Power system reliability control method and system based on distributed generation cooperative optimization

Abstract

This invention provides a power system reliability control method based on distributed generation collaborative optimization, comprising the following steps: Step 1: Obtaining real-time operating parameters of the power system to obtain the optimal output combination scheme of distributed generation and conventional power source collaborative optimization; Step 2: Real-time monitoring of the operating status of distributed generation based on Internet of Things (IoT) technology; Step 3: Scheduling and controlling the active and reactive power output of distributed generation to ensure the reliability of power supply to important loads. This technical solution provides important technical support for constructing a highly robust and adaptable power system optimization scheduling strategy.

Description

Technical Field

[0001] This invention relates to the field of power system reliability control technology, and in particular to a power system reliability control method and system based on distributed power source collaborative optimization. Background Technology

[0002] In existing technologies, reliability control for distributed generation mainly adopts the following methods: First, during the distribution network planning stage, optimizing the location and capacity configuration of distributed generation to improve the static safety margin of the grid; second, during the distribution network operation stage, improving the voltage quality at the connection point of distributed generation through reactive power optimization, voltage control, and other means; and third, in the event of grid faults, reducing the scope of the accident impact through islanding operation and fault isolation. However, these methods are often limited to a single time scale and specific scenarios, lacking a collaborative interaction mechanism between distributed generation and centralized generation, and between the distribution network and active distribution network, making it difficult to leverage the flexible adjustment capabilities and system support role of distributed generation. Summary of the Invention

[0003] In view of this, the purpose of this invention is to provide a power system reliability control method and system for distributed power source collaborative optimization, which provides important technical support for building a highly robust and adaptable power system optimization scheduling strategy.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: a power system reliability control method based on distributed generation collaborative optimization, comprising the following steps:

[0005] Step 1: Obtain the real-time operating parameters of the power system to obtain the optimal output combination scheme of distributed power sources and conventional power sources.

[0006] Step 2: Real-time monitoring of the operating status of distributed power sources based on Internet of Things (IoT) technology;

[0007] Step 3: Dispatch and control the active and reactive power output of distributed power sources to ensure the reliability of power supply to important loads.

[0008] In a preferred embodiment, in step 1, real-time operating parameters of the power system are acquired, including data on load demand, output of conventional generating units, power flow of transmission lines, line voltage phase angle, and bus voltage amplitude. Combined with distributed generation modeling parameters, the maximum entropy principle and adaptive dynamic programming algorithm are used to establish a multi-timescale power system optimal scheduling model that considers the spatiotemporal distribution characteristics and uncertainties of distributed generation. Big data mining technology and deep learning algorithms are used to analyze historical operating data of distributed generation to obtain the output probability distribution characteristics of different types of distributed generation. Combined with static security constraints and stability constraints of the power system, the scheduling model is solved using hybrid robust stochastic optimization theory to obtain the optimal output combination scheme for the coordinated optimization of distributed generation and conventional power sources.

[0009] In a preferred embodiment, in step 2, when a distributed power source failure is detected that leads to a decrease in the power supply reliability of the power system, the power system optimization flow model considering the distributed power source failure scenario is solved to optimize and adjust the output of conventional power sources and the switching status of reactive power compensation equipment in real time to ensure that the N-1 verification reliability of the power system meets the requirements. Using multi-agent collaborative optimization theory, combined with short-term load forecasting and distributed power source output forecasting results, the output of conventional power sources and distributed power sources in each time period of the power system is dynamically optimized and scheduled to minimize the impact of distributed power source access on the power grid, and a rolling optimization strategy is adopted to realize the real-time adaptive control of the power system.

[0010] In a preferred embodiment, in step 3, when an emergency fault occurs in the power system causing a partial power outage, graph theory and network topology analysis algorithms are used to identify and isolate the faulty area. Based on the distributed power source configuration and load importance level within the power outage area, a multi-objective optimization solution is used to solve the regional microgrid recovery and reconfiguration model. The active and reactive power outputs of the distributed power sources are rationally scheduled and controlled. While ensuring the reliability of power supply to important loads, the power supply to non-important loads is maximized, thereby improving the overall power supply reliability level and emergency resilience of the power system.

[0011] This invention also provides a power system reliability control system based on distributed power source collaborative optimization, and implements the power system reliability control method based on distributed power source collaborative optimization.

[0012] Compared with the prior art, the present invention has the following beneficial effects:

[0013] 1. Introducing probabilistic distribution constraints on distributed generation output into the power system optimal scheduling model, considering static security and stability constraints of the power grid, the optimal scheduling model is transformed into a robust optimization problem with probabilistic constraints using hybrid robust stochastic optimization theory. The complexity of the problem solution is reduced through scenario generation and reduction techniques. The hybrid robust stochastic optimization model is solved using column generation algorithms and Benders decomposition algorithms to obtain the optimal output combination scheme that satisfies the coordinated optimization of distributed and conventional power sources under various power system constraints. Post-simulation evaluation is then performed, and the optimal scheduling model is adaptively modified and improved based on the evaluation results.

[0014] 2. Historical operating data of distributed generation sources are analyzed to obtain the output probability distribution characteristics of different types of distributed generation sources. Combined with static security and stability constraints of the power system, a scheduling model is solved using hybrid robust stochastic optimization theory to obtain the optimal output combination scheme for coordinated optimization of distributed generation sources and conventional power sources. By integrating data-driven and model-driven methods, the operating patterns and uncertainties of distributed generation sources are fully explored, providing important technical support for constructing highly robust and adaptable power system optimization scheduling strategies. Detailed Implementation

[0015] The present invention will be further described below with reference to the embodiments.

[0016] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0017] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations according to this application; as used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise; furthermore, it should be understood that when the terms “comprising” and / or “including” are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0018] The power system reliability control method based on distributed generation cooperative optimization includes the following steps:

[0019] S101. Obtain real-time operating parameters of the power system, including load demand, output of conventional generating units, power flow of transmission lines, line voltage phase angle, and bus voltage amplitude. Combined with distributed generation modeling parameters, and using the maximum entropy principle and adaptive dynamic programming algorithm, establish a multi-timescale power system optimal scheduling model that considers the spatiotemporal distribution characteristics and uncertainties of distributed generation. Utilize big data mining technology and deep learning algorithms to analyze historical operating data of distributed generation, obtain the output probability distribution characteristics of different types of distributed generation, and combine the static security constraints and stability constraints of the power system. Solve the scheduling model using hybrid robust stochastic optimization theory to obtain the optimal output combination scheme for the coordinated optimization of distributed generation and conventional power.

[0020] S102. Based on IoT technology, the operating status of distributed power sources is monitored in real time. When a fault in a distributed power source is detected, which leads to a decrease in the reliability of the power supply, the power system optimization flow model considering the fault scenario of the distributed power source is solved to optimize and adjust the output of conventional power sources and the switching status of reactive power compensation equipment in real time to ensure that the reliability of the N-1 verification of the power system meets the requirements. Using the multi-agent collaborative optimization theory, combined with the results of short-term load forecasting and distributed power source output forecasting, the output of conventional power sources and distributed power sources in each period of the power system is dynamically optimized and scheduled to minimize the impact of distributed power source access on the grid. A rolling optimization strategy is adopted to realize the real-time adaptive control of the power system.

[0021] S103. When an emergency fault occurs in the power system, causing power outages in some areas, graph theory and network topology analysis algorithms are used to identify and isolate the faulty areas. Based on the configuration of distributed power sources and the importance level of the loads in the power outage areas, a multi-objective optimization solution is used to solve the regional microgrid recovery and reconfiguration model. The active and reactive power outputs of distributed power sources are rationally scheduled and controlled. While ensuring the reliability of power supply to important loads, the power supply to non-important loads is maximized, thereby improving the overall power supply reliability level and emergency resilience of the power system.

[0022] Example 1:

[0023] By combining distributed generation modeling parameters and applying the maximum entropy principle and adaptive dynamic programming algorithm, a multi-timescale power system optimal scheduling model considering the spatiotemporal distribution characteristics and uncertainties of distributed generation is established, including:

[0024] Key parameters of different types of distributed power sources are obtained through on-site surveys and equipment monitoring. Data cleaning and feature engineering techniques are used to preprocess the obtained distributed power source parameter data. Machine learning algorithms are then used to establish the mapping relationship between distributed power source parameters and their output characteristics, resulting in accurate distributed power source modeling parameters.

[0025] By introducing the maximum entropy principle, the geographical distribution of distributed generation sources in the power system is regarded as a random variable. With the key attribute parameters of distributed generation sources as constraints, the maximum entropy model is established using the Lagrange multiplier method. The probability density function of the location distribution of distributed generation sources is obtained by solving the model. Furthermore, a large number of location distributions of distributed generation sources are randomly generated using the Monte Carlo simulation method. The spatial distribution characteristics of distributed generation sources in the power system and their impact on power grid power flow calculation, voltage distribution, etc. are statistically analyzed.

[0026] Considering the uncertainties of distributed generation, a scenario set is introduced into the power system optimization scheduling model. The complexity of model solution is reduced based on scenario reduction technology. An adaptive dynamic programming algorithm is used to decompose the multi-time-scale optimization scheduling problem into multiple sub-problems. For each sub-problem, a power system optimization power flow model considering the access of distributed generation is established and the spatiotemporal distribution characteristics of distributed generation are embedded. The optimal scheduling scheme for each time period is obtained by solving the problem.

[0027] The state variables of each sub-problem are updated in real time during the dynamic programming process. By adaptively adjusting and optimizing the time domain and spatial granularity, the efficiency and accuracy of model solving are balanced to obtain the power system optimization scheduling strategy for all time periods. The multi-time scale optimization scheduling model is integrated into the power system energy management system and interacts with functional modules such as load forecasting and electricity price forecasting to achieve collaborative optimization scheduling of distributed power sources and conventional power sources.

[0028] For example,

[0029] To establish a multi-timescale power system optimal scheduling model that considers the spatiotemporal distribution characteristics and uncertainties of distributed generation sources, the following methods and steps are required:

[0030] 1. Acquisition and processing of distributed power generation modeling parameters

[0031] 1.1 Obtain key parameters of different types of distributed power sources, including rated capacity, output power, efficiency curve, and failure rate, through on-site surveys and equipment monitoring.

[0032] 1.2 Data cleaning and feature engineering techniques are used to preprocess the acquired distributed power source parameter data, remove outliers, and normalize and standardize the data to achieve a standardized representation.

[0033] 1.3 Machine learning algorithms such as support vector machines and random forests are used to establish the mapping relationship between distributed power source parameters and their output characteristics, so as to obtain accurate distributed power source modeling parameters.

[0034] 2. Spatiotemporal distribution characteristics of distributed power sources based on the maximum entropy principle

[0035] 2.1 Introducing the maximum entropy principle, the geographical distribution of distributed power sources in the power system is regarded as a random variable, and its spatiotemporal distribution characteristics are characterized by constructing a probability distribution function;

[0036] 2.2 Using the key attribute parameters of distributed power sources (such as capacity, type, distance from load, etc.) as constraints, the maximum entropy model is established using the Lagrange multiplier method to solve for the probability density function of the location distribution of distributed power sources;

[0037] 2.3 The Monte Carlo simulation method is adopted to randomly generate the location distribution of a large number of distributed power sources based on the probability density function, and statistical analysis is performed on their spatial distribution characteristics in the power system and their impact on power flow calculation, voltage distribution, etc.

[0038] 3. Adaptive dynamic programming algorithm for solving multi-timescale optimal scheduling models

[0039] 3.1 Considering the uncertainties of distributed power sources, a set of scenarios is introduced into the power system optimization scheduling model, and the complexity of model solution is reduced based on scenario reduction techniques;

[0040] 3.2 An adaptive dynamic programming algorithm is adopted to decompose the multi-time-scale optimization scheduling problem into multiple sub-problems. The sub-problems are coupled through state variables to form a nested recursive optimization structure.

[0041] 3.3 For each sub-problem, a power system optimization flow model considering the access of distributed generation sources is established, and the spatiotemporal distribution characteristics of distributed generation sources are embedded to solve for the optimal scheduling scheme for each time period.

[0042] 3.4 During the dynamic programming process, the state variables of each sub-problem are updated in real time. By adaptively adjusting and optimizing the time and space granularity, the efficiency and accuracy of model solving are balanced, and the power system optimization scheduling strategy for all time periods is obtained.

[0043] 4. Validation and Application of Multi-Time-Scale Optimization Scheduling Model

[0044] 4.1 Build a power system simulation platform with distributed generation, import typical power grid structure and load data, and test and verify the effectiveness of the optimized scheduling model;

[0045] 4.2 To address the issues of high computational complexity and slow convergence speed that may occur during the model solving process, parallel computing, heuristic search, and other techniques are employed for optimization and improvement to enhance the model's practicality.

[0046] 4.3 Integrate the multi-timescale optimization scheduling model into the power system energy management system, and interact with functional modules such as load forecasting and electricity price forecasting to achieve coordinated optimization scheduling of distributed power sources and conventional power sources, thereby improving the economy and reliability of the power system.

[0047] The above methods and steps detail how to combine distributed generation modeling parameters, apply the maximum entropy principle and adaptive dynamic programming algorithm, and establish a multi-timescale power system optimal dispatch model that considers the spatiotemporal distribution characteristics and uncertainties of distributed generation. By employing advanced technologies such as data mining and artificial intelligence, the output characteristics and distribution patterns of distributed generation can be deeply characterized, and this model can be integrated with the power system optimal dispatch model. This allows for a more accurate and efficient analysis of the impact of distributed generation access on the power grid, providing important theoretical basis and methodological support for formulating scientific and reasonable power system optimal dispatch strategies.

[0048] Example 2:

[0049] By utilizing big data mining techniques and deep learning algorithms, historical operating data of distributed generation sources are analyzed to obtain the output probability distribution characteristics of different types of distributed generation sources. Combined with static security and stability constraints of the power system, the scheduling model is solved using hybrid robust stochastic optimization theory to obtain the optimal output combination scheme for coordinated optimization of distributed generation sources and conventional power sources, including:

[0050] Real-time operating parameters of distributed power sources are collected by IoT devices such as smart meters and sensors to form a massive historical operating dataset. The collected historical operating data is cleaned and preprocessed, and time series decomposition technology is used to divide the historical operating data of distributed power sources into trend items, periodic items and random items to characterize their long-term change trends, periodic fluctuation patterns and random disturbance characteristics, respectively.

[0051] The association rule mining algorithm is used to discover the correlation pattern between the output of distributed power sources and environmental factors. The clustering analysis algorithm is used to divide the historical operation data of distributed power sources into several typical operation modes. A probabilistic graphical model is introduced to establish a spatiotemporal correlation model of the output of distributed power sources, which describes the dynamic evolution law of its output in time and space.

[0052] By employing deep learning models such as long short-term memory neural networks, the temporal characteristics and nonlinear relationships contained in the historical operation data of distributed power sources are learned. Attention mechanisms and transfer learning techniques are introduced to construct a multi-timescale distributed power source output probability prediction model with cross-scenario and cross-regional generalization prediction capabilities. Furthermore, the prediction results of multiple deep learning sub-models are fused through a model integration strategy to improve the stability and robustness of distributed power source output probability prediction.

[0053] In the power system optimal scheduling model, a probabilistic distribution constraint on the output of distributed generation sources is introduced. Considering the static security and stability constraints of the power grid, the optimal scheduling model is transformed into a robust optimization problem with probabilistic constraints using hybrid robust stochastic optimization theory. The complexity of solving the problem is reduced by scenario generation and reduction techniques. The hybrid robust stochastic optimization model is solved using column generation algorithms and Benders decomposition algorithms to obtain the optimal output combination scheme that satisfies the coordinated optimization of distributed and conventional power sources under various power system constraints. Post-simulation evaluation is performed, and the optimal scheduling model is adaptively modified and improved based on the evaluation results.

[0054] For example, in order to analyze the historical operating data of distributed power sources using big data mining techniques and deep learning algorithms, obtain the output probability distribution characteristics of different types of distributed power sources, and combine power system constraints with hybrid robust stochastic optimization theory to solve the scheduling model, thereby obtaining the optimal output combination scheme for the coordinated optimization of distributed power sources and conventional power sources, the following methods and steps are required:

[0055] 1. Collection and preprocessing of historical operating data of distributed power sources

[0056] 1.1 Real-time operating parameters of distributed power sources, including active power, reactive power, voltage, and current, are collected through IoT devices such as smart meters and sensors to form a massive historical operating dataset.

[0057] 1.2 The collected historical operation data is cleaned and preprocessed to remove noisy data such as missing values ​​and outliers, and the data is standardized by methods such as data normalization and feature scaling.

[0058] 1.3 Using time series decomposition technology, the historical operating data of distributed power sources are divided into trend terms, periodic terms, and random terms to characterize their long-term changing trends, periodic fluctuation patterns, and random disturbance characteristics, respectively.

[0059] 2. Analysis of the Probability Distribution Characteristics of Distributed Power Generation Output Based on Big Data Mining Technology

[0060] 2.1 Using association rule mining algorithms, we can discover the correlation patterns between distributed power generation output and environmental factors (such as wind speed and light intensity) to characterize the output distribution patterns of distributed power generation under different conditions.

[0061] 2.2 Clustering analysis algorithms, such as K-means and DBSCAN, are used to divide the historical operating data of distributed power sources into several typical operating modes, and the output probability distribution characteristics of each operating mode are statistically analyzed.

[0062] 2.3 Introduce probabilistic graphical models, such as Hidden Markov Models and Conditional Random Fields, to establish a spatiotemporal correlation model of distributed power output, and characterize the dynamic evolution of its output in time and space.

[0063] 3. Deep learning algorithms are used to construct a probability prediction model for distributed power generation output.

[0064] 3.1 Deep learning models such as Long Short-Term Memory Neural Network (LSTM) are used to learn the temporal characteristics and nonlinear relationships contained in the historical operation data of distributed power sources, and to construct a multi-time-scale output probability prediction model.

[0065] 3.2 Introduce attention mechanism and transfer learning technology to improve the generalization prediction ability of deep learning model for different types of distributed power sources, and realize the characterization of output probability distribution characteristics across scenarios and regions;

[0066] 3.3 By employing model ensemble strategies, such as Bagging and Boosting, the prediction results of multiple deep learning sub-models are integrated to improve the stability and robustness of distributed power source output probability prediction.

[0067] 4. Solving power system dispatch models using hybrid robust stochastic optimization theory

[0068] 4.1 Introduce the probability distribution constraint of distributed generation output into the power system optimization scheduling model, and consider the static security constraints (such as power flow balance, line capacity limitation, etc.) and stability constraints (such as voltage stability, frequency stability, etc.) of the power grid.

[0069] 4.2 By adopting hybrid robust stochastic optimization theory, the optimization scheduling model is transformed into a robust optimization problem with probabilistic constraints, and the problem-solving complexity is reduced through scenario generation and reduction techniques;

[0070] 4.3 The hybrid robust stochastic optimization model is solved by using column generation algorithm, Benders decomposition algorithm and other methods to obtain the optimal output combination scheme that satisfies the cooperative optimization of distributed power source and conventional power source under various constraints of power system;

[0071] 4.4 Conduct post-event simulation evaluation of the hybrid robust stochastic optimization scheduling strategy, analyze its impact on power system economics, reliability, environmental protection and other indicators, and make adaptive corrections and improvements to the optimization scheduling model based on the evaluation results.

[0072] The above methods and steps detail how to utilize big data mining techniques and deep learning algorithms to analyze historical operating data of distributed generation sources, obtain the output probability distribution characteristics of different types of distributed generation sources, and combine them with static security constraints and stability constraints of the power system. By solving the scheduling model through hybrid robust stochastic optimization theory, the optimal output combination scheme of distributed generation sources and conventional power sources is obtained. By integrating data-driven and model-driven methods, the operating patterns and uncertainties of distributed generation sources are fully explored, providing important technical support for constructing highly robust and adaptable power system optimization scheduling strategies.

[0073] Example 3:

[0074] Based on IoT technology, the operating status of distributed generation is monitored in real time. When a fault in a distributed generation is detected, leading to a decrease in the reliability of the power supply system, the power system optimization flow model considering the fault scenario of the distributed generation is solved to optimize and adjust the output of conventional power sources and the switching status of reactive power compensation equipment in real time. This ensures that the N-1 reliability check of the power system meets the requirements, including:

[0075] IoT devices such as smart sensors and smart meters are deployed at the distributed power source site to collect the operating parameters of the distributed power source in real time. The collected operating data is transmitted to the cloud server through a wireless communication network to realize centralized monitoring and management of the operating status of the distributed power source. Edge computing technology is used to deploy smart gateways at the distributed power source site to perform local preprocessing and analysis of the collected operating data to realize rapid detection and early warning of distributed power source faults.

[0076] Based on historical operating data, a parameter model of the distributed power supply under normal operating conditions is established using machine learning algorithms. This model is used to determine in real time whether the operating status of the distributed power supply is abnormal. When the operating parameters of the distributed power supply deviate from the normal range, the fault diagnosis process is triggered. The fault tree analysis, expert knowledge base and other methods are used to infer the fault type and cause of the distributed power supply, and corresponding emergency response strategies are formulated for different types of distributed power supply faults.

[0077] Based on the conventional power flow model, a set of distributed generation fault scenarios is introduced. Each scenario corresponds to a possible combination of distributed generation faults and is assigned a corresponding probability of occurrence. To minimize the operating cost of the power system, a stochastic optimization power flow model considering distributed generation fault scenarios is established. Decision variables include the output of conventional power sources and the switching status of reactive power compensation equipment. In addition, the power system N-1 verification reliability constraint is introduced into the optimization power flow model to ensure that the power system can meet the safe operation limit requirements under any distributed generation fault scenario.

[0078] When the IoT monitoring system detects a fault in a distributed power source, it triggers the real-time solution of the power system optimization flow model to generate the optimal scheduling strategy to deal with the distributed power source fault. The optimal scheduling strategy is then distributed to each conventional power source and reactive power compensation device through the energy management system, realizing the real-time optimization and adjustment of the power system operation mode. The system also continuously monitors the changes in the distributed power source fault situation. When the fault recovers or expands, the power system optimization flow model is triggered again to dynamically update the power system's optimal scheduling strategy.

[0079] Considering distributed generation failure scenarios, Monte Carlo simulation and other methods are used to evaluate the power supply reliability indicators of the power system, quantify the impact of distributed generation failures on the reliability of the power system, formulate power system reliability improvement measures for the weaknesses exposed in the reliability assessment, continuously track the operation and failure characteristics of distributed generation, and regularly update the power system optimization power flow model and reliability assessment model to achieve adaptive optimization operation of the power system in high-penetration distributed generation scenarios.

[0080] For example,

[0081] To monitor the operating status of distributed generation in real time based on IoT technology, when a distributed generation failure is detected that leads to a decrease in power system reliability, the following steps are required: Solve the power system optimization power flow model that considers distributed generation failure scenarios, and optimize and adjust the output of conventional power sources and the switching status of reactive power compensation equipment in real time to ensure that the N-1 reliability check of the power system meets the requirements.

[0082] 1. Construct a distributed power supply IoT monitoring system

[0083] 1.1 Deploy IoT devices such as smart sensors and smart meters at the distributed power source site to collect the operating parameters of the distributed power source in real time, including active power, reactive power, voltage, current, temperature, etc.

[0084] 1.2 The collected data on the operation of the distributed power source is transmitted to the cloud server via a wireless communication network (such as 4G / 5G, NB-IoT, etc.) to achieve centralized monitoring and management of the operating status of the distributed power source;

[0085] 1.3 Edge computing technology is adopted to deploy intelligent gateways at the distributed power source site to perform local preprocessing and analysis of the collected operating data, thereby enabling rapid detection and early warning of distributed power source faults.

[0086] 2. Distributed Power Source Fault Detection and Diagnosis Methods

[0087] 2.1 Based on historical operating data, a parameter model of the distributed power supply under normal operating conditions is established using machine learning algorithms (such as support vector machine, random forest, etc.) to determine in real time whether the operating state of the distributed power supply is abnormal;

[0088] 2.2 When the operating parameters of the distributed power source are detected to deviate from the normal range, the fault diagnosis process is triggered. Through methods such as fault tree analysis and expert knowledge base, the fault type and cause of the distributed power source are inferred.

[0089] 2.3 For different types of distributed power source failures, develop corresponding emergency response strategies, including fault isolation, power limiting, and orderly shutdown, to reduce the impact of distributed power source failures on the power system.

[0090] 3. Power system optimization flow model considering distributed generation failure scenarios

[0091] 3.1 Based on the conventional power flow model, a set of distributed generation fault scenarios is introduced. Each scenario corresponds to a possible combination of distributed generation faults and is assigned a corresponding probability of occurrence.

[0092] 3.2 To minimize the operating cost of the power system, a stochastic optimization power flow model considering distributed generation failure scenarios is established. Decision variables include the output of conventional power sources and the switching status of reactive power compensation equipment.

[0093] 3.3 Introduce N-1 verification reliability constraints in the optimized power flow model to ensure that the power system can meet the safe operation limits such as line current and node voltage under any distributed source failure scenario.

[0094] 4. Optimize and adjust the power system operation mode in real time.

[0095] 4.1 When the IoT monitoring system detects a fault in a distributed power source, it triggers the real-time solution of the power system optimization flow model to generate the optimal scheduling strategy to deal with the distributed power source fault.

[0096] 4.2 By distributing optimized scheduling strategies to each conventional power source and reactive power compensation equipment through the energy management system (EMS), the power system operation mode can be optimized and adjusted in real time to maintain the power supply reliability of the power system;

[0097] 4.3 Continuously monitor changes in the fault status of distributed power sources. When the fault is recovered or aggravated, re-trigger the solution of the power system optimization power flow model and dynamically update the power system optimization scheduling strategy.

[0098] 5. Power System Reliability Assessment and Improvement

[0099] 5.1 Considering the failure scenarios of distributed generation sources, Monte Carlo simulation and other methods are used to evaluate the power supply reliability indicators (such as SAIFI, SAIDI, etc.) of the power system and quantify the impact of distributed generation source failures on the reliability of the power system.

[0100] 5.2 To address the weaknesses revealed in the reliability assessment, formulate power system reliability improvement measures, such as optimizing the location of distributed power source access points, increasing distribution automation equipment, and introducing demand-side response mechanisms.

[0101] 5.3 Continuously track the operation status and fault characteristics of distributed power sources, and regularly update the power system optimization power flow model and reliability assessment model to achieve adaptive optimization operation of the power system in scenarios with high penetration of distributed power sources.

[0102] The above methods and steps detail how to monitor the operating status of distributed generation sources in real time based on IoT technology. When a distributed generation failure is detected, leading to a decrease in power system reliability, the power system optimization flow model considering distributed generation failure scenarios is solved to optimize and adjust the output of conventional power sources and the switching status of reactive power compensation equipment in real time, ensuring that the N-1 reliability check of the power system meets the requirements. By deeply integrating IoT technology, intelligent fault diagnosis technology, and power system optimization operation technology, a self-healing and robust smart grid operation control system is constructed, enhancing the power system's ability to cope with the uncertainties of distributed generation sources.

[0103] Example 4:

[0104] Utilizing multi-agent cooperative optimization theory and combining short-term load forecasting and distributed generation output forecasting results, this method dynamically optimizes and schedules the output of conventional and distributed power sources in different time periods of the power system. This minimizes the impact of distributed generation access on the grid, and employs a rolling optimization strategy to achieve real-time adaptive control of the power system, including:

[0105] The power system is divided into multiple sub-regions. In each sub-region, an intelligent dispatch agent is set up to be responsible for the optimized dispatch of conventional and distributed power sources. At the system level, a coordination and control agent is set up to coordinate the power balance and exchange between sub-regions and interact with the power market. A hierarchical distributed information exchange and decision-making mechanism is established to achieve seamless integration of sub-region internal optimization and system-level collaborative optimization.

[0106] For each sub-region, historical load data and distributed power generation output data are collected and preprocessed. Statistical learning and machine learning methods are used to construct a multi-model fusion regional load prediction model. Output prediction models considering physical characteristics and uncertainties are established for different types of distributed power sources.

[0107] Using intelligent dispatch agents as the basic unit, a dynamic optimization dispatch model for distributed power systems is established. The objective functions include minimizing generation costs, maximizing renewable energy consumption, and power balance. Considering the static security constraints and dynamic stability constraints of the power system, a stochastic optimization model with multiple time periods and scenarios is constructed. Game theory and collaborative learning mechanisms are introduced. Through policy iteration and information interaction among intelligent agents, the collaborative optimization dispatch of distributed power sources and conventional power sources is realized. And through the interaction between intelligent agents and the power market, the market-oriented operation of the power system is realized.

[0108] The model predictive control approach transforms the dynamic optimization scheduling problem of the power system into a rolling optimization problem within a finite time domain. At the beginning of each scheduling cycle, the short-term load forecast and distributed generation output forecast results are updated. Combined with the current power system operating status, a multi-agent collaborative optimization algorithm is triggered to generate the optimal scheduling strategy for a future period. The optimized scheduling strategy is decoded into control commands and distributed to each conventional power source and distributed power source through the energy management system to achieve real-time adaptive control of the power system.

[0109] To address the uncertainties in short-term load forecasting and distributed generation output forecasting, robust optimization theory is introduced to construct a power system optimization scheduling model with uncertainty immunity. Scenario generation and reduction techniques are employed to improve the solution efficiency of the robust optimization model. Online self-optimization of power system scheduling is achieved through adaptive dynamic adjustment of the scenario set and robustness parameters. Machine learning methods are introduced to continuously optimize the agent's scheduling strategy through reinforcement learning, contrastive learning, and other techniques, thereby enhancing the resilience and recovery capability of the power system in the face of uncertain disturbances.

[0110] For example, in order to utilize multi-agent cooperative optimization theory, combine short-term load forecasting and distributed generation output forecasting results, dynamically optimize and schedule the output of conventional and distributed power sources in the power system at various time periods, minimize the impact of distributed generation access on the power grid, and achieve real-time adaptive control of the power system using a rolling optimization strategy, the following methods and steps are required:

[0111] 1. Construct a hierarchical distributed power system optimal scheduling framework

[0112] 1.1 Divide the power system into multiple sub-regions, and set up an intelligent dispatch agent in each sub-region to be responsible for the optimized dispatch of conventional power sources and distributed power sources in the region;

[0113] 1.2 Set up a coordination and control agent at the system level to coordinate power balance and exchange between sub-regions and interact with the electricity market;

[0114] 1.3 Establish a hierarchical and distributed information interaction and decision-making mechanism to achieve seamless integration of sub-regional internal optimization and system-level collaborative optimization.

[0115] 2. Develop high-precision short-term load forecasting and distributed generation output forecasting models.

[0116] 2.1 For each sub-region, historical load data and distributed power generation output data are collected, and data cleaning and preprocessing are performed;

[0117] 2.2 By comprehensively utilizing statistical learning and machine learning methods, such as multiple linear regression, time series analysis, support vector machines, and neural networks, a multi-model fusion regional load forecasting model is constructed to improve the accuracy of short-term load forecasting;

[0118] 2.3 For different types of distributed power sources (such as wind power, photovoltaic, small hydropower, etc.), respectively establish output prediction models that take into account their physical characteristics and uncertainties, such as wind power prediction models based on numerical weather prediction (NWP) and photovoltaic power generation prediction models based on sky imaging.

[0119] 3. Dynamic Dispatch Model for Power Systems Based on Multi-Agent Cooperative Optimization Theory

[0120] 3.1 Using intelligent dispatch agent as the basic unit, a dynamic optimization dispatch model for distributed power systems is established. The objective functions include minimizing generation costs, maximizing renewable energy consumption, and achieving power balance.

[0121] 3.2 Considering the static security constraints of the power system (such as line power flow limits, node voltage limits, etc.) and dynamic stability constraints (such as N-1 verification, small disturbance angle stability, etc.), construct a stochastic optimization model with multiple time periods and multiple scenarios;

[0122] 3.3 By introducing game theory and collaborative learning mechanisms, the collaborative optimization scheduling of distributed power sources and conventional power sources can be achieved through policy iteration and information interaction among intelligent agents. Furthermore, the market-oriented operation of the power system can be realized through the interaction between intelligent agents and the power market.

[0123] 4. Rolling Optimization Scheduling Strategy

[0124] 4.1 The model predictive control (MPC) approach is adopted to transform the dynamic optimization scheduling problem of the power system into a rolling optimization problem within a finite time domain;

[0125] 4.2 At the beginning of each scheduling cycle, update the short-term load forecast and distributed generation output forecast results, and combine them with the current power system operating status to trigger a multi-agent collaborative optimization algorithm to generate the optimal scheduling strategy for a future period.

[0126] 4.3 The optimized scheduling strategy is decoded into control commands and distributed to conventional and distributed power sources through the energy management system to achieve real-time adaptive control of the power system.

[0127] 5. Robust optimization enhancement considering uncertainties

[0128] 5.1 To address the uncertainties in short-term load forecasting and distributed generation output forecasting, robust optimization theories, such as stochastic optimization and distributed robust optimization, are introduced to construct a power system optimization scheduling model with uncertainty immunity.

[0129] 5.2 By employing scenario generation and reduction techniques, the solution efficiency of the robust optimization model is improved, and online self-optimization of power system dispatch is achieved through adaptive dynamic adjustment of the scenario set and robustness parameters;

[0130] 5.3 Introduce machine learning methods, and continuously optimize the scheduling strategy of the intelligent agent through techniques such as reinforcement learning and contrastive learning, so as to improve the resilience and recovery capability of the power system in the face of uncertain disturbances.

[0131] 6. Simulation Verification and Practical Application

[0132] 6.1 Build a power system simulation platform, including multiple sub-modules such as conventional power sources, distributed power sources, loads, and power grids, and consider actual engineering constraints and limitations;

[0133] 6.2 The dynamic scheduling method of power system based on multi-agent cooperative optimization theory is tested and verified on the simulation platform to evaluate its optimization performance and computational efficiency, and is compared and analyzed with the traditional centralized scheduling method.

[0134] 6.3 Select a typical power distribution network as a demonstration project, apply the proposed hierarchical distributed dynamic dispatching method in an actual power system, conduct long-term operation monitoring and evaluation, summarize practical experience, and provide a basis for further improving and promoting the dispatching method.

[0135] The above method details how to utilize multi-agent collaborative optimization theory, combined with short-term load forecasting and distributed generation output forecasting results, to dynamically optimize and schedule the output of conventional and distributed power sources in the power system across different time periods. This minimizes the impact of distributed generation access on the grid and employs a rolling optimization strategy to achieve real-time adaptive control of the power system. By constructing a hierarchical distributed power system optimization scheduling framework, developing a high-precision prediction model, and introducing robust optimization and machine learning techniques, the proposed method significantly improves the power system's ability to cope with the uncertainties of distributed generation, enabling efficient, economical, and safe operation in scenarios with a high proportion of renewable energy.

[0136] Example 5:

[0137] When an emergency fault in the power system causes a partial power outage, graph theory and network topology analysis algorithms are used to identify and isolate the faulty area. Based on the distributed power source configuration and load importance level within the outage area, a multi-objective optimization solution is used to solve the regional microgrid recovery and reconfiguration model. This allows for the rational scheduling and control of the active and reactive power output of distributed power sources, including:

[0138] First, a physical topology model of the power system is constructed, which abstracts the physical equipment of the power system as nodes and edges of a graph. Nodes represent buses, edges represent lines or transformers, and the weight of the edges represents the electrical parameters of the lines or transformers. Information of intelligent electronic devices is embedded in the graph model to realize the integrated modeling of the physical layer and information layer of the power system.

[0139] When an emergency fault occurs in the power system, causing power outages in some areas, the power system topology changes are monitored in real time based on the constructed graph model using depth-first search or breadth-first search algorithms to quickly locate the fault point. Starting from the fault point, the minimum path tree algorithm is used to search for all nodes directly or indirectly connected to the fault point to form the fault-affected area. At the boundary bus of the fault-affected area, the fault area is isolated from the external power grid by changing the switch topology state to form an isolated regional microgrid.

[0140] Then, taking the isolated regional microgrid as the research object, considering the distributed power sources and load distribution within it, a multi-time-period, multi-scenario mixed integer nonlinear programming problem is established with the objective functions of maximizing the restoration of power supply to important loads, minimizing the load reduction and minimizing network losses, the active power output, reactive power output and switching status of distributed power sources as decision variables, and the constraints of power flow equations, voltage amplitude limits, line capacity limits and upper and lower limits of distributed power source output as constraints.

[0141] For the established multi-objective optimization model of microgrid restoration and reconfiguration, intelligent optimization algorithms such as genetic algorithm, particle swarm optimization algorithm or ant colony algorithm are used to solve it. Pareto front is generated by constructing fitness function and non-dominated sorting, and the problem is divided into multiple sub-problems by a distributed optimization architecture. The convergence of the global optimal solution is achieved by local intelligent agents through iterative game.

[0142] Based on the solution results of multi-objective optimization, the optimal topology and distributed power output plan for microgrid restoration and reconfiguration are determined. A hierarchical distributed control architecture is adopted to establish a coordination control mechanism between the microgrid control center and the local controllers of each distributed power source, so as to realize the automatic issuance and execution of scheduling instructions. Real-time adaptive control strategies such as rolling optimization and model predictive control are introduced to dynamically adjust the active and reactive power output of distributed power sources to ensure the voltage and frequency stability of the microgrid.

[0143] Finally, after the fault is repaired, the islanded microgrid is smoothly connected to the external power grid through synchronous inspection and automatic reclosing technology. During grid connection, power control and reactive power compensation are adopted to reduce the impact and disturbance at the moment of grid connection. After grid connection, the output scheduling and load management strategies of distributed power sources are dynamically optimized according to the operating status of the microgrid and the external power grid to realize the flexible and autonomous operation of the microgrid.

[0144] For example,

[0145] To address the issue of power system outages caused by emergency faults in certain areas, and to identify and isolate the affected areas using graph theory and network topology analysis algorithms, and to rationally schedule and control the active and reactive power output of distributed power sources by solving a regional microgrid recovery and reconfiguration model based on the distributed power source configuration and load importance levels within the outage area through multi-objective optimization, the following steps are required:

[0146] 1. Construct a physical topology model of the power system.

[0147] 1.1 Collect network topology data of the power system, including the connection relationships and electrical parameters of equipment such as buses, lines, transformers, and switches;

[0148] 1.2 Using graph theory modeling, the physical equipment of the power system is abstracted as nodes and edges of a graph, and a directed weighted graph model is constructed. Nodes represent buses, edges represent lines or transformers, and the weight of the edges represents the electrical parameters of the lines or transformers.

[0149] 1.3 Embed information on intelligent electronic devices (IEDs), such as protection devices and measuring devices, into the graph model to achieve integrated modeling of the physical layer and information layer of the power system.

[0150] 2. Fault Area Identification and Isolation Algorithm

[0151] 2.1 Based on graph model, depth-first search or breadth-first search algorithm can monitor changes in the power system topology in real time and quickly locate the fault point when an emergency fault causes a partial power outage.

[0152] 2.2 Starting from the fault point, a minimum path tree algorithm, such as Dijkstra's algorithm or Bellman-Ford algorithm, is used to search for all nodes directly or indirectly connected to the fault point to form the fault-affected area.

[0153] 2.3 At the boundary bus of the fault-affected area, the fault area is isolated from the external power grid by changing the switch topology state, forming an islanded regional microgrid.

[0154] 3. Regional Microgrid Recovery and Reconfiguration Model

[0155] 3.1 Taking isolated regional microgrids as the research object, considering the distributed power sources (such as micro gas turbines, energy storage, renewable energy, etc.) and load distribution within them, a multi-objective optimization model for microgrid recovery and reconfiguration is established;

[0156] 3.2 The objective function includes maximizing the restoration of power supply to critical loads, minimizing the load reduction, and minimizing network losses. The decision variables include the active power output, reactive power output, and switching status of distributed power sources.

[0157] 3.3 The constraints include power flow equations, voltage amplitude limits, line capacity limits, and upper and lower limits of distributed power generation output, forming a mixed integer nonlinear programming (MINLP) problem with multiple time periods and multiple scenarios.

[0158] 4. Multi-objective optimization solution algorithm

[0159] 4.1 To address the multi-objective characteristics and non-convexity of the microgrid recovery and reconfiguration model, intelligent optimization algorithms, such as genetic algorithm (GA), particle swarm optimization (PSO), and ant colony optimization (ACO), are employed for multi-objective optimization.

[0160] 4.2 The concept of Pareto optimal solution is introduced. By constructing a fitness function and non-dominated sorting, the Pareto front of the microgrid restoration and reconfiguration problem is generated, realizing the trade-offs and compromises among multiple objectives;

[0161] 4.3 A distributed optimization architecture is adopted, which divides the microgrid restoration and reconfiguration problem into multiple sub-problems. Local intelligent agents are responsible for solving each sub-problem, and convergence of the global optimal solution is achieved through iterative game theory.

[0162] 5. Distributed power source optimization scheduling and control strategies

[0163] 5.1 Based on the results of the multi-objective optimization solution, determine the optimal topology and distributed power generation output plan for microgrid restoration and reconfiguration, and generate corresponding control commands;

[0164] 5.2 A hierarchical distributed control architecture is adopted, and a coordination control mechanism is established between the microgrid control center and the local controllers of each distributed power source to realize the automatic issuance and execution of scheduling commands;

[0165] 5.3 To address the intermittent and fluctuating characteristics of distributed power sources, real-time adaptive control strategies, such as rolling optimization and model predictive control, are introduced to dynamically adjust the active and reactive power output of distributed power sources, ensuring the voltage and frequency stability of the microgrid.

[0166] 6. Synchronous grid connection of microgrids with external power grids

[0167] 6.1 After fault repair, synchronous inspection and automatic reclosing technologies are used to achieve a smooth grid connection process between the islanded microgrid and the external power grid.

[0168] 6.2 By implementing measures such as power control and reactive power compensation, the impact and disturbance at the moment of grid connection are reduced, ensuring the synchronization of voltage, frequency and phase angle between the microgrid and the external power grid;

[0169] 6.3 After grid connection, based on the operating status of the microgrid and the external power grid, the output scheduling and load management strategies of distributed power sources are dynamically optimized to achieve flexible and autonomous operation of the microgrid.

[0170] 7. Simulation Analysis and Practical Applications

[0171] 7.1 Based on the physical parameters and topology of the actual power system, a simulation platform for microgrid recovery and reconfiguration was built to test and verify the proposed fault area identification and isolation algorithm, multi-objective optimization model and solution method;

[0172] 7.2 Conduct extensive simulation experiments for different fault scenarios and load demands to evaluate the feasibility, effectiveness, and robustness of the microgrid recovery and reconfiguration strategy, and compare and analyze it with the traditional centralized recovery method;

[0173] 7.3 Select typical power distribution networks and industrial parks as demonstration projects, apply the proposed microgrid restoration and reconfiguration technology to actual fault recovery and emergency management, and conduct long-term operation monitoring and effect evaluation to provide practical basis for the promotion and application of the technology.

[0174] The above methods and steps detail how to identify and isolate faulty areas when a power system experiences an emergency fault leading to partial power outages. This is achieved using graph theory and network topology analysis algorithms. Based on the distributed generation configuration and load importance levels within the outage area, a multi-objective optimization solution is used to solve the regional microgrid recovery and reconfiguration model, enabling rational scheduling and control of the active and reactive power output of distributed generation. Through a series of innovative methods, including constructing a power system physical topology model, developing faulty area identification and isolation algorithms, establishing a microgrid recovery and reconfiguration optimization model, employing intelligent optimization algorithms, implementing optimized scheduling and control of distributed generation, and synchronizing the microgrid with the external power grid, the proposed technical solution significantly improves the power system's self-healing capability and recovery speed in the face of emergency faults, reduces the scope of accident impact, and ensures the reliability of power supply to critical loads, demonstrating broad application prospects.

Claims

1. A power system reliability control method based on distributed generation cooperative optimization, characterized in that, Includes the following steps: Step 1: Obtain the real-time operating parameters of the power system to obtain the optimal output combination scheme of distributed power sources and conventional power sources. Step 2: Real-time monitoring of the operating status of distributed power sources based on Internet of Things (IoT) technology; Step 3: Dispatch and control the active and reactive power output of distributed power sources to ensure the reliability of power supply to critical loads; In step 1, real-time operating parameters of the power system are acquired, including load demand, output of conventional generating units, power flow of transmission lines, line voltage phase angle, and bus voltage amplitude. Combined with distributed generation modeling parameters, the maximum entropy principle and adaptive dynamic programming algorithm are used to establish a multi-timescale power system optimization scheduling model that considers the spatiotemporal distribution characteristics and uncertainties of distributed generation. Using big data mining technology and deep learning algorithms, historical operating data of distributed generation are analyzed to obtain the output probability distribution characteristics of different types of distributed generation. Combined with static security constraints and stability constraints of the power system, the scheduling model is solved through hybrid robust stochastic optimization theory to obtain the optimal output combination scheme of distributed generation and conventional power. In step 2, when a distributed power source failure is detected, leading to a decrease in the power supply reliability of the power system, the power system optimization flow model considering the distributed power source failure scenario is solved to optimize and adjust the output of conventional power sources and the switching status of reactive power compensation equipment in real time to ensure that the N-1 verification reliability of the power system meets the requirements. Using multi-agent collaborative optimization theory, combined with short-term load forecasting and distributed power source output forecasting results, the output of conventional power sources and distributed power sources in each time period of the power system is dynamically optimized and scheduled to minimize the impact of distributed power source access on the grid. A rolling optimization strategy is adopted to achieve real-time adaptive control of the power system. In step 3, when an emergency fault occurs in the power system, causing power outages in some areas, graph theory and network topology analysis algorithms are used to identify and isolate the faulty areas. Based on the configuration of distributed power sources and the importance level of the loads in the power outage areas, a multi-objective optimization solution is used to solve the regional microgrid recovery and reconfiguration model. The active and reactive power outputs of distributed power sources are rationally scheduled and controlled. While ensuring the reliability of power supply to important loads, the power supply to non-important loads is maximized, thereby improving the overall power supply reliability level and emergency resilience of the power system.

2. A power system reliability control system with distributed power source collaborative optimization, characterized in that... The power system reliability control method based on distributed power source collaborative optimization as described in claim 1 is implemented.