Distributed energy dispatching methods and related equipment that do not depend on the physical parameters of the distribution network

Abstract

This invention is primarily applied in the fields of power grid systems and neural network technology. It discloses a distributed energy dispatching method and related equipment that does not rely on the physical parameters of the distribution network. First, it acquires node power injection data of the distribution network system and inputs it into a pre-trained parameter prediction neural network. Then, the neural network generates parameters for a convex optimization substitution model describing the dispatch response characteristics of the distribution network. Next, based on these parameters and the physical constraints of the distributed energy, a convex quadratic programming problem model is constructed. Then, a pre-set convex optimization solver is called to solve the model, obtaining a sequence of power setpoints for each distributed energy source. Finally, these sequences are sent as dispatch commands to the corresponding controllers to achieve precise control of the distributed energy. This method can efficiently generate multi-time-period optimal dispatching strategies that strictly satisfy the physical constraints of the distribution network, significantly improving the system's economy and security.

Description

Technical Field

[0001] This invention relates to the fields of power grid systems and neural network technology, specifically to a distributed energy dispatching method and related equipment that do not depend on the physical parameters of the distribution network. Background Technology

[0002] With the high penetration of distributed energy resources (DERs), the distribution network is undergoing a profound transformation from "passive absorption" to "active regulation." The flexible adjustment capabilities of DERs, such as wind power, photovoltaics, energy storage, and flexible loads, provide considerable resources for optimal scheduling across multiple time periods. Through synergistic optimization in the spatiotemporal dimensions, the system's economy, security, and renewable energy absorption rate can be significantly improved. However, actual distribution networks face two fundamental constraints: unknown system models and data scarcity.

[0003] Under these dual constraints, existing technologies face a dilemma: model-based methods fail due to incomplete physical model information, while data-driven methods fail due to incomplete measurement data. However, physical constraints such as voltage and current in the distribution network must be strictly satisfied by the decision-making process. Therefore, there is an urgent need for an efficient decision optimization method that can still strictly adhere to physical constraints even when data is scarce and the complete system model is unknown. This has become a core technical challenge that urgently needs to be overcome in the field of active distribution network dispatching. Summary of the Invention

[0004] This invention provides a distributed energy dispatching method and related equipment that do not rely on the physical parameters of the distribution network. It can efficiently generate multi-time-period optimal dispatching strategies that strictly meet the physical constraints of the distribution network, significantly improving the system's economy and security.

[0005] This invention provides a distributed energy dispatching method that does not rely on the physical parameters of the distribution network, the method comprising: Obtain node power injection data of the distribution network system and input the node power injection data into a pre-trained parameter prediction neural network; The parameter prediction neural network generates model parameters for a convex optimization alternative model that describes the dispatch response characteristics of the power distribution network system. Based on the model parameters and the physical constraints reflecting the physical limits of distributed energy, a convex quadratic programming problem model is constructed. The convex quadratic programming problem model is solved by calling a preset convex optimization solver to obtain the power setpoint sequence for each distributed energy source. Each of the power setpoint sequences is sent as a scheduling instruction to the corresponding distributed energy controller to control the operation of each of the distributed energy sources.

[0006] Optionally, the step of constructing a convex quadratic programming problem model based on the model parameters and physical constraints reflecting the physical limits of distributed energy includes: The model parameters are divided into first parameters and second parameters; The objective function of the convex quadratic programming problem model is generated based on the first parameter and the preset regularization term. Based on the second parameter, generate linear inequality constraints describing the power flow relationship of the distribution network; The physical constraints are treated as independent inequality constraints and, together with the linear inequality constraints, form the model constraints of the convex quadratic programming problem model. Based on the objective function and the model constraints, the convex quadratic programming problem model is constructed.

[0007] Optionally, the second parameter is a set of dynamic constraint parameters for defining approximate constraints on power flow in the distribution network. The set of dynamic constraint parameters includes at least one constraint tightness control parameter. The value of the constraint tightness control parameter is dynamically adjusted by the parameter prediction neural network based on the system uncertainty information or safety margin information contained in the node power injection data. The step of generating linear inequality constraints describing the power flow relationship of the distribution network based on the second parameter includes: Using the dynamic constraint parameter set, a linear inequality constraint is constructed for the boundary value of the convex quadratic programming problem model, which varies with the constraint tightness control parameter.

[0008] Optionally, the physical constraints include time-coupled physical constraints; When constructing the convex quadratic programming problem model, the time-coupled physical constraints are used as inequality constraints or equality constraints, which together with the objective function of the convex quadratic programming problem model constitute a multi-time-period joint optimization problem.

[0009] Optionally, the training method of the parameter prediction neural network includes: Obtain historical node power injection data samples and corresponding scheduling decision labels; Determine the optimization objective, wherein the optimization objective is to minimize the difference between the solution of the convex quadratic programming problem model and the scheduling decision label; The gradient of the optimization objective with respect to the neural network weights is calculated using gradient calculation methods. Using the gradient obtained from the solution, the weight parameters of the parameter prediction neural network are updated using the gradient descent algorithm, and the parameter prediction neural network is trained in an end-to-end manner.

[0010] Optionally, the step of calculating the gradient of the optimization objective with respect to the neural network weights using gradient calculation includes: Based on the optimality conditions of the convex quadratic programming problem model, a joint matrix is ​​generated with model parameters and optimal solutions as variables; The joint matrix is ​​analyzed according to the implicit function theorem, and the Jacobian matrix of the solution to the convex quadratic programming problem model with respect to the model parameters is generated. Based on the Jacobian matrix, the gradient of the loss function of the optimization objective with respect to the solution of the convex quadratic programming problem model is backpropagated to the gradient with respect to the model parameters using the chain rule, so as to calculate the gradient of the loss function with respect to the neural network weights.

[0011] Optionally, obtaining historical node power injection data samples and corresponding scheduling decision labels includes: Obtain sample pairs consisting of historical operation records of the distribution network, wherein each sample pair includes historical node power injection data and a sequence of historical dispatch instructions that were actually executed or verified to be feasible under target historical conditions; The historical node power injection data is used as the historical node power injection data sample, and the historical scheduling instruction sequence is used as the scheduling decision label.

[0012] The present invention also provides a distributed energy dispatching device that does not depend on the physical parameters of the distribution network, the device comprising: The input module is used to acquire node power injection data of the distribution network system and input the node power injection data into a pre-trained parameter prediction neural network. The first generation module is used to generate model parameters for a convex optimization alternative model that describes the dispatch response characteristics of the power distribution network system through the parameter prediction neural network. The second generation module is used to construct a convex quadratic programming problem model based on the model parameters and physical constraints that reflect the physical limits of distributed energy. The computation module is used to call a preset convex optimization solver to solve the convex quadratic programming problem model and obtain the power setpoint sequence corresponding to each distributed energy source. The output module is used to send each of the power setpoint sequences as scheduling instructions to the corresponding distributed energy controller to control the operation of each of the distributed energy sources.

[0013] The present invention also provides an electronic device, the electronic device including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the distributed energy dispatching method that does not depend on the physical parameters of the distribution network as described in any of the preceding claims.

[0014] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the distributed energy dispatching method as described in any of the preceding claims, which is independent of the physical parameters of the distribution network.

[0015] The present invention has at least the following beneficial effects: This technical solution efficiently generates multi-time-period optimal scheduling strategies that satisfy the physical constraints of the distribution network, significantly improving system economy and security. The core innovation lies in its breakthrough elimination of dependence on the physical model of the distribution network: no complex parameters such as network topology and line impedance are required; precise scheduling can be achieved with only a small amount of node power injection data. By directly generating convex optimization alternative model parameters through a parameter prediction neural network, this method greatly simplifies the modeling process, lowers the data requirement threshold, and ensures that the scheduling strategy strictly meets the physical limits of distributed energy resources. The efficient solution mechanism of the convex quadratic programming model and mature solver further guarantees the rapid generation and feasibility of the strategy. This solution effectively solves the bottlenecks of traditional methods, such as complex models, strong data dependence, and poor real-time performance. It not only improves the economy of scheduling but also enhances system security, making it particularly suitable for complex distribution network environments. Attached Figure Description

[0016] The accompanying drawings are provided to further understand the technical solutions of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the technical solutions of the present invention, and do not constitute a limitation on the technical solutions of the present invention.

[0017] Figure 1 This is a flowchart illustrating the steps of a distributed energy dispatching method that does not rely on the physical parameters of the distribution network. Figure 2 This is a flowchart of step S103 in a distributed energy dispatching method that does not depend on the physical parameters of the distribution network. Figure 3 This is a flowchart of the neural network training process in a distributed energy dispatching method that does not depend on the physical parameters of the distribution network. Figure 4 This is a schematic diagram of the structure of a power distribution network testing system provided in an embodiment of this application; Figure 5 This is a schematic diagram showing the comparison of the breach of operational constraints of the energy storage system caused by this technical solution and existing technical solutions; Figure 6 This is a schematic diagram showing the comparison results of the violation of distribution network operation constraints caused by this technical solution and existing technical solutions; Figure 7 This is a schematic diagram showing the comparison of the operating costs of the proposed technical solution and existing technical solutions under conditions of high new energy penetration. Figure 8This is a schematic diagram showing the comparison of the operating costs of the proposed technical solution and existing technical solutions under conditions of low new energy penetration. Figure 9 This is a schematic diagram of a distributed energy dispatching device that does not depend on the physical parameters of the distribution network. Figure 10 This is a schematic diagram of the structure of an electronic device. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] The researchers in this application found that existing end-to-end learning methods (such as direct solution mapping, DSM) attempt to directly fit high-dimensional decision spaces, which are prone to overfitting when there are only a small number of samples (such as 100 days of data), resulting in extremely poor performance on the test set; methods that directly predict decisions cannot guarantee that the output results meet the physical constraints of the equipment (such as battery SOC limits), which may lead to the inability to execute scheduling instructions; and traditional optimization methods cannot operate in power grids with unknown topologies and parameters.

[0020] To address the aforementioned technical problems, this application provides a distributed energy dispatching method and related equipment that do not rely on the physical parameters of the distribution network. This method can efficiently generate multi-time-period optimal dispatching strategies that strictly satisfy the physical constraints of the distribution network, significantly improving the system's economy and security. The following are various embodiments of the technical solution presented in this application.

[0021] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating the steps of a distributed energy dispatching method that does not rely on the physical parameters of the distribution network.

[0022] This embodiment provides a distributed energy dispatching method that does not depend on the physical parameters of the distribution network, including: S101. Obtain the node power injection data of the distribution network system and input the node power injection data into the pre-trained parameter prediction neural network.

[0023] S102. Generate model parameters for a convex optimization alternative model that describes the dispatch response characteristics of the power distribution network system using a parameter prediction neural network.

[0024] S103. Based on the model parameters and the physical constraints that reflect the physical limits of distributed energy, a convex quadratic programming problem model is constructed.

[0025] S104. Call the preset convex optimization solver to solve the convex quadratic programming problem model and obtain the power setpoint sequence corresponding to each distributed energy source.

[0026] S105. Send each power setpoint sequence as a scheduling instruction to the corresponding distributed energy controller to control the operation of each distributed energy source.

[0027] In some embodiments, current grid operating conditions θ are collected, such as load forecasting and photovoltaic power generation forecasting.

[0028] In some embodiments, the parameters of the convex substitution model can be ,in, These parameters are dynamically generated by a neural network based on the current operating condition θ. It is the learned cost vector.

[0029] Understandably, this embodiment efficiently generates multi-period optimal scheduling strategies that satisfy the physical constraints of the distribution network, significantly improving system economy and security. The core innovation lies in its breakthrough elimination of dependence on the physical model of the distribution network: no complex parameters such as network topology and line impedance are required; precise scheduling can be achieved with only a small amount of node power injection data. By directly generating convex optimization alternative model parameters through a parameter prediction neural network, this method greatly simplifies the modeling process, lowers the data requirement threshold, and ensures that the scheduling strategy strictly meets the physical limits of distributed energy resources. The efficient solution mechanism of the convex quadratic programming model and mature solver further guarantees the rapid generation and feasibility of the strategy. This solution effectively solves the bottlenecks of traditional methods, such as complex models, strong data dependencies, and poor real-time performance. It not only improves the economy of scheduling but also enhances system security, making it particularly suitable for complex distribution network environments.

[0030] Please refer to Figure 2 , Figure 2 This is a flowchart of step S103 in a distributed energy dispatching method that does not depend on the physical parameters of the distribution network.

[0031] In some embodiments, step S103 includes: S201. Divide the model parameters into the first parameter and the second parameter.

[0032] S202. Generate the objective function of the convex quadratic programming problem model based on the first parameter and the preset regularization term.

[0033] S203. Generate linear inequality constraints describing the power flow relationship of the distribution network based on the second parameter.

[0034] S204. Physical constraints are treated as independent inequality constraints and combined with linear inequality constraints to form the model constraints of a convex quadratic programming problem model.

[0035] S205. Based on the objective function and model constraints, construct a convex quadratic programming problem model.

[0036] In some embodiments, to determine the objective function, it is constructed as a strongly convex quadratic programming form to guarantee the uniqueness and differentiability of the solution:

[0037] in, It is the learned cost vector. It is a very small regularization coefficient.

[0038] Understandably, in this embodiment, the model parameters are decomposed into a first parameter and a second parameter, which are used to generate the objective function and the linear inequality constraints describing the power flow relationship in the distribution network, respectively. Physical constraints are then integrated as independent inequality constraints. This improvement makes the construction of the convex quadratic programming problem model more accurate, the division between the objective function and constraints clearer, and further enhances the model's adaptability to the physical characteristics of the distribution network and the strictness of the constraints. Consequently, the generated scheduling strategy is more aligned with actual operational needs, achieves higher optimization accuracy, and can more efficiently realize optimal scheduling across multiple time periods, significantly improving system economy and security while ensuring the feasibility and reliability of the scheduling scheme.

[0039] In some embodiments, the second parameter is a set of dynamic constraint parameters for defining approximate constraints on power flow in the distribution network. The set of dynamic constraint parameters includes at least one constraint tightness control parameter. The value of the constraint tightness control parameter is dynamically adjusted by the parameter prediction neural network based on the system uncertainty information or safety margin information contained in the node power injection data. The specific implementation of step S203 is to use a set of dynamic constraint parameters to construct linear inequality constraints for the boundary values ​​of the convex quadratic programming problem model that vary with the constraint tightness control parameters.

[0040] In some embodiments, physical constraints include time-coupled physical constraints; when constructing a convex quadratic programming problem model, time-coupled physical constraints are used as inequality constraints or equality constraints, which together with the objective function of the convex quadratic programming problem model constitute a multi-time-period joint optimization problem.

[0041] It is understandable that constraints can be divided into two parts: unknown constraints and physical constraints.

[0042] Unknown constraints are approximations of unknown power flow constraints using a system of linear inequalities:

[0043] in, These parameters are dynamically generated by a neural network based on the current operating condition θ. This "instance-level local approximation" is more accurate and easier to train than the global approximation.

[0044] Physical constraints refer to explicitly embedded device limitations (such as battery capacity and charge / discharge power limitations):

[0045] This embodiment explicitly preserves known constraints, ensuring that the final output is consistent regardless of the neural network's predictions. It must meet the physical limitations of the equipment, thus ensuring the basic safety of scheduling.

[0046] Please refer to Figure 3 , Figure 3 This is a flowchart of the neural network training process in a distributed energy dispatching method that does not depend on the physical parameters of the distribution network.

[0047] In some embodiments, the training methods for the parameter prediction neural network include: S301. Obtain historical node power injection data samples and corresponding scheduling decision labels.

[0048] S302. Determine the optimization objective, whereby the optimization objective is to minimize the difference between the solution of the convex quadratic programming problem model and the scheduling decision label.

[0049] S303. Using gradient calculation, solve for the gradient of the optimization target with respect to the neural network weights.

[0050] S304. Using the gradient obtained from the solution, update the weight parameters of the parameter prediction neural network using the gradient descent algorithm, and train the parameter prediction neural network in an end-to-end manner.

[0051] In this embodiment, the training and inference processes do not require physical parameters such as the power grid's admittance matrix and topology, thus achieving partially model-free scheduling.

[0052] In some embodiments, the neural network weights are updated end-to-end using stochastic gradient descent (SGD) or the Adam optimizer.

[0053] In some embodiments, the optimization objective is to minimize scheduling decisions. With optimal label Mean squared error (MSE) between:

[0054] Understandably, in this embodiment, by acquiring historical operational data and scheduling decision labels, and aiming to minimize the difference between the solution and the label of the convex quadratic programming problem model, a gradient descent algorithm is used to train the neural network end-to-end. This training method enables the neural network to learn parameter generation rules that better reflect actual operational patterns, thereby more accurately predicting model parameters. This not only improves the optimization accuracy of the scheduling strategy but also enhances the model's generalization ability to different operational scenarios, further improving the system's economy and security in multi-time-period scheduling, and ensuring the efficiency and reliability of scheduling instructions.

[0055] In some embodiments, step S301 includes: Obtain sample pairs consisting of historical operation records of the distribution network. Each sample pair includes historical node power injection data and a sequence of historical dispatch instructions that were actually executed or verified to be feasible under the target historical conditions. The historical node power injection data is used as the historical node power injection data sample, and the historical dispatch instruction sequence is used as the dispatch decision label.

[0056] In some embodiments, the sample pair is ,in, It can be a historically suboptimal but feasible scheduling record.

[0057] Understandably, in this embodiment, the method for acquiring training data for the parameter prediction neural network is further refined, specifying that sample pairs consist of historical node power injection data of the distribution network and corresponding historical dispatch instruction sequences. This improvement makes the training data more targeted and practical, with the historical dispatch instruction sequences serving as labels for dispatch decisions, providing the neural network with more accurate learning targets. The neural network trained in this way can more accurately capture the intrinsic relationship between distribution network operation and dispatch instructions, thereby generating model parameters that better meet actual operational needs. This further improves the accuracy and reliability of the dispatch strategy, enhances the system's adaptability under complex operating conditions, significantly improves the economy and security of the distribution network, and optimizes dispatch performance.

[0058] In some embodiments, step S303 includes: Based on the optimality condition of the convex quadratic programming problem model, a joint matrix with model parameters and optimal solution as variables is generated. The joint matrix is ​​analyzed according to the implicit function theorem, and the Jacobian matrix of the solution of the convex quadratic programming problem model with respect to the model parameters is generated. Based on the Jacobian matrix, the gradient of the loss function of the optimization objective with respect to the solution of the convex quadratic programming problem model is backpropagated to the gradient with respect to the model parameters using the chain rule, so as to calculate the gradient of the loss function with respect to the neural network weights.

[0059] In this embodiment, differentiation is performed using the KKT conditions and the implicit function theorem.

[0060] The KKT conditions of the hybrid substitution model can be written as matrix equations:

[0061] The Jacobian matrix of the optimization solution with respect to the parameters is obtained using the implicit function theorem:

[0062] Calculate the loss function for neural network weights using the chain rule. gradient:

[0063] Understandably, in this embodiment, the newly added step S303, through the calculation of the joint matrix and Jacobian matrix, combined with the implicit function theorem and the chain rule, achieves precise gradient backpropagation from the optimized objective loss function to the neural network weights. This improvement makes the training of the neural network more accurate and enables more efficient adjustment of weight parameters to minimize the difference between the model solution and the scheduling decision label. This not only improves the accuracy of parameter prediction but also further enhances the model's adaptability to complex scheduling problems and its solution accuracy. Ultimately, this technique significantly improves the optimization effect of distributed energy scheduling strategies, ensuring that the scheduling scheme achieves higher economy and security while meeting physical constraints, thus optimizing the overall system performance.

[0064] Compared to the Direct Mapping Method (DSM), which requires thousands of samples, this technical solution introduces the structure of the optimization problem as an inductive bias, requiring only a very small number of samples (e.g., 100 samples) to train a model with excellent generalization performance, thus solving the problem of scarce distribution network scheduling data.

[0065] By explicitly preserving known constraints in the hybrid model, the scheduling instructions generated by this technical solution strictly meet the physical limitations of the device under any circumstances (such as the battery not being overcharged or over-discharged), which is significantly better than pure deep learning methods that often generate non-compliant instructions.

[0066] Please see Figure 4 , Figure 4 This is a schematic diagram of a distribution network test system provided in an embodiment of this application. The aforementioned distributed energy dispatch method, which is independent of the physical parameters of the distribution network, is applied to this test system, which includes a 33-node distribution network. DG represents distributed renewable energy generating units, and the battery symbol represents distributed energy storage facilities.

[0067] Please see Figures 5 to 8 , Figure 5This demonstrates the results of comparing the default amounts of energy storage system operation constraints caused by the distributed energy dispatch method of this application, which does not rely on the physical parameters of the distribution network, with those caused by the DSM method (Direct Solution Mapping, DSM). Figure 6 The results show the comparison of the default amounts of distribution network operation constraints caused by the distributed energy dispatch method of this application, which does not depend on the physical parameters of the distribution network, and the DSM method; Figure 7 The presentation shows the comparison of the operating costs of the distributed energy dispatching method, DSM method, and physical model-based method obtained under conditions of high renewable energy penetration rate, which do not rely on the physical parameters of the distribution network. Figure 8 This paper presents a comparison of the operating costs of the distributed energy dispatching method, the DSM method, and the physical model-based method obtained under conditions of high renewable energy penetration.

[0068] Please refer to Figure 9 , Figure 9 This is a schematic diagram of a distributed energy dispatching device that does not depend on the physical parameters of the distribution network.

[0069] This embodiment also provides a distributed energy dispatching device that does not depend on the physical parameters of the distribution network, including: Input module 401 is used to acquire node power injection data of the distribution network system and input the node power injection data into a pre-trained parameter prediction neural network; The first generation module 402 is used to generate model parameters for a convex optimization alternative model that describes the dispatch response characteristics of the power distribution network system through a parameter prediction neural network. The second generation module 403 is used to construct a convex quadratic programming problem model based on model parameters and physical constraints that reflect the physical limits of distributed energy. The computation module 404 is used to call a preset convex optimization solver to solve the convex quadratic programming problem model and obtain the power setpoint sequence corresponding to each distributed energy source. The output module 405 is used to send each power setpoint sequence as a scheduling instruction to the corresponding distributed energy controller to control the operation of each distributed energy source.

[0070] It will be understood by those skilled in the art that all or some of the steps and apparatuses in the methods disclosed above can be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components can be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. As is known to those skilled in the art, communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0071] It is understood that the content of the above method embodiments is applicable to the present device embodiments. The specific functions implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0072] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement any of the above-mentioned distributed energy dispatching methods that do not depend on the physical parameters of the distribution network.

[0073] refer to Figure 10 , Figure 10 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes: The processor 501 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 502 can be implemented as a read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 502 can store operating devices and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 502 and is called and executed by the processor 501 to execute the distributed energy dispatching method independent of the physical parameters of the distribution network according to the embodiments of this application. The input / output interface 503 is used to implement information input and output; The communication interface 504 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 505 transmits information between various components of the device (e.g., processor 501, memory 502, input / output interface 503, and communication interface 504); The processor 501, memory 502, input / output interface 503, and communication interface 504 are connected to each other within the device via bus 505.

[0074] It is understood that the content of the above method embodiments is applicable to the embodiments of this electronic device. The specific functions implemented by the embodiments of this electronic device are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0075] This application also provides a computer-readable storage medium storing a processor-executable program, which, when executed by a processor, is used to implement the distributed energy dispatching method independent of the physical parameters of the distribution network as described in any of the above specific embodiments.

[0076] This application also discloses a computer program product, including a computer program or computer instructions, which are stored in a computer-readable storage medium. The processor of the computer device reads the computer program or computer instructions from the computer-readable storage medium and executes the computer program or computer instructions, causing the computer device to perform the distributed energy dispatching method that does not depend on the physical parameters of the distribution network as described in any of the preceding embodiments.

[0077] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0078] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices. It should be understood that in this application, “at least one” means one or more, and “more than one” means two or more.

[0079] In the several embodiments provided in this application, it should be understood that the disclosed apparatus, devices, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0080] The units described 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 can be selected to achieve the purpose of this embodiment according to actual needs.

[0081] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0082] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0083] Although the description of this application has been quite detailed and particularly focused on several of the described embodiments, it is not intended to limit itself to any of these details or embodiments or any particular embodiment. Rather, it should be considered as effectively covering the intended scope of this application by referring to the appended claims and taking into account the prior art, which provides for a broad possible interpretation of these claims. Furthermore, the foregoing description of this application with respect to embodiments foreseeable by the inventors is intended to provide a useful description, and non-substantial modifications to this application that have not yet been foreseen may still represent equivalent modifications.

Claims

1. A distributed energy dispatching method independent of distribution network physical parameters, characterized in that, The method includes: Obtain node power injection data of the distribution network system and input the node power injection data into a pre-trained parameter prediction neural network; The parameter prediction neural network generates model parameters for a convex optimization alternative model that describes the dispatch response characteristics of the power distribution network system. Based on the model parameters and the physical constraints reflecting the physical limits of distributed energy, a convex quadratic programming problem model is constructed. The convex quadratic programming problem model is solved by calling a preset convex optimization solver to obtain the power setpoint sequence for each distributed energy source. Each of the power setpoint sequences is sent as a scheduling instruction to the corresponding distributed energy controller to control the operation of each of the distributed energy sources.

2. The method according to claim 1, characterized in that, The convex quadratic programming problem model is constructed based on the model parameters and the physical constraints reflecting the physical limits of distributed energy, including: The model parameters are divided into first parameters and second parameters; The objective function of the convex quadratic programming problem model is generated based on the first parameter and the preset regularization term. Based on the second parameter, generate linear inequality constraints describing the power flow relationship of the distribution network; The physical constraints are treated as independent inequality constraints and, together with the linear inequality constraints, form the model constraints of the convex quadratic programming problem model. Based on the objective function and the model constraints, the convex quadratic programming problem model is constructed.

3. The method according to claim 2, characterized in that, The second parameter is a set of dynamic constraint parameters used to define approximate constraints on power flow in the distribution network. The set of dynamic constraint parameters includes at least one constraint tightness control parameter. The value of the constraint tightness control parameter is dynamically adjusted by the parameter prediction neural network based on the system uncertainty information or safety margin information contained in the node power injection data. The step of generating linear inequality constraints describing the power flow relationship of the distribution network based on the second parameter includes: Using the dynamic constraint parameter set, a linear inequality constraint is constructed for the boundary value of the convex quadratic programming problem model, which varies with the constraint tightness control parameter.

4. The method according to claim 2, characterized in that, The physical constraints include time-coupled physical constraints; When constructing the convex quadratic programming problem model, the time-coupled physical constraints are used as inequality constraints or equality constraints, which together with the objective function of the convex quadratic programming problem model constitute a multi-time-period joint optimization problem.

5. The method according to claim 1, characterized in that, The training methods for the parameter prediction neural network include: Obtain historical node power injection data samples and corresponding scheduling decision labels; Determine the optimization objective, wherein the optimization objective is to minimize the difference between the solution of the convex quadratic programming problem model and the scheduling decision label; The gradient of the optimization objective with respect to the neural network weights is calculated using gradient calculation methods. Using the gradient obtained from the solution, the weight parameters of the parameter prediction neural network are updated using the gradient descent algorithm, and the parameter prediction neural network is trained in an end-to-end manner.

6. The method according to claim 5, characterized in that, The step of calculating the gradient of the optimization objective with respect to the neural network weights using gradient calculation includes: Based on the optimality conditions of the convex quadratic programming problem model, a joint matrix is ​​generated with model parameters and optimal solutions as variables; The joint matrix is ​​analyzed according to the implicit function theorem, and the Jacobian matrix of the solution to the convex quadratic programming problem model with respect to the model parameters is generated. Based on the Jacobian matrix, the gradient of the loss function of the optimization objective with respect to the solution of the convex quadratic programming problem model is backpropagated to the gradient with respect to the model parameters using the chain rule, so as to calculate the gradient of the loss function with respect to the neural network weights.

7. The method according to claim 5, characterized in that, The acquisition of historical node power injection data samples and corresponding scheduling decision tags includes: Obtain sample pairs consisting of historical operation records of the distribution network, wherein each sample pair includes historical node power injection data and a sequence of historical dispatch instructions that were actually executed or verified to be feasible under target historical conditions; The historical node power injection data is used as the historical node power injection data sample, and the historical scheduling instruction sequence is used as the scheduling decision label.

8. A distributed energy dispatching device that does not rely on the physical parameters of the distribution network, characterized in that, The device includes: The input module is used to acquire node power injection data of the distribution network system and input the node power injection data into a pre-trained parameter prediction neural network. The first generation module is used to generate model parameters for a convex optimization alternative model that describes the dispatch response characteristics of the power distribution network system through the parameter prediction neural network. The second generation module is used to construct a convex quadratic programming problem model based on the model parameters and physical constraints that reflect the physical limits of distributed energy. The computation module is used to call a preset convex optimization solver to solve the convex quadratic programming problem model and obtain the power setpoint sequence corresponding to each distributed energy source. The output module is used to send each of the power setpoint sequences as scheduling instructions to the corresponding distributed energy controller to control the operation of each of the distributed energy sources.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the distributed energy dispatching method that does not depend on the physical parameters of the distribution network as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the distributed energy dispatching method that does not depend on the physical parameters of the distribution network as described in any one of claims 1 to 7.

Images