let be the wind power error covariance matrix. This is the robust scaling factor. , These represent the total disturbance standard deviation and the corresponding robustness margin for the CHP unit, respectively.

[0103] (3) Redundancy constraint selection algorithm:

[0104] To further improve the solution speed and reduce the scale of the optimization scheduling problem, after representing and reconstructing the uncertainty of renewable power sources using chance constraints, this invention employs a neural network combined with an iterative constraint selection method. (See [reference needed]). Figure 2 Redundant constraints and variables in the optimization scheduling problem of electrothermal coupling systems are screened.

[0105] (3-1) Training a neural network based on constraint learning:

[0106] In practical engineering, line overloads corresponding to violations of transmission line constraints are permissible for short periods or when the violations are minor. However, if the output of conventional generators cannot exceed or fall below their generating capacity limits in practice, the power imbalance will exceed the power system's balancing capacity, leading to serious accidents. Therefore, based on this understanding, the constraint reduction method of this invention aims to reduce unnecessary line transmission constraints. The basic idea is as follows: First, a commercial solver (GUROBI) is used to solve the electro-thermal coupling system optimization scheduling problem without soft constraints. A trained neural network (named Vio_MLP) is used to predict violations of line transmission constraints. Then, necessary line transmission constraints are added based on the MLP output, thereby eliminating redundant transmission constraints. The following details the neural network structure and the constraint violation identification method:

[0107] 1) Expression for the degree of violation of transmission constraints

[0108] For different power supply and demand combinations, a formula needs to be defined to compare whether transmission constraints are violated on each line and the degree of violation. Let For connecting nodes If the transmission constraints of the line are violated, This indicates the line. If the transmission power on the line does not violate the transmission constraints, then the transmission power on the line is overloaded. It is the set of all transmission constraint violations in the power grid.

[0109] in, The following is derived from the opportunity constraint expression of the power system:

[0110]

[0111]

[0112]

[0113] 2) Structure of Vio_MLP

[0114] This paper trains a Vio_MLP to describe the mapping from historical supply and demand data to transmission constraint violations. This neural network has one input layer. There are one hidden layer and one output layer. Each neuron includes a linear mapping and a non-linear activation function; this paper uses the Corrected Linear Unit (ReLU) as the activation function. Its input is... This represents the power of each node in the power grid, including historical data for each generator node and load node. Output A quantitative representation of the violation of transmission constraints on each line:

[0115]

[0116]

[0117] in All are row vectors, with each element arranged in the order of node and line numbers.

[0118] Use symbols For hidden layers After numbering, the regression MLP expression is as follows:

[0119]

[0120] in , These are hidden layers Linear function and output, parameters , These are hidden layers The weight matrix and biases need to be obtained through training.

[0121] The trained neural network can predict the violation of transmission constraints on each line under this condition based on the injected / output power of the power grid nodes.

[0122] (3-2) Redundancy transmission constraint selection algorithm:

[0123] This invention, considering the problem background, employs a neural network to predict line power flow constraint violations under different generator output conditions, thereby replacing the time-consuming steps of traditional iterative methods and obtaining an improved redundancy constraint selection algorithm. The specific steps are shown in Table 1. Specifically, first, the optimal scheduling problem of a relaxed electrothermal coupling system without any transmission constraints is solved, and the obtained generator power combination results are used as input data for the neural network. The degree of transmission constraint violation on each line in the power grid can be obtained by Vio_MLP fitting. If all lines are within the safe operating range, the iteration ends and the output of each conventional generator unit is output as the final result; otherwise, the line with the highest violation degree is selected. indivual( (The value is a constant, set according to the power grid scale). The transmission constraints corresponding to the lines are added to the original relaxation problem, and the problem is solved again. Repeat the above steps until a solution obtained satisfies all line transmission constraints.

[0124] Table 1

[0125]

[0126] In summary, this invention proposes an optimal scheduling method for electro-thermal coupled systems with uncertainties in renewable resources based on constraint learning. It applies a renewable energy generation uncertainty encoding representation method to electro-thermal coupled systems, using chance constraints to represent the impact of renewable energy generation prediction errors on the power and thermal systems, thus reducing the error between the model and the actual problem situation. Simultaneously, this invention reconstructs and linearizes chance constraints using generation prediction error data samples, simplifying the optimal scheduling problem model and reducing the solution difficulty. Furthermore, compared to traditional iterative redundancy constraint reduction methods, the Vio_MLP neural network trained by this invention, combined with constraint learning, can quickly predict line flow constraint violations under different grid supply and demand conditions. Especially for large-scale, high-dimensional, complex integrated energy systems with a large number of lines, it can further improve the efficiency of redundancy constraint identification while maintaining solution accuracy, reducing the scale of the power system optimal scheduling problem, and increasing the solution speed. This invention applies the redundancy constraint reduction method to uncertain scenarios in high-dimensional, multi-heterogeneous energy systems, improving the performance in handling complex constraints, expanding the application scope of constraint reduction methods, and enriching research on power system energy management.

[0127] This invention also provides a storage medium for storing a computer program, which, when executed, performs at least the methods described above.

[0128] This invention also provides a control device, including a processor and a storage medium for storing a computer program; wherein the processor executes the computer program by performing at least the method described above.

[0129] This invention also provides a processor that executes a computer program, at least performing the methods described above.

[0130] The storage medium can be implemented by any type of non-volatile storage device, or a combination thereof. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), magnetic random access memory (FRAM), flash memory, magnetic surface memory, optical disc or CD-ROM; magnetic surface memory can be disk storage or magnetic tape storage. The storage media described in the embodiments of this invention are intended to include, but are not limited to, these and any other suitable types of memory.

[0131] In the several embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0132] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0133] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0134] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0135] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0136] The methods disclosed in the several method embodiments provided by this invention can be arbitrarily combined without conflict to obtain new method embodiments.

[0137] The features disclosed in the several product embodiments provided by this invention can be arbitrarily combined without conflict to obtain new product embodiments.

[0138] The features disclosed in the several method or device embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new method or device embodiments.

[0139] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various equivalent substitutions or obvious modifications can be made without departing from the concept of the present invention, and all such modifications, achieving the same performance or application, should be considered within the scope of protection of the present invention.

Claims

1. A method for optimal scheduling of an electrothermal coupling system of renewable resources with uncertainties based on constraint learning, characterized in that, Includes the following steps: S1. Construct a structural model of the electrothermal coupling system, including a power generation / heat generation unit model, a heat network model, and a transmission network model, and consider the uncertainty of renewable energy power generation. S2. Establish an optimization scheduling problem model for the electrothermal coupling system, including the objective function and constraints, wherein the constraints include power system constraints, heating network constraints and electrothermal coupling constraints, and use opportunity constraints to handle the uncertainty of renewable energy. Step S2 involves establishing an optimization scheduling problem model, including: constructing a multi-objective function that minimizes the total day-ahead scheduling operating cost and the real-time scheduling generator unit adjustment amount, and transforming the multi-objective function into a single-objective optimization problem; introducing opportunity constraints into the constraints to handle the uncertainties of renewable energy, including the upper and lower limits of generator output opportunity constraints and the line transmission power opportunity constraints; transforming the opportunity constraints into a deterministic equivalent form; and introducing heat balance opportunity constraints on the heating network side to ensure that the heat load demand is met under a given confidence level, and calculating and handling the impact of uncertainties. S3. Train a neural network model to predict the constraint violations of transmission lines under different combinations of power and heat supply and demand. Step S3 includes training the neural network model by constructing a neural network model with the power injected into the grid nodes as input and the degree of constraint violation of each line as output. In the calculation of the degree of line constraint violation, the parameters after the reconstruction of the opportunity constraint are combined, including the line power flow term, offset term and margin term. The activation function of the hidden layer is used to learn the weight matrix and bias parameters through training to establish a nonlinear mapping relationship from the generation dispatch scheme to the line constraint violation. S4. Based on the output of the neural network model, an iterative redundancy constraint filtering algorithm is used to dynamically identify and add necessary transmission constraints, reduce redundant constraints, thereby reducing the scale of the optimization problem and improving the solution efficiency.

2. The method for optimal scheduling of an electrothermal coupling system with uncertainties based on constraint learning according to claim 1, characterized in that, Step S1, which involves constructing the power generation / heat generation unit model, includes: The deviation between the predicted and actual values ​​of renewable energy power generation is modeled and represented as a random variable with mean and covariance characteristics; The total mismatch of renewable energy generation power is proportionally allocated to each conventional generator unit through participation factors in order to maintain system power balance; A convex combinatorial model is used to model the electrothermal coupling feasible region of the cogeneration unit to describe the relationship between its power generation and heating power. Modeling an electric boiler includes the linear conversion relationship between its electrical power consumption and thermal power output, and takes into account its conversion efficiency.

3. The method for optimal scheduling of an electrothermal coupling system with uncertainties based on constraint learning according to claim 1, characterized in that, Step S1, which involves constructing the heating network model, includes: The generalized phasor method is used to transform the dynamic characteristics of heating network pipelines, the nodal temperature mixing equation, and the temperature continuity equation from the time domain to the frequency domain. The temperature decay and phase delay effects during pipeline transmission are simulated using an exponential decay term. Establish the heat power balance equation for the nodes of the heating network, taking into account the heat balance relationship of heat source injection, heat load demand and pipeline inflow and outflow; The historical heat load data of the day before the scheduling date is converted into phasor form by Fourier transform and used as the initial state input of the heating network.

4. The method for optimal scheduling of an electrothermal coupling system with uncertainties based on constraint learning according to claim 1, characterized in that, Step S1, which involves constructing the transmission network model, includes: A DC power flow approximation model is adopted, and the power transmission distribution factor matrix is ​​used to quickly calculate the power flow distribution of the line. Line power flow is expressed as a linear combination of generating power, load power, and renewable energy generation deviation; Only the uncertainty of renewable energy power generation is explicitly modeled, while load power is treated as a deterministic variable.

5. The method for optimal scheduling of an electrothermal coupling system with uncertainties based on constraint learning according to claim 1, characterized in that, In step S2, The multi-objective problem is transformed into a single-objective optimization problem by using a linear weighted normalization method; The chance constraint is transformed into a deterministic equivalent form by using a first-order Taylor expansion and an inverse cumulative distribution function; The impact of uncertainty is addressed through robust margin calculation.

6. The method for optimal scheduling of an electrothermal coupling system with uncertainties based on constraint learning according to claim 1, characterized in that, In step S3, The neural network model is a multilayer perceptron model; Use the modified linear unit as the activation function for the hidden layer.

7. The method for optimal scheduling of an electrothermal coupling system with uncertainties based on constraint learning according to claim 1, characterized in that, The redundancy constraint filtering algorithm in step S4 includes: First, solve the relaxation optimization problem without transmission constraints to obtain the generator set output scheme; The power output scheme is input into the trained neural network model to predict the degree of constraint violation for each line; Determine which transmission constraints have been violated based on whether the degree of violation exceeds a threshold; Select the transmission constraints corresponding to the lines with the highest degree of violation and add them to the optimization model; Iteratively solve the optimization problem and add necessary constraints until all transmission constraints are satisfied, and output the final optimized scheduling scheme.

8. The method for optimal scheduling of an electrothermal coupling system with uncertainties based on constraint learning according to claim 7, characterized in that, The redundant constraint screening algorithm replaces the traditional iterative calculation process with neural network prediction, dynamically identifies the most critical violated constraints, and gradually constructs the minimum necessary constraint set, which significantly reduces the complexity of the problem while ensuring the accuracy of the solution.

9. The method for optimal scheduling of an electrothermal coupling system with uncertainties based on constraint learning according to claim 1, characterized in that, The method further includes: The mathematical programming solver is used to solve the optimization problem in each iteration, and the constraint set is dynamically updated during the iteration process to finally obtain an optimized scheduling scheme that satisfies all safety constraints.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the constraint learning-based optimal scheduling method for electrothermal coupling systems with uncertainties, as described in any one of claims 1 to 9.

Images