Adaptive optimization method and system for power distribution network planning

Abstract

The invention relates to the technical field of power distribution networks, and discloses a self-adaptive optimization method and system for power distribution network planning, and the system comprises a data collection and processing module, a prediction analysis module, a scheme generation module, an evaluation and decision module, an execution and feedback module, and a self-adaptive optimization module. Through constructing a power distribution network multi-source data set and carrying out standardization processing, and setting a multi-objective optimization function for different planning scenes, pertinence of generation of different scales of power distribution network planning schemes is ensured, and meanwhile, load and new energy output prediction is carried out by adopting a spatial-temporal distribution prediction algorithm. The method can effectively cope with the uncertainty caused by the fluctuation of the operation data of the power grid, guarantees the unification of the perspectiveness and scientificity of the planning scheme, further reduces the planning decision deviation, and introduces the preference learning algorithm in the scheme evaluation process, so that the finally recommended planning scheme meets the technical economy requirements and also meets the actual management requirements.

Description

Technical Field

[0001] This invention relates to the field of power distribution network technology, specifically to an adaptive optimization method and system for power distribution network planning. Background Technology

[0002] A power distribution network is a network that receives electrical energy from the transmission network and regional power plants, and distributes it locally and tiered according to voltage to various users through distribution facilities. It consists of overhead lines, cables, poles, distribution transformers, disconnect switches, reactive power compensators and some auxiliary facilities, and plays an important role in distributing electrical energy in the power grid.

[0003] Currently, due to the involvement of various influencing factors in the distribution network planning process, the traditional evaluation models relied upon when selecting the optimal distribution network planning scheme cannot respond to changes in the grid operation status in real time. When load fluctuations and abnormal output of new energy sources occur, the planning scheme will deviate significantly from the actual needs, and the applicability of the planning scheme cannot be guaranteed.

[0004] Therefore, an adaptive optimization method and system for distribution network planning are proposed to solve the above problems. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an adaptive optimization method and system for distribution network planning, which solves the problem that the planning schemes proposed in the background art deviate significantly from actual needs and cannot guarantee the applicability of the planning schemes.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an adaptive optimization method and system for distribution network planning, the method comprising the following steps: S1. Collect multi-source heterogeneous data of the distribution network, including historical load data, real-time operation data, equipment status data, geographic information data and distributed energy access data, and generate a multi-source data set of the distribution network; S2. Perform data cleaning, normalization, and feature extraction processing on the multi-source data set of the distribution network to construct a distribution network planning feature dataset; S3. Based on the aforementioned power distribution network planning feature dataset, load forecasting and new energy output forecasting are performed using a spatiotemporal distribution prediction algorithm to generate power distribution network planning prediction data. S4. Based on the power distribution network planning and forecasting data and combined with the planning constraints, generate multiple alternative power distribution network planning schemes through a multi-objective optimization algorithm. S5. Use an adaptive evaluation model to comprehensively evaluate the multiple alternative power distribution network planning schemes and generate scheme evaluation result data; S6. Based on the evaluation results of the above schemes, a preference learning algorithm is used to identify decision preferences, and a recommended scheme is output through an interactive decision interface; S7. Implement the distribution network planning scheme according to the recommended scheme, monitor the operation effect in real time, and generate feedback data; S8. Based on the feedback data, dynamically update the adaptive evaluation model and preference learning algorithm to achieve continuous optimization of the power distribution network planning scheme.

[0007] Preferably, the collection of multi-source heterogeneous data from the distribution network in step S1 includes: S11. Collect multi-source heterogeneous data of the power distribution network through smart meters, SCADA systems, IoT sensors and GIS platforms; S12. Perform time alignment and spatial correlation processing on the collected data to generate a multi-source data set of the distribution network with a unified spatiotemporal reference.

[0008] Preferably, the construction of the distribution network planning feature dataset in step S2 includes: S21. Wavelet transform and principal component analysis are used to perform noise filtering and feature dimensionality reduction on the multi-source data set of the power distribution network. S22. Extract load time sequence characteristics, network topology characteristics, and environmental influencing factor characteristics to construct a distribution network planning feature dataset.

[0009] Preferably, the generation of distribution network planning and forecasting data in step S3 includes: S31. Short-term and medium-to-long-term load forecasting based on Long Short-Term Memory Network (LSTM) and Spatiotemporal Graph Convolutional Network (STGCN); S32. Considering the impact of weather factors and equipment aging, a probabilistic prediction method is used to generate a new energy output prediction range.

[0010] Preferably, the generation of multiple distribution network planning alternatives in step S4 includes: S41. Define a multi-objective function that considers economy, reliability, environmental friendliness, and scalability. S42. The multi-objective particle swarm optimization algorithm (MOPSO) is used to solve the Pareto optimal solution set and generate multiple alternative schemes for distribution network planning.

[0011] Preferably, the generated scheme evaluation result data in step S5 includes: S51. Construct a comprehensive evaluation indicator system that includes technical indicators, economic indicators, and risk indicators; S52. Calculate the index weights based on the entropy weight method and the CRITIC method, and rank the schemes using the TOPSIS method to generate scheme evaluation result data.

[0012] Preferably, the output recommendation scheme in step S6 includes: S61. A preference learning algorithm based on deep reinforcement learning is used to learn decision preferences from historical decision data; S62. Display the evaluation results of the solution through a human-computer interaction interface, receive feedback from decision-makers, and dynamically adjust the recommended solution.

[0013] Preferably, the continuous optimization of the distribution network planning scheme in step S8 includes: S81. Collect distribution network operation data in real time and calculate the deviation between the actual effect and the expected effect of the scheme; S82. The evaluation model weights and preference learning model parameters are updated through an online learning mechanism to achieve adaptive optimization of the model.

[0014] Preferably, the system includes: The data acquisition and processing module collects heterogeneous data from the distribution network through the multi-source data access unit and generates a multi-source data set of the distribution network using the data cleaning and fusion unit. The predictive analysis module receives the multi-source data set of the distribution network and generates distribution network planning and prediction data through the load prediction unit and the new energy prediction unit. The scheme generation module generates multiple alternative schemes for power distribution network planning based on the predicted data through a multi-objective optimization algorithm unit. The evaluation and decision-making module receives the alternative solutions, generates solution evaluation results using the adaptive evaluation unit, and outputs recommended solutions through the preference learning unit and the interactive decision-making unit. The execution and feedback module performs deployment according to the recommended scheme, collects feedback data through the operation monitoring unit, and transmits it to the model update unit; The adaptive optimization module receives the feedback data and dynamically adjusts the evaluation model and preference learning algorithm through the model update unit to achieve continuous system optimization.

[0015] Preferably, the evaluation and decision-making module further includes: The visual interactive unit displays the evaluation results and recommended solutions, and provides a human-computer interaction interface for decision-makers to input their preferences and feedback. The real-time correction unit dynamically adjusts the recommended solution based on feedback data to ensure that the solution meets the actual decision-making needs.

[0016] Compared with existing technologies, the present invention provides an adaptive optimization method and system for distribution network planning, which has the following advantages: 1. In this invention, when optimizing the distribution network planning scheme, the standardization process is achieved by constructing and merging multi-source datasets of the distribution network, and by setting multi-objective optimization functions for different planning scenarios. This ensures the relevance of the distribution network planning schemes generated for different scales. At the same time, the spatiotemporal distribution prediction algorithm is used to predict load and new energy output, which can effectively cope with the uncertainty caused by fluctuations in power grid operation data, ensure the unity of the forward-looking and scientific nature of the planning scheme, and further reduce the deviation of planning decisions.

[0017] 2. In this invention, when conducting a comprehensive evaluation of power distribution network planning schemes, a comprehensive evaluation system including technical indicators, economic indicators, and risk indicators is established. An adaptive evaluation model is used to dynamically calculate the weights of the indicators, enabling the system to objectively quantify the comprehensive benefits of each alternative scheme. Furthermore, a preference learning algorithm is introduced during the scheme evaluation process, which can integrate the decision-maker's subjective preferences with the objective evaluation results in real time. This ensures that the final recommended planning scheme meets both technical and economic requirements and actual management needs, guaranteeing the feasibility of the planning scheme.

[0018] 3. In this invention, during the implementation of the distribution network planning scheme, a real-time feedback mechanism is established to continuously collect operational data, and online learning technology is used to dynamically update the evaluation model and optimization algorithm. This enables the system to continuously optimize the planning scheme according to the actual operating status of the power grid, realize the full life cycle management of the planning scheme, effectively improve the adaptability and long-term operational benefits of the distribution network planning, and further improve the refinement level and dynamic optimization capability of distribution network planning management. Attached Figure Description

[0019] Figure 1 This is a flowchart of an adaptive optimization method for distribution network planning according to the present invention; Figure 2 This is a diagram illustrating the architecture of an adaptive optimization system for power distribution network planning according to the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see Figure 1 - Figure 2 This is an adaptive optimization method and system for distribution network planning. The method includes the following steps: S1. Collect multi-source heterogeneous data of the distribution network, including historical load data, real-time operation data, equipment status data, geographic information data and distributed energy access data, and generate a multi-source data set of the distribution network; S2. Perform data cleaning, normalization, and feature extraction processing on the multi-source data set of the distribution network to construct a distribution network planning feature dataset; Data cleaning: refers to the process of identifying and processing errors, missing values, outliers, and inconsistencies in raw data to ensure data quality; Normalization: also known as standardization, is the process of scaling data of different scales and dimensions proportionally to make them fall into a specific range, thereby eliminating the influence of dimensions and accelerating model convergence. Commonly used methods include Min-Max normalization and Z-score standardization. formula: Min-Max normalization: ; in, Here, X represents the normalized value, and X represents the original data value. , These are the minimum and maximum values ​​of this feature in the sample; Z-score standardization: ; in, The standardized value. The mean of the sample. The standard deviation of the sample; Feature extraction: The process of constructing new features from raw data that are more informative, relevant, and suitable for the model; S3. Based on the distribution network planning feature dataset, a spatiotemporal distribution prediction algorithm is used to predict load and new energy output, generating distribution network planning prediction data. Spatiotemporal distribution prediction: refers to a prediction method that considers both time and space dimensions. Load and new energy output change with time and are also affected by geographical location and network topology factors. S4. Based on the distribution network planning and forecasting data and combined with the planning constraints, generate multiple alternative distribution network planning schemes through a multi-objective optimization algorithm. Multi-objective optimization function: In planning, it is necessary to optimize multiple conflicting objectives simultaneously. The claims mention economy, reliability, environmental protection, and scalability. Example of objective function dimension normalization: Economic objectives ; The total investment cost of the plan is... The highest cost among all alternative options; Reliability targets ; SAIDI is a system average outage time indicator; S5. Use an adaptive evaluation model to comprehensively evaluate multiple alternative plans for power distribution network planning and generate evaluation result data for the plans. S6. Based on the solution evaluation results data, a preference learning algorithm is used to identify decision preferences, and recommended solutions are output through an interactive decision interface. S7. Implement the distribution network planning scheme according to the recommended scheme, monitor the operation effect in real time, and generate feedback data; S8. Based on feedback data, the adaptive evaluation model and preference learning algorithm are dynamically updated to achieve continuous optimization of the power distribution network planning scheme; The collection of multi-source heterogeneous data from the distribution network in step S1 includes: S11. Collect multi-source heterogeneous data of the power distribution network through smart meters, SCADA systems, IoT sensors and GIS platforms; S12. Perform time alignment and spatial correlation processing on the collected data to generate a multi-source data set of the distribution network with a unified spatiotemporal reference. The construction of the distribution network planning feature dataset in step S2 includes: S21. Wavelet transform and principal component analysis are used to filter noise and reduce the dimensionality of features in the multi-source data set of the distribution network. S22. Extract load time-series characteristics, network topology characteristics, and environmental influencing factor characteristics to construct a distribution network planning feature dataset; The generation of distribution network planning and forecasting data in step S3 includes: S31. Short-term and medium-to-long-term load forecasting based on Long Short-Term Memory Network (LSTM) and Spatiotemporal Graph Convolutional Network (STGCN); LSTM: A special type of recurrent neural network that solves the gradient vanishing and exploding problems in long sequence training by introducing a "gate" mechanism, making it very suitable for time series prediction; Simplified illustration of the core computation of an LSTM unit: For a time step t: Forgotten Gate: ; Input Gate: ; Candidate cell status: ; Cell status update: ; Output gate: ; Hidden output: ; in, It is the Sigmoid activation function. For the time input, The hidden state at time t-1 The cell state at time t-1. , Here are the parameter matrix and bias vector to be trained, and * indicates element-wise multiplication; STGCN: A deep learning architecture that combines graph convolutional networks and time series processing models. It abstracts the power distribution network as a graph, uses graph convolution to capture spatial features, and uses temporal convolution and recurrent units to capture temporal features. It is very suitable for predicting spatiotemporal data with topological structures, such as power grids. S32. Considering the impact of weather factors and equipment aging, a probabilistic prediction method is used to generate a new energy output prediction range. Step S4, which generates multiple alternative distribution network planning schemes, includes: S41. Define a multi-objective function that considers economy, reliability, environmental friendliness, and scalability. S42. The multi-objective particle swarm optimization algorithm (MOPSO) is used to solve the Pareto optimal solution set and generate multiple alternative schemes for distribution network planning. Pareto optimal solution set: In multi-objective optimization, there is no single solution that is best for all objectives. A Pareto optimal solution is a solution that can be used to modify any objective without making any other objective worse. The set of all such solutions constitutes the Pareto front.

[0022] The general form of a multi-objective optimization problem: ; ; ; ;

[0023] in, Let m be a vector containing m objective functions. For the decision variable vector, Let m be the objective function. , For inequality and equality constraints, Let T be the feasible region of the decision variables, T be the transpose symbol, indicating that the objective function is a column vector, j be the index symbol, indicating the j-th inequality constraint, representing the various physical constraints and safety boundaries that must be satisfied in the distribution network planning, and k be the index symbol, indicating the k-th equality constraint, representing the physical laws that the distribution network must strictly follow. MOPSO is an extension of the particle swarm optimization algorithm in the multi-objective domain. It maintains an external archive to store non-dominated solutions and uses this archive to guide the flight and search of the particle swarm, ultimately outputting a set of approximate Pareto optimal solutions. The generated scheme evaluation result data in step S5 includes: S51. Construct a comprehensive evaluation indicator system that includes technical indicators, economic indicators, and risk indicators; S52. Calculate the index weights based on the entropy weight method and the CRITIC method, and rank the schemes using the TOPSIS method to generate scheme evaluation result data. Entropy weighting method: an objective weighting method that determines the weight based on the degree of variation of each indicator value. The greater the difference in the data sequence of an indicator, the more information it contains, and the greater its weight.

[0024] CRITIC method: Another objective weighting method that considers not only the contrast strength within indicators, but also the conflict between indicators. The greater the contrast strength and the stronger the conflict with other indicators, the greater the weight.

[0025] TOPSIS method: A multi-attribute decision-making method that constructs a "positive ideal solution" and a "negative ideal solution", and ranks each solution by calculating the relative distance between each solution and these two ideal solutions. The closer the solution is to the positive ideal solution and the farther it is from the negative ideal solution, the better the solution is. The core steps of the TOPSIS method: Constructing a normalized decision matrix: Normalizing the original data matrix; Construct a weighted normalized matrix: multiply each column of the normalized matrix by its corresponding weight; Determine the ideal solution: Positive Ideal Solution ,in ; Negative ideal solution ,in ; Calculate the distance: the distance from each solution to the ideal solution. ; Distance from each solution to the negative ideal solution ; Calculate relative proximity: ; The larger the value, the closer the solution is to the ideal solution, and the higher its ranking. The recommended output scheme in step S6 includes: S61. A preference learning algorithm based on deep reinforcement learning is used to learn decision preferences from historical decision data; Preference learning: learning from decision-makers' past choices and feedback to understand their implicit preferences for various goals and attributes, rather than directly assigning fixed weights.

[0026] Deep Reinforcement Learning (DRL): Combining the perceptual capabilities of deep learning with the decision-making capabilities of reinforcement learning, a framework is constructed in this context as follows: Agent: Recommendation system; State: The current set of alternative solutions and their evaluation metrics; Action: Recommend a specific option and adjust the weights to generate a new recommendation ranking; Reward: The decision-maker ultimately adopts the recommended solution, or rejects the recommended solution and makes other choices; By continuously trying and receiving rewards, the DRL model learns the recommendation strategy that best meets the decision-maker's preferences; S62. Display the evaluation results of the solution through the human-computer interaction interface, receive feedback from decision-makers, and dynamically adjust the recommended solution; The continuous optimization of the distribution network planning scheme in step S8 includes: S81. Collect distribution network operation data in real time and calculate the deviation between the actual effect and the expected effect of the scheme; S82. Update the evaluation model weights and preference learning model parameters through an online learning mechanism to achieve adaptive optimization of the model; Online learning: In contrast to batch learning, the model receives new data one by one and in small batches as a data stream, and immediately updates the model parameters incrementally, enabling the model to quickly adapt to changes in data distribution; The system includes: The data acquisition and processing module collects heterogeneous data from the distribution network through the multi-source data access unit and generates a multi-source data set of the distribution network using the data cleaning and fusion unit. The predictive analysis module receives multi-source data sets from the distribution network and generates distribution network planning and prediction data through the load prediction unit and the new energy prediction unit. The scheme generation module generates multiple alternative schemes for power distribution network planning based on the predicted data and through a multi-objective optimization algorithm unit. The evaluation and decision-making module receives alternative solutions, generates evaluation results using the adaptive evaluation unit, and outputs recommended solutions through the preference learning unit and the interactive decision-making unit. The execution and feedback module performs deployment according to the recommended plan, collects feedback data through the operation monitoring unit, and transmits it to the model update unit; The adaptive optimization module receives feedback data and dynamically adjusts the evaluation model and preference learning algorithm through the model update unit to achieve continuous system optimization. The assessment and decision-making module also includes: The visual interactive unit displays the evaluation results and recommended solutions, and provides a human-computer interaction interface for decision-makers to input their preferences and feedback. The real-time correction unit dynamically adjusts the recommended solution based on feedback data to ensure that the solution meets the actual decision-making needs.

[0027] An adaptive optimization method and system operation steps for distribution network planning are as follows: Step 1: Multi-source data fusion and feature engineering The system begins with the extensive collection of multi-source heterogeneous data from the power distribution network. This includes historical and real-time load data, equipment status, geographic information, and renewable energy access data obtained from smart meters, SCADA systems, and sensors. These raw data undergo rigorous cleaning, time alignment, and spatial correlation processing to form a unified spatiotemporal reference dataset. Subsequently, wavelet transform and principal component analysis techniques are used for noise filtering and feature dimensionality reduction to extract key load time-series features, network topology features, and environmental impact features, thus constructing a high-quality planning feature dataset that lays a solid data foundation for subsequent analysis.

[0028] Step 2: Spatiotemporal Prediction and Uncertainty Quantification The system utilizes the feature dataset constructed in the previous step and employs advanced spatiotemporal distribution prediction algorithms to predict load and renewable energy output. This step not only generates point predictions but, more importantly, considers weather and equipment aging factors. It outputs prediction intervals through probabilistic prediction methods, quantifying the uncertainty of future power grid conditions, thereby ensuring the foresight of the planning scheme and its adaptability to fluctuations.

[0029] Step 3: Multi-objective optimization to generate Pareto scheme Based on predictive data, the system combines multiple objectives and physical constraints of distribution network planning, including economy, reliability, environmental protection and scalability, and uses the multi-objective particle swarm optimization (MOPSO) algorithm to solve the problem. Its core is to find and generate a set of Pareto optimal solutions, that is, a series of alternative planning schemes that achieve the best balance among multiple objectives, rather than a single solution, thus providing decision-makers with a rich selection space.

[0030] Step 4: Adaptive Evaluation and Preference Learning Decision The system constructs a comprehensive evaluation index system and uses an adaptive model to dynamically calculate index weights to objectively and quantitatively evaluate and rank alternative solutions. At the same time, the system uses a preference learning algorithm based on deep reinforcement learning to mine and learn the decision-makers' implicit preferences from historical decision data. Through a human-computer interaction interface, the evaluation results and recommended solutions are presented to the decision-makers, and feedback can be received to dynamically adjust the output, ensuring that the final recommended solution is both scientific and objective and meets actual management needs.

[0031] Step 5: Execution Feedback and Model Self-Evolution After the recommended scheme is deployed and implemented, the system does not stop operating. It collects the actual operating effect data of the planning scheme through the real-time monitoring system and compares it with the expected effect to generate feedback deviation data. This valuable field data is input into the system and continuously and dynamically updates the weights of the evaluation model and the parameters of the preference learning algorithm through the online learning mechanism, forming a closed loop of "evaluation-decision-execution-feedback-optimization". This enables the entire system and planning scheme to evolve along with the operation of the power grid, achieving continuous adaptive optimization of the entire life cycle of distribution network planning.

[0032] In summary, the core principle of this system lies in constructing a complete adaptive loop of "data-driven prediction, multi-objective optimization generation, human-machine collaborative decision-making, and closed-loop feedback optimization," which ensures that the planning scheme has foresight, scientificity, and high adaptability throughout the entire process from formulation to implementation and updating.

[0033] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0034] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An adaptive optimization method for distribution network planning, characterized in that, The method includes the following steps: S1. Collect multi-source heterogeneous data of the distribution network, including historical load data, real-time operation data, equipment status data, geographic information data and distributed energy access data, and generate a multi-source data set of the distribution network; S2. Perform data cleaning, normalization, and feature extraction processing on the multi-source data set of the distribution network to construct a distribution network planning feature dataset; S3. Based on the aforementioned power distribution network planning feature dataset, load forecasting and new energy output forecasting are performed using a spatiotemporal distribution prediction algorithm to generate power distribution network planning prediction data. S4. Based on the power distribution network planning and forecasting data and combined with the planning constraints, generate multiple alternative power distribution network planning schemes through a multi-objective optimization algorithm. S5. Use an adaptive evaluation model to comprehensively evaluate the multiple alternative power distribution network planning schemes and generate scheme evaluation result data; S6. Based on the evaluation results of the above schemes, a preference learning algorithm is used to identify decision preferences, and a recommended scheme is output through an interactive decision interface; S7. Implement the distribution network planning scheme according to the recommended scheme, monitor the operation effect in real time, and generate feedback data; S8. Based on the feedback data, dynamically update the adaptive evaluation model and preference learning algorithm to achieve continuous optimization of the power distribution network planning scheme.

2. The adaptive optimization method for distribution network planning according to claim 1, characterized in that, The collection of multi-source heterogeneous data from the distribution network in step S1 includes: S11. Collect multi-source heterogeneous data of the power distribution network through smart meters, SCADA systems, IoT sensors and GIS platforms; S12. Perform time alignment and spatial correlation processing on the collected data to generate a multi-source data set of the distribution network with a unified spatiotemporal reference.

3. The adaptive optimization method for distribution network planning according to claim 1, characterized in that, The construction of the power distribution network planning feature dataset in step S2 includes: S21. Wavelet transform and principal component analysis are used to perform noise filtering and feature dimensionality reduction on the multi-source data set of the power distribution network. S22. Extract load time sequence characteristics, network topology characteristics, and environmental influencing factor characteristics to construct a distribution network planning feature dataset.

4. The adaptive optimization method for distribution network planning according to claim 1, characterized in that: The generation of distribution network planning and forecasting data in step S3 includes: S31. Short-term and medium-to-long-term load forecasting based on Long Short-Term Memory Network (LSTM) and Spatiotemporal Graph Convolutional Network (STGCN); S32. Considering the impact of weather factors and equipment aging, a probabilistic prediction method is used to generate a new energy output prediction range.

5. The adaptive optimization method for distribution network planning according to claim 1, characterized in that, The generation of multiple distribution network planning alternatives in step S4 includes: S41. Define a multi-objective function that considers economy, reliability, environmental friendliness, and scalability. S42. The multi-objective particle swarm optimization algorithm (MOPSO) is used to solve the Pareto optimal solution set and generate multiple alternative schemes for distribution network planning.

6. The adaptive optimization method for distribution network planning according to claim 1, characterized in that, The generated scheme evaluation result data in step S5 includes: S51. Construct a comprehensive evaluation indicator system that includes technical indicators, economic indicators, and risk indicators; S52. Calculate the index weights based on the entropy weight method and the CRITIC method, and rank the schemes using the TOPSIS method to generate scheme evaluation result data.

7. The adaptive optimization method for distribution network planning according to claim 1, characterized in that, The output recommendation scheme in step S6 includes: S61. A preference learning algorithm based on deep reinforcement learning is used to learn decision preferences from historical decision data; S62. Display the evaluation results of the solution through a human-computer interaction interface, receive feedback from decision-makers, and dynamically adjust the recommended solution.

8. The adaptive optimization method for distribution network planning according to claim 1, characterized in that, The continuous optimization of the distribution network planning scheme in step S8 includes: S81. Collect real-time power distribution network operation data and calculate the deviation between the actual and expected results of the scheme; S82. The evaluation model weights and preference learning model parameters are updated through an online learning mechanism to achieve adaptive optimization of the model.

9. An adaptive optimization system for distribution network planning, implementing the adaptive optimization method for distribution network planning as described in any one of claims 1-8, characterized in that, The system includes: The data acquisition and processing module collects heterogeneous data from the distribution network through the multi-source data access unit and generates a multi-source data set of the distribution network using the data cleaning and fusion unit. The predictive analysis module receives the multi-source data set of the distribution network and generates distribution network planning and prediction data through the load prediction unit and the new energy prediction unit. The scheme generation module generates multiple alternative schemes for power distribution network planning based on the predicted data through a multi-objective optimization algorithm unit. The evaluation and decision-making module receives the alternative solutions, generates solution evaluation results using the adaptive evaluation unit, and outputs recommended solutions through the preference learning unit and the interactive decision-making unit. The execution and feedback module performs deployment according to the recommended scheme, collects feedback data through the operation monitoring unit, and transmits it to the model update unit; The adaptive optimization module receives the feedback data and dynamically adjusts the evaluation model and preference learning algorithm through the model update unit to achieve continuous system optimization.

10. An adaptive optimization system for distribution network planning according to claim 9, characterized in that, The assessment and decision-making module also includes: The visual interactive unit displays the evaluation results and recommended solutions, and provides a human-computer interaction interface for decision-makers to input their preferences and feedback. The real-time correction unit dynamically adjusts the recommended solution based on feedback data to ensure that the solution meets the actual decision-making needs.

Images