Source-network-storage integrated power distribution network operation optimization method

Abstract

The invention provides a source-network-storage integrated power distribution network operation optimization method, which relates to the technical field of power distribution networks and comprises the steps of setting an intelligent sensor, collecting multi-dimensional power distribution operation parameters, performing semantic extraction and automatic labeling, determining logic association after conflict correction, obtaining a power distribution knowledge graph, constructing a dynamic prediction model and performing power distribution index prediction. Obtaining a prediction result and generating a key index prediction map; constructing a multi-temporal-spatial-scale collaborative optimization model, setting an objective function, constructing a physical optimization sub-model and a market game optimization sub-model, carrying out interactive coupling, and carrying out collaboration, interactive feedback and iterative convergence to obtain a multi-scale collaborative optimization scheme; and carrying out distributed evaluation, generating an evaluation result, if the evaluation result does not reach the standard, packaging the multi-temporal-spatial-scale collaborative optimization model into an optimization contract, triggering the optimization contract, generating a candidate optimization scheme, carrying out repeated evaluation until the standard is reached, and taking the candidate optimization scheme as a consensus scheme and executing the candidate optimization scheme.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network, and in particular to a method for optimizing the operation of a distribution network integrating source, network and storage. Background Art

[0002] With the large-scale access of new energy and the deepening of power market reform, distribution networks are facing multiple challenges such as supply and demand balance, power quality maintenance, and economic operation. Traditional distribution network operation optimization methods are mainly focused on a single time scale and a single scenario, which is difficult to adapt to the complexity and variability of distribution networks in the context of source-grid-storage integration.

[0003] At present, there have been some studies and explorations on the operation optimization of the source-grid-storage integrated distribution network. However, the existing research mainly focuses on the optimization at a single time scale, such as day-ahead optimization or real-time optimization, and lacks consideration of multi-time scale collaborative optimization. In addition, most of the existing methods adopt centralized optimization, which has limited computational efficiency and scalability, and it is difficult to meet the optimization needs of large-scale distribution networks. At the same time, the existing methods lack flexibility and adaptability in the modeling and solution process of the distribution network, and it is difficult to quickly respond to changes in the operating status of the distribution network.

[0004] Therefore, a solution is urgently needed to solve the problems existing in the prior art. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide a distribution network operation optimization method with integrated source, grid and storage, which realizes the efficient optimization and flexible control of the distribution network with integrated source, grid and storage, improves the economy, reliability and adaptability of the distribution network operation, and provides strong support for the development of energy Internet and smart energy.

[0006] To achieve the above-mentioned purpose, the present invention adopts the following technical scheme: a distribution network operation optimization method integrating source, network and storage, wherein various types of intelligent sensors are respectively arranged at substations, distribution lines and distributed power sources in the distribution network, and multi-dimensional distribution operation parameters are collected by the intelligent sensors, and uploaded to the edge computing server through the multi-protocol fusion distribution Internet of Things. The edge computing server performs semantic extraction and automatic annotation by combining energy field knowledge and distribution network operation law through a knowledge graph construction algorithm based on ontology reasoning, and determines the corresponding logical association after conflict correction of the annotated semantic content to obtain a distribution knowledge graph, and constructs a dynamic prediction model based on the distribution knowledge graph, and performs distribution index prediction in combination with a graph convolutional neural network to obtain prediction results and generate a key index prediction graph;

[0007] Based on the power distribution knowledge graph and the key indicator prediction graph, a multi-time and space scale collaborative optimization model is constructed and an objective function is set through the integrated time and space data modeling technology. For different time and space scales, a multi-time scale rolling optimization algorithm and a spatial hierarchical optimization algorithm are used to construct a plurality of physical optimization sub-models of different scales. A market game optimization sub-model is constructed through an agent-based distributed game optimization method and interactively coupled with the physical optimization sub-model. The physical optimization sub-model is coordinated through multi-agent reinforcement learning. The market game optimization sub-model and the coordinated physical optimization sub-model are interactively fed back and iteratively converged in combination with an adaptive evolutionary game algorithm to obtain a multi-scale collaborative optimization solution.

[0008] The evaluation rules and evaluation indicator system corresponding to the multi-scale collaborative optimization scheme are encapsulated into an evaluation contract, the multi-scale collaborative optimization scheme is executed, and the multi-scale collaborative optimization scheme is distributedly evaluated based on the real-time operation data and the evaluation contract through the blockchain nodes in the pre-set blockchain network to generate an evaluation result. If the evaluation result does not meet the standard, the multi-spatiotemporal scale collaborative optimization model is encapsulated into an optimization contract through the blockchain smart contract technology, the optimization contract is collaboratively triggered by the blockchain nodes, and the candidate optimization scheme is re-solved based on the real-time operation data to generate the candidate optimization scheme, and the evaluation is repeated until the generated candidate optimization scheme meets the standard, and the candidate optimization scheme of the last iteration is used as the consensus scheme and executed.

[0009] In a preferred embodiment, the intelligent sensor includes a phasor measurement unit, a micro-meteorological station, a distribution line monitoring device, a distribution fault indicator and a power quality monitoring device. Based on the intelligent sensor, multiple types of heterogeneous data are collected and combined into a data set to obtain the multi-dimensional distribution operation parameters;

[0010] Based on the distribution Internet of Things with multi-protocol fusion, the communication protocol is selected on the sensor node side according to the data type and communication requirements. On the aggregation node side, the communication protocol is fused and converted through the edge computing gateway, and the fused and converted data is uploaded to the edge computing server corresponding to the distribution Internet of Things in combination with the high-speed communication link.

[0011] The edge computing server constructs an energy domain ontology according to the concepts, terms and relationships in the energy domain, defines the equipment, topology and environment in the distribution network as core concepts, takes attribute association, spatial adjacency and logical dependency as semantic relationships, formally represents the energy domain ontology through an ontology description language, generates an energy domain knowledge base, calculates the support between the load, voltage and network loss in the distribution network in the historical operation data of the distribution network through an association rule mining algorithm, determines the association pattern based on the support, determines the operation mode corresponding to different time scales in combination with a time series clustering algorithm, and determines the causal relationship between equipment failure and power quality non-compliance events in the distribution network in combination with a causal inference algorithm, and obtains the operation law of the distribution network;

[0012] The edge computing server semantically represents the received fused and converted data according to time, space and attributes and maps it to the concepts and relationships in the energy field ontology. Through the ontology reasoning algorithm, semantic enrichment is performed based on the energy field ontology and the operating rules of the distribution network, and the business semantics, spatiotemporal semantics and associated semantics corresponding to the current data are automatically labeled. The labeled semantic content is conflict-detected through a rule-based data quality verification method to obtain the labeled semantic content, and the labeled semantic content is converted into triples and the nodes and edges of the distribution knowledge graph are constructed, wherein the nodes represent the equipment, users and environment in the distribution network, and the edges represent the association relationship between entities, so as to obtain the distribution knowledge graph;

[0013] Based on the power distribution knowledge graph and the power grid operation and management requirements, load demand, network loss level and voltage qualification rate are used as key indicators. Based on the key indicators, the power distribution knowledge graph is feature learned in combination with a graph convolutional neural network, and node information and structure information are aggregated to obtain a low-dimensional vector representation of each node. An initial prediction model is established in combination with a long short-term memory network. The historical operation data of the distribution network is added to the initial prediction model as a training set, and prediction results are generated and compared with the true labels corresponding to each data. The loss value is calculated and the hyperparameter optimization of the initial prediction model is performed based on the loss value. The optimization is repeated until the model converges to obtain the dynamic prediction model. The key indicators of the distribution network at different time scales are predicted by the dynamic prediction model to obtain the key indicator prediction graph.

[0014] In a preferred embodiment, the support between load, voltage and network loss in the distribution network is determined by calculating the historical operation data of the distribution network through an association rule mining algorithm as shown in the following formula:

[0015]

[0016] Among them, SUP represents support, α represents the time decay factor, t0 represents the current time, t irepresents the timestamp of each historical data, β represents the spatial impact factor, d(X, Y, Z) represents the spatial distance function, which is used to measure the spatial distance between load, voltage and network loss, X represents load, Y represents voltage, Z represents network loss, I() represents the indicative function, T i represents the i-th historical data, and N represents the total number of historical data.

[0017] In a preferred embodiment, based on the power distribution knowledge graph and the key indicator prediction graph, a multi-spatiotemporal scale collaborative optimization model is constructed by spatiotemporal data integration modeling technology and an objective function is set, wherein the multi-spatiotemporal scale collaborative optimization model is composed of multiple sub-models;

[0018] For different spatiotemporal scales, the time window and the optimized time domain are determined by a multi-time-scale rolling optimization algorithm. In each window, based on the multi-time-space scale data and the power distribution knowledge graph, the optimized time domain and the prediction information are updated according to the rolling mechanism to generate a multi-time-scale physical optimization sub-model. Based on the spatial characteristics of the distribution network, the spatial topological structure of the distribution network is determined and in each area, a spatial hierarchical physical optimization sub-model is constructed according to physical constraints and operating rules, wherein the spatial hierarchical physical optimization sub-model and the multi-time-scale physical optimization sub-model together constitute the physical optimization sub-model;

[0019] Based on the distributed game optimization algorithm, the market entities in the distribution network are regarded as players, and the utility function and strategy space corresponding to each player are determined. Each player optimizes its own decision-making behavior based on the distributed game, and the optimal response dynamic algorithm is used to make the players repeat the game and reach the game equilibrium, and the game hyperparameters are obtained. The market game optimization sub-model is constructed according to the game hyperparameters;

[0020] The decision results of the market players in the market game optimization submodel are transmitted to the physical optimization submodel and used as boundary conditions and input parameters of physical optimization. The physical optimization submodel optimizes the operation plan of the distribution network to obtain the distribution network state variables. The decision space of the market players is updated based on the distribution network state variables and the game optimization is repeated. The pre-set coupling variables and coupling functions are combined to interactively couple with the market game optimization submodel to generate an interactive coupling optimization model and solve it to obtain a decoupled optimal solution, wherein the decoupled optimal solution includes the optimal parameters of the market game optimization submodel and the physical optimization submodel;

[0021] The physical optimization sub-model is updated based on the optimal parameters, each physical optimization sub-model is used as an agent in combination with multi-agent reinforcement learning, a multi-agent system is constructed, and the optimization decision strategy is learned according to the reinforcement learning algorithm, a reward function is constructed with the goal of maximizing the cumulative reward, a state space, an action space and a state transfer function are defined, and the strategy function corresponding to each agent is dynamically adjusted according to real-time data until the cumulative reward reaches the maximum and the collaboration is completed;

[0022] For the physical optimization sub-model and the market game optimization sub-model after collaborative processing, an initial game population is generated by an adaptive evolutionary game algorithm, wherein each individual in the initial game population represents a decision, and the fitness of each individual is calculated according to the strategy distribution of the initial game population and the game environment, and an evolutionary operation is selected based on the fitness and a second-generation game population is generated, and the second-generation game population dynamically adjusts its own strategy according to its own fitness and the strategy distribution of individuals in the second-generation game population, and the evolutionary operation and the individuals in the second-generation game population are returned to the physical optimization sub-model and the market game optimization sub-model and the decision space is updated, and iterations are repeated until the strategy distribution converges, and the individual with the highest fitness value in the last iteration is output as a multi-scale collaborative optimization solution.

[0023] In a preferred embodiment, the utility function corresponding to each player is determined as shown in the following formula:

[0024] U i (s i ,s -i )=w r ·R i (s i ,s -i )-w c ·C i (s i )+w f ·F i (s i ,s -i );

[0025] Among them, U i (s i ,s -i ) represents the utility function, w r Represents the return weight coefficient, R i (s i ,s -i ) represents the payoff function of the ith player, w c represents the cost weight coefficient, C i (s i ) represents the cost function of the ith player, w f represents the fairness weight coefficient, Fi (s i ,s -i ) is the fairness function of the ith player, which indicates the fairness of the income distribution between different players in the game process, s i represents the strategy of the ith player, s -i Represents the strategy combination of all players except the i-th player.

[0026] In a preferred embodiment, the objective function of the multi-temporal and spatial scale collaborative optimization model is as follows:

[0027]

[0028] Where j represents the index of the spatial dimension, h represents the index of the temporal dimension, M represents the total number of spatial dimensions, H represents the total number of temporal dimensions, and W represents the total number of spatial dimensions. c Represents the cost weight coefficient of the optimization process, C jh (x jh ) represents the cost function of the jth region at time h, x jh represents the decision variable, W g represents the constraint violation weight coefficient, λ jh represents the constraint violation penalty coefficient, g jh (x jh ) represents the constraint function of the jth region at time h, W r represents the risk weight coefficient, R jh (x jh ) represents the operating risk function of the jth region at time h.

[0029] In a preferred embodiment, an evaluation index system and evaluation rules corresponding to the pre-set multi-scale collaborative optimization scheme are obtained, wherein the evaluation index system includes technical indicators, economic indicators and environmental indicators, and the evaluation rules include evaluation process, data source and evaluation cycle, and the evaluation rules and the evaluation index system are encapsulated as an evaluation contract;

[0030] Execute the multi-scale collaborative optimization scheme to generate optimization results, upload the optimization results and the parameters of the multi-scale collaborative optimization scheme to the blockchain network, collect the real-time operation data of the distribution network through the blockchain node and automatically trigger the execution of the evaluation contract according to the preset evaluation cycle and evaluation conditions, calculate the real-time operation data according to the evaluation function in the evaluation contract, and generate the evaluation results in combination with the evaluation indicator system and the evaluation rules;

[0031] Whether the evaluation result meets the standard is determined according to the task requirements and pre-set constraints. If the evaluation result does not meet the standard, the objective function, constraints and decision variables of the multi-spatiotemporal scale collaborative optimization model are converted into functions and variables in the optimization contract according to the blockchain smart contract technology to generate an optimization contract. The blockchain nodes collaboratively trigger the optimization contract to obtain the real-time operation data and the evaluation result as input data, solve the multi-spatiotemporal scale collaborative optimization model, obtain the candidate optimization plan, upload the candidate optimization plan to the blockchain network and evaluate it again, and repeat the evaluation until the candidate optimization plan meets the standard;

[0032] The candidate optimization solution generated by the last iteration is uploaded to the blockchain network as the consensus solution. The blockchain node verifies the constraints of the consensus solution. If the consensus solution meets the constraints of all blockchain nodes, the consensus solution is executed. Otherwise, the consensus solution is deleted and the candidate optimization solution is regenerated.

[0033] The present invention also provides a source-grid-storage integrated distribution network operation optimization system, which is used to implement the source-grid-storage integrated distribution network operation optimization method, including:

[0034] The first unit is used to respectively set up various types of smart sensors at the substations, distribution lines and distributed power sources in the distribution network, collect multi-dimensional distribution operation parameters through the smart sensors, and upload them to the edge computing server through the multi-protocol fusion distribution Internet of Things. The edge computing server uses the knowledge graph construction algorithm based on ontology reasoning to combine energy field knowledge and distribution network operation rules for semantic extraction and automatic annotation, and performs conflict correction on the annotated semantic content and determines the corresponding logical association to obtain the distribution knowledge graph. A dynamic prediction model is constructed based on the distribution knowledge graph, and the distribution index is predicted in combination with the graph convolutional neural network to obtain the prediction results and generate the key index prediction graph;

[0035] The second unit is used to construct a multi-space-time scale collaborative optimization model and set an objective function based on the power distribution knowledge graph and the key indicator prediction graph through the space-time data integration modeling technology, and for different space-time scales, construct multiple physical optimization sub-models of different scales through a multi-time scale rolling optimization algorithm and a spatial hierarchical optimization algorithm, construct a market game optimization sub-model through an agent-based distributed game optimization method and interact with the physical optimization sub-model, collaborate with the physical optimization sub-model through multi-agent reinforcement learning, and interactively feedback and iteratively converge the market game optimization sub-model and the physical optimization sub-model after collaborative processing in combination with an adaptive evolutionary game algorithm to obtain a multi-scale collaborative optimization solution;

[0036] The third unit is used to encapsulate the evaluation rules and evaluation indicator system corresponding to the multi-scale collaborative optimization scheme into an evaluation contract, execute the multi-scale collaborative optimization scheme, and perform distributed evaluation on the multi-scale collaborative optimization scheme based on real-time operation data and the evaluation contract through the blockchain nodes in the pre-set blockchain network to generate an evaluation result. If the evaluation result does not meet the standard, the multi-spatiotemporal scale collaborative optimization model is encapsulated as an optimization contract through the blockchain smart contract technology, and the optimization contract is collaboratively triggered by the blockchain nodes and a candidate optimization scheme is re-solved based on the real-time operation data. The evaluation is repeated until the generated candidate optimization scheme meets the standard, and the candidate optimization scheme of the last iteration is used as the consensus scheme and executed.

[0037] The present invention also provides an electronic device, characterized in that it includes:

[0038] processor;

[0039] a memory for storing processor-executable instructions;

[0040] Among them, the processor is configured to call the instructions stored in the memory to execute the source-grid-storage integrated distribution network operation optimization method.

[0041] The present invention also provides a computer-readable storage medium on which computer program instructions are stored, characterized in that when the computer program instructions are executed by a processor, the source-grid-storage integrated distribution network operation optimization method is implemented.

[0042] Compared with the prior art, the present invention has the following beneficial effects: by setting up intelligent sensors at different locations of the distribution network, collecting multi-dimensional distribution operation parameters, and uploading them to the edge computing server through the distribution Internet of Things with multi-protocol fusion, the knowledge graph based on ontology reasoning is used to build an algorithm, and semantic extraction and automatic labeling are performed in combination with energy field knowledge and distribution network operation rules to obtain a distribution knowledge graph, thereby realizing efficient collection and semantic processing of distribution network data, providing rich and accurate data support for subsequent optimization, making full use of the semantic association information contained in the distribution knowledge graph, and combining deep learning technology to accurately capture key indicators of the distribution network. The dynamic change trend of the target provides forward-looking guidance for optimization decision-making, fully considers the characteristics of the distribution network in time and space dimensions, realizes the joint optimization of the physical level and the market level, and improves the globality and effectiveness of the optimization. The automatic evaluation and iterative optimization mechanism utilizes the decentralization and non-tamperability of the blockchain to ensure the fairness, transparency and efficiency of the evaluation process, while realizing continuous improvement and adaptive adjustment of the optimization scheme. In summary, the present invention realizes the efficient optimization and flexible control of the source-grid-storage integrated distribution network, improves the economy, reliability and adaptability of the distribution network operation, and provides strong support for the development of energy Internet and smart energy. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A schematic diagram of a method flow chart of a preferred embodiment of the present invention;

[0044] Figure 2 It is a schematic diagram of the structure of a system according to a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0045] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0046] It should be noted that the following detailed descriptions are illustrative and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0047] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application; as used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or their combinations.

[0048] Figure 1 FIG. 1 is a flow chart of a method for optimizing the operation of a distribution network with integrated source, grid and storage according to an embodiment of the present invention. Figure 1As shown, the method includes:

[0049] S1. Various types of smart sensors are respectively set at the substations, distribution lines and distributed power sources in the distribution network. Multi-dimensional distribution operation parameters are collected by the smart sensors and uploaded to the edge computing server through the multi-protocol fusion distribution Internet of Things. The edge computing server uses the knowledge graph construction algorithm based on ontology reasoning to combine energy field knowledge and distribution network operation rules for semantic extraction and automatic annotation, and performs conflict correction on the annotated semantic content and determines the corresponding logical association to obtain the distribution knowledge graph. A dynamic prediction model is constructed based on the distribution knowledge graph, and the distribution index is predicted in combination with the graph convolutional neural network to obtain the prediction results and generate the key index prediction graph;

[0050] The multi-protocol integrated distribution Internet of Things aims to enable devices using different communication protocols to communicate seamlessly in the same network, thereby realizing unified management and analysis of data. The edge computing server is a computing device located at the edge of the network, close to the data source. The knowledge graph construction algorithm based on ontology reasoning uses the ontology model to infer and construct complex relationships between entities. By associating data with ontology, it can automatically generate or expand the knowledge graph and improve the semantic understanding ability of information. The semantic extraction is the process of extracting meaningful information from text, images or other data forms. The conflict correction refers to identifying and resolving inconsistencies or contradictions between different sources or systems in the data or decision-making process. The distribution indicator prediction involves the use of technologies such as data analysis and machine learning to predict the performance indicators of the distribution system.

[0051] In an optional embodiment,

[0052] Various types of smart sensors are respectively set at substations, distribution lines and distributed power sources in the distribution network. Multi-dimensional distribution operation parameters are collected by the smart sensors and uploaded to the edge computing server through the multi-protocol fusion distribution Internet of Things. The edge computing server uses the knowledge graph construction algorithm based on ontology reasoning to combine energy field knowledge and distribution network operation rules for semantic extraction and automatic annotation, and performs conflict correction on the annotated semantic content and determines the corresponding logical association to obtain the distribution knowledge graph. A dynamic prediction model is constructed based on the distribution knowledge graph, and the distribution index is predicted in combination with the graph convolutional neural network to obtain the prediction results and generate the key index prediction graph including:

[0053] Various types of intelligent sensors are respectively arranged at the substations, distribution lines and distributed power sources in the distribution network, wherein the intelligent sensors include phasor measurement units, micro-meteorological stations, distribution line monitoring devices, distribution fault indicators and power quality monitoring devices. Based on the intelligent sensors, various types of heterogeneous data are collected and combined into data sets to obtain the multi-dimensional distribution operation parameters;

[0054] Based on the distribution Internet of Things with multi-protocol fusion, the communication protocol is selected on the sensor node side according to the data type and communication requirements. On the aggregation node side, the communication protocol is fused and converted through the edge computing gateway, and the fused and converted data is uploaded to the edge computing server corresponding to the distribution Internet of Things in combination with the high-speed communication link.

[0055] The edge computing server constructs an energy domain ontology according to the concepts, terms and relationships in the energy domain, defines the equipment, topology and environment in the distribution network as core concepts, takes attribute association, spatial adjacency and logical dependency as semantic relationships, formally represents the energy domain ontology through an ontology description language, generates an energy domain knowledge base, calculates the support between the load, voltage and network loss in the distribution network in the historical operation data of the distribution network through an association rule mining algorithm, determines the association pattern based on the support, determines the operation mode corresponding to different time scales in combination with a time series clustering algorithm, and determines the causal relationship between equipment failure and power quality non-compliance events in the distribution network in combination with a causal inference algorithm, and obtains the operation law of the distribution network;

[0056] The edge computing server semantically represents the received fused and converted data according to time, space and attributes and maps it to the concepts and relationships in the energy field ontology. Through the ontology reasoning algorithm, semantic enrichment is performed based on the energy field ontology and the operating rules of the distribution network, and the business semantics, spatiotemporal semantics and associated semantics corresponding to the current data are automatically labeled. The labeled semantic content is conflict-detected through a rule-based data quality verification method to obtain the labeled semantic content, and the labeled semantic content is converted into triples and the nodes and edges of the distribution knowledge graph are constructed, wherein the nodes represent the equipment, users and environment in the distribution network, and the edges represent the association relationship between entities, so as to obtain the distribution knowledge graph;

[0057] Based on the power distribution knowledge graph and the power grid operation and management requirements, load demand, network loss level and voltage qualification rate are used as key indicators. Based on the key indicators, the power distribution knowledge graph is feature learned in combination with a graph convolutional neural network, and node information and structure information are aggregated to obtain a low-dimensional vector representation of each node. An initial prediction model is established in combination with a long short-term memory network. The historical operation data of the distribution network is added to the initial prediction model as a training set, and prediction results are generated and compared with the true labels corresponding to each data. The loss value is calculated and the hyperparameter optimization of the initial prediction model is performed based on the loss value. The optimization is repeated until the model converges to obtain the dynamic prediction model. The key indicators of the distribution network at different time scales are predicted by the dynamic prediction model to obtain the key indicator prediction graph.

[0058] The phasor measurement unit is a device for accurately measuring voltage and current phasors in a power system, which can provide high-precision time synchronization data to help monitor the status of the power system in real time. The micro-meteorological station is a compact device for monitoring and recording local environmental conditions (such as temperature, humidity, wind speed, wind direction, rainfall, etc.). The distribution fault indicator is a device installed on a power line for detecting and indicating the fault location in the line. The power quality monitoring device is used to monitor power quality parameters in the power system, such as voltage deviation, frequency deviation, current harmonics and voltage sag, etc. The aggregation node is a node in the network for centrally collecting and processing data from multiple sensors or devices. The high-speed communication link refers to a communication channel that provides high-speed data transmission in the network. The network loss (network loss) refers to the loss of electric energy caused by factors such as resistance and leakage during power transmission and distribution. The causal inference algorithm is used to determine the causal relationship between data or events. The ontology reasoning algorithm is based on a predefined ontology structure and discovers implicit knowledge and relationships through logical reasoning. The rule-based data quality verification method evaluates and verifies the quality of data through a set of predefined rules to ensure the accuracy, consistency and integrity of the data.

[0059] Various types of intelligent sensors are installed at the substations, distribution lines and distributed power sources of the distribution network, including phasor measurement units, micro-meteorological stations, distribution line monitoring devices, distribution fault indicators and power quality monitoring devices. Various types of heterogeneous data in the operation of the distribution network are collected through intelligent sensors, including voltage, current, power, temperature, humidity, wind speed, wind direction, fault information, harmonic content, etc. The collected heterogeneous data are combined into a data set to form multi-dimensional distribution operation parameters;

[0060] On the sensor node side, select the appropriate communication protocol according to the data type and communication requirements. On the aggregation node side, deploy the edge computing gateway to fuse and convert the data of different communication protocols to achieve interoperability between protocols. Use high-speed communication links (such as 5G, optical fiber, etc.) to upload the fused and converted data to the edge computing server corresponding to the distribution Internet of Things. The edge computing server constructs the energy domain ontology based on the concepts, terms and relationships in the energy field, defines the equipment, topology and environment in the distribution network as core concepts, and takes attribute association, spatial adjacency and logical dependency as semantic relationships. Use ontology description languages ​​(such as OWL, RDF) to formally represent the energy domain ontology, generate an energy domain knowledge base, apply association rule mining algorithms to calculate the support between load, voltage and network loss in the historical operation data of the distribution network, determine the association pattern between them, combine the time series clustering algorithm to cluster the operation data of different time scales, determine the corresponding operation mode, and apply the causal inference algorithm to analyze the causal relationship between equipment failure and power quality substandard events in the distribution network to obtain the operation law of the distribution network.

[0061] The edge computing server receives the fused and converted data, represents it semantically according to time, space and attributes, and maps it to the concepts and relationships in the energy domain ontology. It applies ontology reasoning algorithms (such as rule-based reasoning and description logic-based reasoning) and combines the energy domain ontology and the operation rules of the distribution network to enrich the data semantically. It automatically annotates the business semantics, spatiotemporal semantics and association semantics corresponding to the current data. It uses a rule-based data quality verification method to detect conflicts in the annotated semantic content to ensure the correctness and consistency of the semantic annotation. It converts the annotated semantic content into a triple (subject-predicate-object) form to construct nodes and edges of the distribution knowledge graph, where nodes represent equipment, users and environments in the distribution network, and edges represent the association relationship between entities.

[0062] Based on the distribution knowledge graph and the needs of power grid operation and management, load demand, network loss level and voltage qualification rate are determined as key indicators. The graph convolutional neural network is used to perform feature learning on the distribution knowledge graph, aggregate node information and structural information, and obtain a low-dimensional vector representation of each node. The initial prediction model is established in combination with the long short-term memory network. The historical operation data of the distribution network is used as a training set to input the model, generate prediction results and compare them with the true labels corresponding to each data, calculate the loss value, and optimize the hyperparameters of the initial prediction model based on the loss value, such as adjusting the learning rate, number of hidden layers, regularization parameters, etc. The prediction is repeated until the model converges to obtain a dynamic prediction model. The trained dynamic prediction model is used to predict the key indicators (load demand, network loss level, voltage qualification rate) of the distribution network at different time scales (such as short-term, medium-term, and long-term). The prediction results are visualized to generate a prediction map of key indicators to intuitively display the operation trends and potential risks of the distribution network at different time scales in the future.

[0063] In this embodiment, by deploying various types of intelligent sensors at key nodes of the distribution network, various heterogeneous data in the operation process of the distribution network can be comprehensively collected to form multi-dimensional distribution operation parameters, which provides sufficient data support for subsequent data analysis and decision optimization. The edge computing gateway realizes the fusion conversion of data of different protocols, and uses high-speed communication links to upload data, ensuring the efficiency and reliability of data transmission. By applying association rule mining, time series clustering, causal inference and other algorithms, the association patterns and operation laws of load, voltage and network loss are discovered from the historical operation data of the distribution network, revealing the causal relationship between equipment failure and power quality problems, providing decision support for intelligent operation management, and using ontology reasoning algorithms and data quality verification methods to perform Perform semantic enrichment and consistency check, build a distribution knowledge graph containing entities and associations such as equipment, users, and environment, form a semantic representation of the global state of the distribution network, and facilitate knowledge reasoning and intelligent analysis. Combined with graph convolutional neural networks and long short-term memory networks, make full use of the structural information and timing characteristics of the distribution knowledge graph, establish a dynamic prediction model, and predict key indicators such as load demand, network loss level, and voltage qualification rate at multiple time scales, providing a basis for early warning and optimized control of the distribution network. In summary, this embodiment realizes comprehensive perception of the operating status of the distribution network, intelligent mining of operating laws, and dynamic prediction and early warning of key indicators by constructing a knowledge graph and dynamic prediction model for the energy field, providing strong support for the intelligent operation and optimized control of the distribution network.

[0064] In an optional embodiment,

[0065] The support between load, voltage and network loss in the distribution network is determined by calculating the historical operation data of the distribution network through the association rule mining algorithm as shown in the following formula

[0066]

[0067] Among them, SUP represents support, α represents the time decay factor, t0 represents the current time, t i represents the timestamp of each historical data, β represents the spatial impact factor, d(X, Y, Z) represents the spatial distance function, which is used to measure the spatial distance between load, voltage and network loss, X represents load, Y represents voltage, Z represents network loss, I() represents the indicative function, T i represents the i-th historical data, and N represents the total number of historical data.

[0068] In this embodiment, by introducing a time attenuation factor to assign different weights to historical data at different times, the impact of past data on the current association rule mining results can be effectively reduced, and the timeliness and accuracy of the mined association rules can be improved. By introducing spatial influencing factors and spatial distance functions, the impact of the relative positions of load, voltage and network loss in space on their association relationships is considered, which can better reflect the actual operation characteristics of the distribution network and improve the applicability of association rules. Through the improved support calculation formula, strongly correlated load-voltage-network loss patterns can be quickly identified in massive historical data, reducing invalid calculations and the generation of redundant rules. In summary, this embodiment improves the rationality of support calculation and the efficiency of association rule mining, provides a more accurate basis for distribution network optimization control, and has significant technical effects and application value.

[0069] S2. Based on the power distribution knowledge graph and the key indicator prediction graph, a multi-time and space scale collaborative optimization model is constructed and an objective function is set through the space-time data integration modeling technology. For different space-time scales, a plurality of physical optimization sub-models of different scales are constructed through the multi-time scale rolling optimization algorithm and the spatial hierarchical optimization algorithm. A market game optimization sub-model is constructed through the agent-based distributed game optimization method and interactively coupled with the physical optimization sub-model. The physical optimization sub-model is coordinated through multi-agent reinforcement learning. The market game optimization sub-model and the physical optimization sub-model after collaborative processing are interactively fed back and iteratively converged in combination with the adaptive evolutionary game algorithm to obtain a multi-scale collaborative optimization solution.

[0070] The spatiotemporal data integrated modeling technology integrates and unifies time and space data to simultaneously consider the dynamic changes of time and space in the analysis and optimization process. The multi-time scale rolling optimization algorithm makes optimization decisions at different time scales and continuously updates the optimization results over time. The spatial hierarchical optimization algorithm is an optimization method that handles spatial problems in a hierarchical manner. It divides the system into multiple spatial levels, optimizes each level separately, and solves the problem hierarchically, thereby achieving global optimality in complex spatial systems. The physical optimization submodel is a submodel built based on physical principles in large-scale optimization problems, which is used to simulate and optimize specific physical processes. The agent-based distributed game optimization algorithm uses agents to The market game optimization sub-model simulates the competition and cooperation behaviors of different market participants (such as producers, consumers, and traders). Through game theory analysis, the resource allocation and price formation mechanism in the market are optimized to achieve the goal of maximizing overall efficiency or fair distribution. The interactive coupling refers to the interaction and influence between different subsystems or elements in a complex system. The multi-agent reinforcement learning is a reinforcement learning method involving multiple agents learning and acting in the same environment. The adaptive evolutionary game algorithm combines evolutionary game theory and adaptive mechanisms to achieve optimization goals by simulating competition and adaptation behaviors in the process of biological evolution. The interactive feedback is a continuous information exchange and response mechanism between various parts of the system.

[0071] In an optional embodiment,

[0072] Based on the power distribution knowledge graph and the key indicator prediction graph, a multi-time and space scale collaborative optimization model is constructed and an objective function is set by using the space-time data integration modeling technology. For different time and space scales, multiple physical optimization sub-models of different scales are constructed by using the multi-time scale rolling optimization algorithm and the spatial hierarchical optimization algorithm. A market game optimization sub-model is constructed by using the agent-based distributed game optimization method and interactively coupled with the physical optimization sub-model. The physical optimization sub-model is coordinated by multi-agent reinforcement learning. The market game optimization sub-model and the coordinated physical optimization sub-model are interactively fed back and iteratively converged in combination with the adaptive evolutionary game algorithm. The multi-scale collaborative optimization solution includes:

[0073] Based on the power distribution knowledge graph and the key indicator prediction graph, a multi-spatiotemporal scale collaborative optimization model is constructed through the spatiotemporal data integration modeling technology and an objective function is set, wherein the multi-spatiotemporal scale collaborative optimization model is composed of multiple sub-models;

[0074] For different spatiotemporal scales, the time window and the optimized time domain are determined by a multi-time-scale rolling optimization algorithm. In each window, based on the multi-time-space scale data and the power distribution knowledge graph, the optimized time domain and the prediction information are updated according to the rolling mechanism to generate a multi-time-scale physical optimization sub-model. Based on the spatial characteristics of the distribution network, the spatial topological structure of the distribution network is determined and in each area, a spatial hierarchical physical optimization sub-model is constructed according to physical constraints and operating rules, wherein the spatial hierarchical physical optimization sub-model and the multi-time-scale physical optimization sub-model together constitute the physical optimization sub-model;

[0075] Based on the distributed game optimization algorithm, the market entities in the distribution network are regarded as players, and the utility function and strategy space corresponding to each player are determined. Each player optimizes its own decision-making behavior based on the distributed game, and the optimal response dynamic algorithm is used to make the players repeat the game and reach the game equilibrium, and the game hyperparameters are obtained. The market game optimization sub-model is constructed according to the game hyperparameters;

[0076] The decision results of the market players in the market game optimization submodel are transmitted to the physical optimization submodel and used as boundary conditions and input parameters of physical optimization. The physical optimization submodel optimizes the operation plan of the distribution network to obtain the distribution network state variables. The decision space of the market players is updated based on the distribution network state variables and the game optimization is repeated. The pre-set coupling variables and coupling functions are combined to interactively couple with the market game optimization submodel to generate an interactive coupling optimization model and solve it to obtain a decoupled optimal solution, wherein the decoupled optimal solution includes the optimal parameters of the market game optimization submodel and the physical optimization submodel;

[0077] The physical optimization sub-model is updated based on the optimal parameters, each physical optimization sub-model is used as an agent in combination with multi-agent reinforcement learning, a multi-agent system is constructed, and the optimization decision strategy is learned according to the reinforcement learning algorithm, a reward function is constructed with the goal of maximizing the cumulative reward, a state space, an action space and a state transfer function are defined, and the strategy function corresponding to each agent is dynamically adjusted according to real-time data until the cumulative reward reaches the maximum and the collaboration is completed;

[0078] For the physical optimization sub-model and the market game optimization sub-model after collaborative processing, an initial game population is generated by an adaptive evolutionary game algorithm, wherein each individual in the initial game population represents a decision, and the fitness of each individual is calculated according to the strategy distribution of the initial game population and the game environment, and an evolutionary operation is selected based on the fitness and a second-generation game population is generated, and the second-generation game population dynamically adjusts its own strategy according to its own fitness and the strategy distribution of individuals in the second-generation game population, and the evolutionary operation and the individuals in the second-generation game population are returned to the physical optimization sub-model and the market game optimization sub-model and the decision space is updated, and iterations are repeated until the strategy distribution converges, and the individual with the highest fitness value in the last iteration is output as a multi-scale collaborative optimization solution.

[0079] The multi-timescale physical optimization submodel is used to optimize physical processes at different time scales. The spatial hierarchical physical optimization submodel achieves more efficient optimization by hierarchically processing the physical processes of the system according to the spatial dimension. The player is a participant in game theory. Each player takes actions or strategies in the game to maximize its own utility. The utility function is a function in game theory used to measure the satisfaction or benefits obtained by the player after taking a specific strategy. The optimal response dynamic algorithm is an algorithm for solving game equilibrium. It iteratively updates the strategy of each player so that it can respond best to the strategies of other players. The game hyperparameter is a parameter that controls the behavior of the game model. Parameters, usually including learning rate, discount factor, strategy update rules, etc. The distribution network state variables are parameters used to describe the current state of the distribution network, such as voltage, current, power flow, etc. The multi-agent system is a system composed of multiple agents, which can make independent decisions, interact and collaborate. The agent can be a robot, a software agent or any entity with autonomous behavior. The reward is a feedback signal used in reinforcement learning and game theory to motivate the agent or player to take a certain behavior. The reward function is a function that defines the reward obtained by the agent after taking a certain action in a specific state. The evolutionary operation is a basic operation in the evolutionary algorithm, which is used to simulate the process of natural selection and genetic variation.

[0080] Based on the distribution knowledge graph and key indicator prediction graph, a multi-time and space scale collaborative optimization model is constructed through the integrated spatiotemporal data modeling technology, the optimization objective function is set, the optimization objectives and constraints are clarified, and the multi-time and space scale collaborative optimization model is divided into multiple sub-models, including a physical optimization sub-model and a market game optimization sub-model. For different time scales, the time window and optimization time domain are determined by a multi-time scale rolling optimization algorithm. In each time window, based on the multi-time and space scale data and the distribution knowledge graph, the optimization time domain and prediction information are updated according to the rolling mechanism, and a multi-time scale physical optimization sub-model is generated to model and solve optimization problems of different time scales.

[0081] Based on the spatial characteristics of the distribution network, the spatial topological structure of the distribution network is determined. In each spatial area, a spatial hierarchical physical optimization sub-model is constructed according to physical constraints and operating rules. The spatial hierarchical physical optimization sub-model and the multi-time scale physical optimization sub-model are combined to form a complete physical optimization sub-model. Based on the distributed game optimization algorithm, the market entities in the distribution network are regarded as players, and the utility function and strategy space corresponding to each player are determined to describe the decision-making behavior and goals of the players. Each player optimizes its own decision-making behavior based on the distributed game. Through the optimal response dynamic algorithm, the players are made to repeat the game and reach the game equilibrium, and the game hyperparameters are obtained. According to the game hyperparameters, the market game optimization sub-model is constructed;

[0082] The decision results of the market players in the market game optimization sub-model are transferred to the physical optimization sub-model as the boundary conditions and input parameters of the physical optimization. The physical optimization sub-model optimizes the operation plan of the distribution network to obtain the distribution network state variables, updates the decision space of the market players based on the distribution network state variables, and repeats the game optimization. Combined with the pre-set coupling variables and coupling functions, it is interactively coupled with the market game optimization sub-model to generate an interactive coupling optimization model, solve the interactive coupling optimization model, and obtain the decoupling optimal solution, including the optimal parameters of the market game optimization sub-model and the physical optimization sub-model;

[0083] Based on the optimal parameters, the physical optimization sub-model is updated. Combined with multi-agent reinforcement learning, each physical optimization sub-model is regarded as an agent to build a multi-agent system. The decision-making strategy is optimized according to the reinforcement learning algorithm. The reward function is constructed with the goal of maximizing the cumulative reward. The state space, action space and state transfer function are defined to describe the decision-making process of the agent. The strategy function corresponding to each agent is dynamically adjusted according to real-time data. Through continuous trial and error and learning, the cumulative reward is maximized to complete collaborative optimization.

[0084] For the physical optimization sub-model and market game optimization sub-model after collaborative processing, the initial game population is generated by the adaptive evolutionary game algorithm. Each individual represents a decision. The fitness of each individual is calculated according to the strategy distribution of the initial game population and the game environment. The evolutionary operation is selected based on the fitness to generate the second-generation game population. The second-generation game population dynamically adjusts its own strategy according to its own fitness and the strategy distribution of individuals in the population. The evolutionary operation and the individuals in the second-generation game population are returned to the physical optimization sub-model and the market game optimization sub-model, and the decision space is updated. The iteration is repeated until the strategy distribution converges, and the individual with the highest fitness value in the last iteration is output as the multi-scale collaborative optimization solution.

[0085] In this embodiment, by constructing a multi-time and space scale collaborative optimization model, the optimization requirements of the distribution network at different time scales and space scales are considered, so that the optimization scheme can adapt to the dynamic changes and regional differences of the distribution network, and improve the refinement and real-time performance of the optimization. By coupling the physical optimization sub-model and the market game optimization sub-model, the physical operation constraints of the distribution network and the game behavior of the market players are combined, and the coordinated optimization of technology and economy is achieved. While ensuring the safe and stable operation of the distribution network, the interests of the market players are taken into account, and the comprehensiveness and feasibility of the optimization scheme are improved. By introducing multi-agent reinforcement learning, each physical optimization sub-model is used as an intelligent agent, so that it can autonomously learn and adjust the optimization strategy according to real-time data and reward functions, and can adapt to changes in the operating environment of the distribution network, continuously improve the optimization performance, reduce manual intervention, and improve the adaptability and robustness of the optimization. Through the adaptive evolutionary game algorithm, the optimization problem is transformed into the evolutionary process of the game individual, which can jump out of the local optimum and find the global optimal solution, and improve the quality and reliability of the optimization scheme. In summary, this embodiment significantly improves the performance and effect of the distribution network optimization control, and provides a strong technical support for the safe, economical and efficient operation of the distribution network, and has broad application prospects and promotion value.

[0086] In an optional embodiment,

[0087] The utility function corresponding to each player is determined as follows:

[0088] U i (s i ,s -i )=w r ·R i (s i ,d -i )-w c ·C i (s i )+w f ·F i (s i ,s-i );

[0089] Among them, U i (s i ,s -i ) represents the utility function, w r Represents the return weight coefficient, R i (s i ,s -i ) represents the payoff function of the ith player, w c represents the cost weight coefficient, C i (s i ) represents the cost function of the ith player, w f represents the fairness weight coefficient, F i (s i ,s -i ) is the fairness function of the ith player, which indicates the fairness of the income distribution between different players in the game process, s i represents the strategy of the ith player, s -i Represents the strategy combination of all players except the i-th player.

[0090] In this embodiment, the utility function includes the player's benefit function, cost function and fairness function, so that the player will consider the cost constraints and the interests of other players while pursuing the maximization of his own interests, which promotes the fairness and rationality of the game process. The benefit function and fairness function in the utility function depend not only on the player's own strategy, but also on the strategy combination of other players, taking into account the mutual influence between players and the dynamics of the game process, so that the game model can more accurately describe the game relationship between market players. The parameterized design enables the game model to adapt to different market environments and player preferences, thereby improving the versatility and practicality of the model. In summary, this embodiment determines the utility function corresponding to each player, comprehensively considers the player's benefit, cost and fairness, introduces the mutual influence between players, improves the flexibility and adaptability of the game model, lays a solid foundation for constructing a market game optimization sub-model, provides a clear optimization goal for solving the game equilibrium, promotes the reasonable decision-making and behavior of market players, and provides a theoretical basis for the design and optimization of market mechanisms.

[0091] In an optional embodiment,

[0092] The objective function of the multi-temporal and spatial scale collaborative optimization model is shown in the following formula:

[0093]

[0094] Where j represents the index of the spatial dimension, h represents the index of the temporal dimension, M represents the total number of spatial dimensions, H represents the total number of temporal dimensions, and W represents the total number of spatial dimensions. c Represents the cost weight coefficient of the optimization process, C jh (x jh ) represents the cost function of the jth region at time h, x jh represents the decision variable, W g represents the constraint violation weight coefficient, λ jh represents the constraint violation penalty coefficient, g jh (x jh ) represents the constraint function of the jth region at time h, W r represents the risk weight coefficient, R jh (x jh ) represents the operating risk function of the jth region at time h.

[0095] In this embodiment, the optimization objectives in the spatial dimension and the time dimension are unified into a framework in the form of double summation, so that the optimization model can simultaneously consider the operating status of the distribution network in different regions and at different times, and realize the collaborative optimization of multiple spatiotemporal scales. The objective function contains three weighted items, which respectively represent the cost of the optimization process, the constraint violation penalty and the operating risk. By introducing the corresponding weight coefficients, the importance of these three factors in the optimization objectives can be flexibly adjusted to achieve the balance and trade-off of multiple objectives. By introducing the constraint violation penalty item, the objective function can handle the constraint violation in the optimization process, so that the optimization model can allow a certain degree of constraint violation while satisfying the key constraints, thereby improving the adaptability and robustness of the optimization. In summary, this embodiment comprehensively considers the optimization objectives and constraints of the distribution network in different spatiotemporal dimensions, introduces regional and temporal differences, improves the adaptability and robustness of the optimization model, promotes risk management and safe operation, and provides a clear optimization direction and evaluation criteria for the collaborative optimization of multiple spatiotemporal scales, laying a solid foundation for realizing the economic, safe and reliable operation of the distribution network, and has important technical significance and application value.

[0096] S3. Encapsulate the evaluation rules and evaluation indicator system corresponding to the multi-scale collaborative optimization scheme into an evaluation contract, execute the multi-scale collaborative optimization scheme, and perform distributed evaluation on the multi-scale collaborative optimization scheme based on real-time operation data and the evaluation contract through the blockchain nodes in the pre-set blockchain network to generate an evaluation result. If the evaluation result does not meet the standard, the multi-spatiotemporal scale collaborative optimization model is encapsulated as an optimization contract through the blockchain smart contract technology, and the optimization contract is collaboratively triggered by the blockchain nodes and the candidate optimization scheme is re-solved based on the real-time operation data. Repeat the evaluation until the generated candidate optimization scheme meets the standard, and the candidate optimization scheme of the last iteration is used as the consensus scheme and executed.

[0097] The evaluation contract is a contract or agreement that defines how to evaluate a system, service or product. The evaluation rules are specific criteria or standards used to guide the evaluation process. The evaluation indicator system is a structured set of indicators used to systematically measure and evaluate the performance of specific goals. The blockchain node is a computing device participating in the blockchain network, responsible for maintaining the blockchain ledger, verifying transactions, disseminating information and executing consensus protocols. The distributed evaluation is an evaluation process performed simultaneously on multiple nodes or devices. The real-time operation data refers to the real-time data generated by the system during operation. The consensus scheme is a protocol or algorithm used to reach consensus among multiple nodes in blockchain and distributed systems.

[0098] In an optional embodiment,

[0099] The evaluation rules and evaluation index system corresponding to the multi-scale collaborative optimization scheme are encapsulated into an evaluation contract, the multi-scale collaborative optimization scheme is executed, and the multi-scale collaborative optimization scheme is distributedly evaluated based on the real-time operation data and the evaluation contract through the blockchain nodes in the pre-set blockchain network to generate an evaluation result. If the evaluation result does not meet the standard, the multi-spatiotemporal scale collaborative optimization model is encapsulated into an optimization contract through the blockchain smart contract technology, the optimization contract is collaboratively triggered by the blockchain nodes, and the candidate optimization scheme is re-solved based on the real-time operation data to generate a candidate optimization scheme, and the evaluation is repeated until the generated candidate optimization scheme meets the standard, and the candidate optimization scheme of the last iteration is used as the consensus scheme and executed, including:

[0100] Obtaining a preset evaluation index system and evaluation rules corresponding to the multi-scale collaborative optimization scheme, wherein the evaluation index system includes technical indicators, economic indicators and environmental indicators, and the evaluation rules include evaluation process, data source and evaluation cycle, and encapsulating the evaluation rules and the evaluation index system into an evaluation contract;

[0101] Execute the multi-scale collaborative optimization scheme to generate optimization results, upload the optimization results and the parameters of the multi-scale collaborative optimization scheme to the blockchain network, collect the real-time operation data of the distribution network through the blockchain node and automatically trigger the execution of the evaluation contract according to the preset evaluation cycle and evaluation conditions, calculate the real-time operation data according to the evaluation function in the evaluation contract, and generate the evaluation results in combination with the evaluation indicator system and the evaluation rules;

[0102] Whether the evaluation result meets the standard is determined according to the task requirements and pre-set constraints. If the evaluation result does not meet the standard, the objective function, constraints and decision variables of the multi-spatiotemporal scale collaborative optimization model are converted into functions and variables in the optimization contract according to the blockchain smart contract technology to generate an optimization contract. The blockchain nodes collaboratively trigger the optimization contract to obtain the real-time operation data and the evaluation result as input data, solve the multi-spatiotemporal scale collaborative optimization model, obtain the candidate optimization plan, upload the candidate optimization plan to the blockchain network and evaluate it again, and repeat the evaluation until the candidate optimization plan meets the standard;

[0103] The candidate optimization solution generated by the last iteration is uploaded to the blockchain network as the consensus solution. The blockchain node verifies the constraints of the consensus solution. If the consensus solution meets the constraints of all blockchain nodes, the consensus solution is executed. Otherwise, the consensus solution is deleted and the candidate optimization solution is regenerated.

[0104] The evaluation function is a mathematical expression or algorithm used to quantify or evaluate the performance of a system, process or solution, and the constraint verification is the process of checking whether a given solution, solution or process satisfies predefined constraints.

[0105] Determine the evaluation indicator system, including technical indicators, economic indicators and environmental indicators, determine the evaluation rules, including the evaluation process, data sources and evaluation cycle, encapsulate the evaluation rules and evaluation indicator system into an evaluation contract, execute the multi-scale collaborative optimization plan, generate optimization results, and upload the optimization results and parameters of the multi-scale collaborative optimization plan to the blockchain network;

[0106] The blockchain node collects the real-time operation data of the distribution network, automatically triggers the execution of the evaluation contract according to the pre-set evaluation cycle and evaluation conditions, calculates the real-time operation data according to the evaluation function in the evaluation contract, generates the evaluation results by combining the evaluation index system and evaluation rules, and judges whether the evaluation results meet the standards according to the task requirements and pre-set constraints. If the evaluation results do not meet the standards, the objective function, constraints and decision variables of the multi-temporal and spatial scale collaborative optimization model are converted into functions and variables in the optimization contract according to the blockchain smart contract technology to generate the optimization contract;

[0107] The blockchain nodes collaboratively trigger the optimization contract, obtain real-time operation data and evaluation results as input data, solve the multi-time and space scale collaborative optimization model, obtain candidate optimization solutions, upload the candidate optimization solutions to the blockchain network, evaluate them again, repeat the evaluation until the candidate optimization solutions meet the standards, take the candidate optimization solutions generated by the last iteration as the consensus solution, and upload the consensus solution to the blockchain network;

[0108] The blockchain nodes verify the constraints of the consensus plan. If the consensus plan meets the constraints of all blockchain nodes, the consensus plan is executed. If the consensus plan does not meet the constraints of all blockchain nodes, the consensus plan is deleted and a new candidate optimization plan is generated.

[0109] In this embodiment, by presetting the evaluation index system and evaluation rules and encapsulating them into an evaluation contract, the automatic evaluation of the distribution network optimization plan is realized. The automated evaluation mechanism improves the evaluation efficiency, reduces manual intervention, ensures the consistency and reliability of the evaluation process, and converts the objective function, constraint conditions and decision variables of the multi-temporal and spatial scale collaborative optimization model into functions and variables in the optimization contract. The optimization model is solved by using the computing power of the blockchain node to obtain the candidate optimization plan. The iterative optimization mechanism can continuously improve the quality and performance of the distribution network optimization plan and adapt to the dynamic changes in the operation status of the distribution network. Once the consensus plan is verified, it can be executed in the blockchain network, and all nodes recognize and follow the plan. The consensus execution mechanism utilizes the characteristics of decentralization, immutability and traceability of the blockchain, and improves the transparency, credibility and consistency of the execution of the distribution network optimization plan. In summary, this embodiment significantly improves the efficiency, reliability and intelligence level of the distribution network optimization, provides new ideas and methods for the operation optimization of the distribution network, has broad application prospects and promotion value, and is of great significance to promoting the development of energy Internet and smart energy.

[0110] Figure 2 FIG. 1 is a schematic diagram of a distribution network operation optimization system with source-grid-storage integration according to an embodiment of the present invention. Figure 2 As shown, the system comprises:

[0111] The first unit is used to respectively set up various types of smart sensors at the substations, distribution lines and distributed power sources in the distribution network, collect multi-dimensional distribution operation parameters through the smart sensors, and upload them to the edge computing server through the multi-protocol fusion distribution Internet of Things. The edge computing server uses the knowledge graph construction algorithm based on ontology reasoning to combine energy field knowledge and distribution network operation rules for semantic extraction and automatic annotation, and performs conflict correction on the annotated semantic content and determines the corresponding logical association to obtain the distribution knowledge graph. A dynamic prediction model is constructed based on the distribution knowledge graph, and the distribution index is predicted in combination with the graph convolutional neural network to obtain the prediction results and generate the key index prediction graph;

[0112] The second unit is used to construct a multi-space-time scale collaborative optimization model and set an objective function based on the power distribution knowledge graph and the key indicator prediction graph through the space-time data integration modeling technology, and for different space-time scales, construct multiple physical optimization sub-models of different scales through a multi-time scale rolling optimization algorithm and a spatial hierarchical optimization algorithm, construct a market game optimization sub-model through an agent-based distributed game optimization method and interact with the physical optimization sub-model, collaborate with the physical optimization sub-model through multi-agent reinforcement learning, and interactively feedback and iteratively converge the market game optimization sub-model and the physical optimization sub-model after collaborative processing in combination with an adaptive evolutionary game algorithm to obtain a multi-scale collaborative optimization solution;

[0113] The third unit is used to encapsulate the evaluation rules and evaluation indicator system corresponding to the multi-scale collaborative optimization scheme into an evaluation contract, execute the multi-scale collaborative optimization scheme, and perform distributed evaluation on the multi-scale collaborative optimization scheme based on real-time operation data and the evaluation contract through the blockchain nodes in the pre-set blockchain network to generate an evaluation result. If the evaluation result does not meet the standard, the multi-spatiotemporal scale collaborative optimization model is encapsulated as an optimization contract through the blockchain smart contract technology, and the optimization contract is collaboratively triggered by the blockchain nodes and a candidate optimization scheme is re-solved based on the real-time operation data. The evaluation is repeated until the generated candidate optimization scheme meets the standard, and the candidate optimization scheme of the last iteration is used as the consensus scheme and executed.

[0114] According to a third aspect of the embodiments of the present invention,

[0115] An electronic device is provided, comprising:

[0116] processor;

[0117] a memory for storing processor-executable instructions;

[0118] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0119] According to a fourth aspect of the embodiments of the present invention,

[0120] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0121] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0122] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A distribution network operation optimization method integrating source, network and storage, characterized in that: Various types of smart sensors are respectively set at substations, distribution lines and distributed power sources in the distribution network. Multi-dimensional distribution operation parameters are collected by the smart sensors and uploaded to the edge computing server through the multi-protocol fusion distribution Internet of Things. The edge computing server uses a knowledge graph construction algorithm based on ontology reasoning to combine energy field knowledge and distribution network operation rules for semantic extraction and automatic annotation, and performs conflict correction on the annotated semantic content and determines the corresponding logical association to obtain a distribution knowledge graph. A dynamic prediction model is constructed based on the distribution knowledge graph, and distribution index prediction is performed in combination with a graph convolutional neural network to obtain prediction results and generate a key indicator prediction graph; Based on the power distribution knowledge graph and the key indicator prediction graph, a multi-time and space scale collaborative optimization model is constructed and an objective function is set through the integrated time and space data modeling technology. For different time and space scales, a multi-time scale rolling optimization algorithm and a spatial hierarchical optimization algorithm are used to construct a plurality of physical optimization sub-models of different scales. A market game optimization sub-model is constructed through an agent-based distributed game optimization method and interactively coupled with the physical optimization sub-model. The physical optimization sub-model is coordinated through multi-agent reinforcement learning. The market game optimization sub-model and the coordinated physical optimization sub-model are interactively fed back and iteratively converged in combination with an adaptive evolutionary game algorithm to obtain a multi-scale collaborative optimization solution. The evaluation rules and evaluation indicator system corresponding to the multi-scale collaborative optimization scheme are encapsulated into an evaluation contract, the multi-scale collaborative optimization scheme is executed, and the multi-scale collaborative optimization scheme is distributedly evaluated based on the real-time operation data and the evaluation contract through the blockchain nodes in the pre-set blockchain network to generate an evaluation result. If the evaluation result does not meet the standard, the multi-spatiotemporal scale collaborative optimization model is encapsulated into an optimization contract through the blockchain smart contract technology, the optimization contract is collaboratively triggered by the blockchain nodes, and the candidate optimization scheme is re-solved based on the real-time operation data to generate the candidate optimization scheme, and the evaluation is repeated until the generated candidate optimization scheme meets the standard, and the candidate optimization scheme of the last iteration is used as the consensus scheme and executed.

2. The source-grid-storage integrated distribution network operation optimization method according to claim 1 is characterized in that: The intelligent sensor includes a phasor measurement unit, a micro-meteorological station, a distribution line monitoring device, a distribution fault indicator and a power quality monitoring device. Based on the intelligent sensor, various types of heterogeneous data are collected and combined into a data set to obtain the multi-dimensional distribution operation parameters; Based on the distribution Internet of Things with multi-protocol fusion, the communication protocol is selected on the sensor node side according to the data type and communication requirements. On the aggregation node side, the communication protocol is fused and converted through the edge computing gateway, and the fused and converted data is uploaded to the edge computing server corresponding to the distribution Internet of Things in combination with the high-speed communication link. The edge computing server constructs an energy domain ontology according to the concepts, terms and relationships in the energy domain, defines the equipment, topology and environment in the distribution network as core concepts, takes attribute association, spatial adjacency and logical dependency as semantic relationships, formally represents the energy domain ontology through an ontology description language, generates an energy domain knowledge base, calculates the support between the load, voltage and network loss in the distribution network in the historical operation data of the distribution network through an association rule mining algorithm, determines the association pattern based on the support, determines the operation mode corresponding to different time scales in combination with a time series clustering algorithm, and determines the causal relationship between equipment failure and power quality non-compliance events in the distribution network in combination with a causal inference algorithm, and obtains the operation law of the distribution network; The edge computing server semantically represents the received fused and converted data according to time, space and attributes and maps it to the concepts and relationships in the energy field ontology. Through the ontology reasoning algorithm, semantic enrichment is performed based on the energy field ontology and the operating rules of the distribution network, and the business semantics, spatiotemporal semantics and associated semantics corresponding to the current data are automatically labeled. The labeled semantic content is conflict-detected through a rule-based data quality verification method to obtain the labeled semantic content, and the labeled semantic content is converted into triples and the nodes and edges of the distribution knowledge graph are constructed, wherein the nodes represent the equipment, users and environment in the distribution network, and the edges represent the association relationship between entities, so as to obtain the distribution knowledge graph; Based on the power distribution knowledge graph and the power grid operation and management requirements, load demand, network loss level and voltage qualification rate are used as key indicators. Based on the key indicators, the power distribution knowledge graph is feature learned in combination with a graph convolutional neural network, and node information and structure information are aggregated to obtain a low-dimensional vector representation of each node. An initial prediction model is established in combination with a long short-term memory network. The historical operation data of the distribution network is added to the initial prediction model as a training set, and prediction results are generated and compared with the true labels corresponding to each data. The loss value is calculated and the hyperparameter optimization of the initial prediction model is performed based on the loss value. The optimization is repeated until the model converges to obtain the dynamic prediction model. The key indicators of the distribution network at different time scales are predicted by the dynamic prediction model to obtain the key indicator prediction graph.

3. The source-grid-storage integrated distribution network operation optimization method according to claim 2 is characterized in that: The support between load, voltage and network loss in the distribution network is determined by calculating the historical operation data of the distribution network through the association rule mining algorithm as shown in the following formula Among them, SUP represents support, α represents the time decay factor, t0 represents the current time, t i represents the timestamp of each historical data, β represents the spatial impact factor, d(X, Y, Z) represents the spatial distance function, which is used to measure the spatial distance between load, voltage and network loss, X represents load, Y represents voltage, Z represents network loss, I() represents the indicative function, T i represents the i-th historical data, and N represents the total number of historical data.

4. The source-grid-storage integrated distribution network operation optimization method according to claim 1, characterized in that: Based on the power distribution knowledge graph and the key indicator prediction graph, a multi-spatiotemporal scale collaborative optimization model is constructed through the spatiotemporal data integration modeling technology and an objective function is set, wherein the multi-spatiotemporal scale collaborative optimization model is composed of multiple sub-models; For different spatiotemporal scales, the time window and the optimized time domain are determined by a multi-time-scale rolling optimization algorithm. In each window, based on the multi-time-space scale data and the power distribution knowledge graph, the optimized time domain and the prediction information are updated according to the rolling mechanism to generate a multi-time-scale physical optimization sub-model. Based on the spatial characteristics of the distribution network, the spatial topological structure of the distribution network is determined and in each area, a spatial hierarchical physical optimization sub-model is constructed according to physical constraints and operating rules, wherein the spatial hierarchical physical optimization sub-model and the multi-time-scale physical optimization sub-model together constitute the physical optimization sub-model; Based on the distributed game optimization algorithm, the market entities in the distribution network are regarded as players, and the utility function and strategy space corresponding to each player are determined. Each player optimizes its own decision-making behavior based on the distributed game, and the optimal response dynamic algorithm is used to make the players repeat the game and reach the game equilibrium, and the game hyperparameters are obtained. The market game optimization sub-model is constructed according to the game hyperparameters; The decision results of the market players in the market game optimization submodel are transmitted to the physical optimization submodel and used as boundary conditions and input parameters of physical optimization. The physical optimization submodel optimizes the operation plan of the distribution network to obtain the distribution network state variables. The decision space of the market players is updated based on the distribution network state variables and the game optimization is repeated. The pre-set coupling variables and coupling functions are combined to interactively couple with the market game optimization submodel to generate an interactive coupling optimization model and solve it to obtain a decoupled optimal solution, wherein the decoupled optimal solution includes the optimal parameters of the market game optimization submodel and the physical optimization submodel; The physical optimization sub-model is updated based on the optimal parameters, each physical optimization sub-model is used as an agent in combination with multi-agent reinforcement learning, a multi-agent system is constructed, and the optimization decision strategy is learned according to the reinforcement learning algorithm, a reward function is constructed with the goal of maximizing the cumulative reward, a state space, an action space and a state transfer function are defined, and the strategy function corresponding to each agent is dynamically adjusted according to real-time data until the cumulative reward reaches the maximum and the collaboration is completed; For the physical optimization sub-model and the market game optimization sub-model after collaborative processing, an initial game population is generated by an adaptive evolutionary game algorithm, wherein each individual in the initial game population represents a decision, and the fitness of each individual is calculated according to the strategy distribution of the initial game population and the game environment, and an evolutionary operation is selected based on the fitness and a second-generation game population is generated, and the second-generation game population dynamically adjusts its own strategy according to its own fitness and the strategy distribution of individuals in the second-generation game population, and the evolutionary operation and the individuals in the second-generation game population are returned to the physical optimization sub-model and the market game optimization sub-model and the decision space is updated, and iterations are repeated until the strategy distribution converges, and the individual with the highest fitness value in the last iteration is output as a multi-scale collaborative optimization solution.

5. The source-grid-storage integrated distribution network operation optimization method according to claim 4 is characterized in that: The utility function corresponding to each player is determined as follows: U i (s i ,s -i )=w r ·R i (s i ,s -i )-w c ·C i (s i )+w f ·F i (s i ,s -i ); Among them, U i (s i ,s -i ) represents the utility function, w r Represents the return weight coefficient, R i (s i ,s -i ) represents the payoff function of the ith player, w c represents the cost weight coefficient, C i (s i ) represents the cost function of the ith player, w f represents the fairness weight coefficient, F i (s i ,s -i ) is the fairness function of the ith player, which indicates the fairness of the income distribution between different players in the game process, s i represents the strategy of the ith player, s -i Represents the strategy combination of all players except the i-th player.

6. The source-grid-storage integrated distribution network operation optimization method according to claim 4 is characterized in that: The objective function of the multi-temporal and spatial scale collaborative optimization model is shown in the following formula: Where j represents the index of the spatial dimension, h represents the index of the temporal dimension, M represents the total number of spatial dimensions, H represents the total number of temporal dimensions, and W represents the total number of spatial dimensions. c Represents the cost weight coefficient of the optimization process, C jh (x jh ) represents the cost function of the jth region at time h, x jh represents the decision variable, W g represents the constraint violation weight coefficient, λ jh represents the constraint violation penalty coefficient, g jh (x jh ) represents the constraint function of the jth region at time h, W r represents the risk weight coefficient, R jh (x jh ) represents the operating risk function of the jth region at time h.

7. The source-grid-storage integrated distribution network operation optimization method according to claim 1, characterized in that: Obtaining a preset evaluation index system and evaluation rules corresponding to the multi-scale collaborative optimization scheme, wherein the evaluation index system includes technical indicators, economic indicators and environmental indicators, and the evaluation rules include evaluation process, data source and evaluation cycle, and encapsulating the evaluation rules and the evaluation index system into an evaluation contract; Execute the multi-scale collaborative optimization scheme to generate optimization results, upload the optimization results and the parameters of the multi-scale collaborative optimization scheme to the blockchain network, collect the real-time operation data of the distribution network through the blockchain node and automatically trigger the execution of the evaluation contract according to the preset evaluation cycle and evaluation conditions, calculate the real-time operation data according to the evaluation function in the evaluation contract, and generate the evaluation results in combination with the evaluation indicator system and the evaluation rules; Whether the evaluation result meets the standard is determined according to the task requirements and pre-set constraints. If the evaluation result does not meet the standard, the objective function, constraints and decision variables of the multi-spatiotemporal scale collaborative optimization model are converted into functions and variables in the optimization contract according to the blockchain smart contract technology to generate an optimization contract. The blockchain nodes collaboratively trigger the optimization contract to obtain the real-time operation data and the evaluation result as input data, solve the multi-spatiotemporal scale collaborative optimization model, obtain the candidate optimization plan, upload the candidate optimization plan to the blockchain network and evaluate it again, and repeat the evaluation until the candidate optimization plan meets the standard; The candidate optimization solution generated by the last iteration is uploaded to the blockchain network as the consensus solution. The blockchain node verifies the constraints of the consensus solution. If the consensus solution meets the constraints of all blockchain nodes, the consensus solution is executed. Otherwise, the consensus solution is deleted and the candidate optimization solution is regenerated.

8. A source-grid-storage integrated distribution network operation optimization system, used to implement the source-grid-storage integrated distribution network operation optimization method described in any one of claims 1 to 7, characterized in that: include: The first unit is used to respectively set up various types of smart sensors at the substations, distribution lines and distributed power sources in the distribution network, collect multi-dimensional distribution operation parameters through the smart sensors, and upload them to the edge computing server through the multi-protocol fusion distribution Internet of Things. The edge computing server uses the knowledge graph construction algorithm based on ontology reasoning to combine energy field knowledge and distribution network operation rules for semantic extraction and automatic annotation, and performs conflict correction on the annotated semantic content and determines the corresponding logical association to obtain the distribution knowledge graph. A dynamic prediction model is constructed based on the distribution knowledge graph, and the distribution index is predicted in combination with the graph convolutional neural network to obtain the prediction results and generate the key index prediction graph; The second unit is used to construct a multi-space-time scale collaborative optimization model and set an objective function based on the power distribution knowledge graph and the key indicator prediction graph through the space-time data integration modeling technology, and for different space-time scales, construct multiple physical optimization sub-models of different scales through a multi-time scale rolling optimization algorithm and a spatial hierarchical optimization algorithm, construct a market game optimization sub-model through an agent-based distributed game optimization method and interact with the physical optimization sub-model, collaborate with the physical optimization sub-model through multi-agent reinforcement learning, and interactively feedback and iteratively converge the market game optimization sub-model and the physical optimization sub-model after collaborative processing in combination with an adaptive evolutionary game algorithm to obtain a multi-scale collaborative optimization solution; The third unit is used to encapsulate the evaluation rules and evaluation indicator system corresponding to the multi-scale collaborative optimization scheme into an evaluation contract, execute the multi-scale collaborative optimization scheme, and perform distributed evaluation on the multi-scale collaborative optimization scheme based on real-time operation data and the evaluation contract through the blockchain nodes in the pre-set blockchain network to generate an evaluation result. If the evaluation result does not meet the standard, the multi-spatiotemporal scale collaborative optimization model is encapsulated as an optimization contract through the blockchain smart contract technology, and the optimization contract is collaboratively triggered by the blockchain nodes and a candidate optimization scheme is re-solved based on the real-time operation data. The evaluation is repeated until the generated candidate optimization scheme meets the standard, and the candidate optimization scheme of the last iteration is used as the consensus scheme and executed.

9. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; Wherein, the processor is configured to call the instructions stored in the memory to execute the source-grid-storage integrated distribution network operation optimization method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by the processor, the source-grid-storage integrated distribution network operation optimization method described in any one of claims 1 to 7 is implemented.

Images