Rural comprehensive energy planning method and system based on operation simulation and intelligent acceleration

Abstract

This invention discloses a method and system for rural integrated energy planning based on operational simulation and intelligent acceleration, relating to the field of new energy planning technology. The method includes the following steps: constructing a two-layer planning model with embedded time-series operational simulation; the planning layer model obtains a preliminary planning scheme and investment cost based on the network topology and resource parameters of the rural integrated energy system; the lower-layer operational model uses outlier detection to identify and evaluate outlier samples in the preliminary planning scheme, obtaining the operational cost, and uses neural networks to accelerate or refine the time-series operational simulation to solve the operational layer model; a genetic algorithm is used as a global search framework to select the optimal planning scheme from multiple candidate schemes, obtaining the final rural integrated energy planning scheme. This invention aims to solve the problem of excessively long planning time caused by using operational simulation to consider multiple operating modes for performance evaluation in the planning of rural integrated energy systems with a high proportion of wind and solar power.

Description

Technical Field

[0001] This invention relates to the field of new energy planning technology, and in particular to a rural integrated energy planning method and system based on operation simulation and intelligent acceleration. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Compared with traditional urban energy systems, rural integrated energy systems rely on relatively weak power distribution networks, which are prone to difficulties in absorbing renewable energy due to its volatility and intermittency. How to scientifically plan for a high proportion of renewable energy in rural areas, taking into account the characteristics of the aforementioned new energy sources and the physical properties of various types of controllable loads and adjustable resources such as energy storage, remains a key issue in the development of rural integrated energy systems.

[0004] Currently, commonly used integrated energy system planning methods often employ a two-tier planning framework. The upper-tier planning focuses on optimizing the configuration of distributed power sources or energy storage capacity, typically targeting investment costs, operating costs, or comprehensive benefit indicators, and outputting several candidate planning schemes. The lower-tier framework is responsible for evaluating the operational performance of these candidate schemes to verify their feasibility and rationality. Due to the significant volatility, intermittency, and randomness of wind and solar power, and the increasingly complex and diverse operation modes of energy systems, the accuracy of calculating operational indicators based solely on typical modes is gradually becoming insufficient to meet evaluation requirements. Methods that evaluate the operational performance of schemes based on refined operational simulations that take into account the spatiotemporal distribution characteristics of wind and solar power and loads in detail are receiving increasing attention.

[0005] Refined operation simulation refers to the process of computer simulation of the operation of the power grid and its equipment under given power system structure, operating parameters, and external environmental conditions, using models and algorithms to obtain system operating status, energy flow characteristics, and performance indicators. Existing research mainly adopts two approaches: one is based on optimized scheduling, simulating the actual system's operation process hourly or daily to accurately consider equipment physical characteristics, grid power flow constraints, power quality constraints, and network losses, thus obtaining relatively realistic operational performance evaluation results. The advantage of this method is its strong physical interpretability and high reliability, reflecting the actual system operation well. However, its disadvantage is the long computation time. For example, if the daily operation simulation takes 5 seconds, the annual operation time is approximately 30 minutes. If 100 candidate schemes need to be evaluated during the scheme selection process, the total time is approximately 51 hours, which is insufficient to meet the needs of rapid evaluation of multiple schemes. The second method directly maps performance metrics based on intelligent algorithms such as neural networks. This approach first constructs a training sample set based on a certain number of optimized scheduling results. Through iterative training via forward and backward propagation of the neural network, the network weights and biases are gradually adjusted, enabling the model to minimize the error between the predicted output and the true label. After training, for any given new scheme parameters, the neural network can directly output its corresponding performance metric, achieving a non-linear mapping from the input parameter space to the output performance metric space. The advantage of this method is its extremely high computational speed, capable of evaluating the performance of a single scheme in milliseconds, making it particularly suitable for the rapid screening and optimization of large-scale schemes. However, its disadvantage is that model performance depends on the coverage and quality of the training samples, resulting in limited generalization ability. When candidate schemes exceed the distribution range of the training samples, the prediction results may show significant deviations, making it difficult to guarantee the accuracy and reliability of the results.

[0006] In summary, relying solely on optimization simulation leads to insufficient computational efficiency, while relying solely on neural network prediction carries accuracy risks. Neither approach can simultaneously meet the dual requirements of "efficiency and accuracy" in large-scale rural integrated energy planning. A better approach would be to organically integrate these two methods, utilizing optimization simulation to ensure physical interpretability and result reliability, while leveraging neural networks for rapid mapping and scheme selection, thereby achieving a balance between efficiency and accuracy.

[0007] However, although the fusion of optimization simulation and neural networks can theoretically achieve a balance between efficiency and accuracy, practical applications still face shortcomings such as imperfect fusion mechanisms, poor sample representativeness, weak physical consistency, poor adaptability, and insufficient collaborative optimization. These problems limit its reliable promotion in large-scale, high-proportion new energy rural energy system planning. Summary of the Invention

[0008] To address the shortcomings of existing technologies, the purpose of this invention is to provide a rural integrated energy planning method and system based on operational simulation and intelligent acceleration, aiming to solve the problem of excessively long planning time caused by using operational simulation to consider multiple operating modes for performance evaluation in the planning of rural integrated energy systems with a high proportion of wind and solar power.

[0009] To achieve the above objectives, the present invention is implemented through the following technical solution: The first aspect of this invention provides a rural integrated energy planning method based on operational simulation and intelligent acceleration, comprising the following steps: The network topology and resource parameters of the rural integrated energy system are obtained, and a two-layer planning model with embedded time-series operation simulation is constructed. The two-layer planning model includes an upper planning layer model and a lower operation layer model. The planning layer model obtains the preliminary planning scheme and investment cost based on the network topology and resource parameters of the rural integrated energy system. The lower-level operation layer model uses outlier detection to identify and evaluate outlier samples in the initial planning scheme, thereby obtaining the operating cost. Specifically, during the outlier detection process, for schemes with high outlier scores, time-series operation simulations with embedded power flow constraints, voltage quality indicators, and equipment physical characteristics are performed; for conventional schemes with low outlier scores, a pre-trained neural network model is directly called to perform millisecond-level performance indicator mapping; all outlier data generated by the time-series operation simulations are transmitted back in real time and dynamically enriched into the training library as new samples. When the accumulation of new samples reaches a preset trigger threshold, incremental retraining of the neural network is automatically performed. A genetic algorithm is used as a global search framework to select the optimal planning scheme from multiple candidate schemes, and the final rural integrated energy planning scheme is obtained. The evaluation index of the candidate scheme is obtained by combining the investment cost calculated by the upper planning layer model and the operating cost calculated by the lower operation layer model after outlier detection.

[0010] A second aspect of the present invention provides a rural integrated energy planning system based on operational simulation and intelligent acceleration, comprising: The model building module is configured to obtain the network topology and resource parameters of the rural integrated energy system, construct a two-layer planning model with embedded time-series operation simulation, the two-layer planning model includes an upper planning layer model and a lower operation layer model, and the planning layer model obtains the preliminary planning scheme and investment cost based on the network topology and resource parameters of the rural integrated energy system. The model simulation and optimization module is configured as the lower-level runtime model to identify and evaluate outlier samples in the preliminary planning scheme through outlier detection, thereby obtaining the operating cost. Specifically, during the outlier detection process, for schemes with high outlier scores, time-series simulations with embedded power flow constraints, voltage quality indicators, and equipment physical characteristics are executed; for conventional schemes with low outlier scores, a pre-trained neural network model is directly called to perform millisecond-level performance indicator mapping; all outlier data generated by the time-series simulations are transmitted back in real time and dynamically enriched into the training library as new samples. When the accumulation of new samples reaches a preset trigger threshold, incremental retraining of the neural network is automatically executed. The planning and decision-making module is configured to use a genetic algorithm as a global search framework to select the optimal planning scheme from multiple candidate schemes and obtain the final rural integrated energy planning scheme. The evaluation index of the candidate scheme is obtained by combining the investment cost calculated by the upper planning layer model and the operating cost calculated after outlier detection by the lower operation layer model.

[0011] A third aspect of the present invention provides a computer-readable storage medium storing a computer program adapted for loading by a processor and executing the steps of the rural integrated energy planning method based on operational simulation and intelligent acceleration as described in the first aspect of the present invention.

[0012] A fourth aspect of the present invention provides a computer device comprising: A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the rural integrated energy planning method based on operational simulation and intelligent acceleration as described in the first aspect of the present invention.

[0013] The above one or more technical solutions have the following beneficial effects: This invention discloses a rural integrated energy planning method and system based on operational simulation and intelligent acceleration. First, a two-layer planning model with embedded time-series operational simulation is constructed, forming an upper planning layer that processes the installed capacity of various types of equipment, and a lower operational layer that processes the optimized operation of various types of equipment. Then, addressing the issues of high computational load and long processing time in operational simulation, a deep function-based outlier detection method is used to identify differences between candidate schemes and existing training samples. When the outlier degree of a scheme is small, a neural network is used for rapid performance prediction; when the outlier degree is large, a complete operational simulation is still used, while newly accumulated outlier samples are dynamically retrained to continuously optimize the accuracy and applicability of the prediction model. Finally, a non-dominated sorting genetic algorithm is used as an external search framework to find the optimal balance between investment cost and operating cost, achieving the optimal selection of planning schemes. This invention combines refined time-series operational simulation with neural network prediction to accelerate the evaluation of planning schemes; simultaneously, it introduces a dynamic training mechanism based on outlier discrimination, enabling neural network prediction to adapt to new schemes, improving accuracy and reliability, thereby significantly improving planning computation efficiency while ensuring physical feasibility and economic rationality.

[0014] This invention employs a two-layer planning architecture. The upper layer comprehensively considers multi-objective optimization of investment costs and operational indicators, and uses a genetic algorithm to generate typical candidate planning schemes. The lower layer uses a high-efficiency operational simulation method that integrates operational simulation and intelligent acceleration to balance solution accuracy and efficiency. Through interaction and collaboration between the upper and lower layers, the optimal planning scheme is achieved. This method significantly improves computational efficiency while ensuring the physical feasibility and economic rationality of the planning results. It breaks through the bottleneck of traditional planning methods that make it difficult to balance accuracy and efficiency, and provides an efficient, intelligent, and widely applicable technical approach for rural integrated energy system planning under conditions of high proportion of renewable energy.

[0015] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of the rural integrated energy planning method based on operation simulation and intelligent acceleration in Embodiment 1 of the present invention; Figure 2This is a flowchart of the two-layer planning model with embedded timing simulation in Embodiment 1 of the present invention; Figure 3 This is a flowchart of the efficient simulation computing engine for constructing the fusion intelligent algorithm in Embodiment 1 of the present invention; Figure 4 This is a flowchart of the preferred planning scheme combining genetic algorithms in Embodiment 1 of the present invention; Figure 5 This is a schematic diagram of the IEEE 33-node distribution network system architecture in Embodiment 1 of the present invention; Figure 6 This is a graph of the extended programming scan points in the evolutionary results of the non-dominated sorting genetic algorithm in Embodiment 1 of the present invention. Figure 7 This is a schematic diagram of the extended programming evolution process in the extended programming evolution result of the non-dominated sorting genetic algorithm in Embodiment 1 of the present invention. Detailed Implementation

[0018] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. 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 invention pertains.

[0019] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. 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. The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0020] Example 1: Embodiment 1 of this invention provides a rural integrated energy planning method based on operational simulation and intelligent acceleration, applicable to the planning and operation optimization of rural integrated energy systems under conditions of high-proportion renewable energy access, such as... Figure 1As shown, the main components include three parts: constructing a two-layer planning model with embedded time-series operation simulation, constructing a high-efficiency operation simulation calculation engine integrating intelligent algorithms, and optimizing the planning scheme using genetic algorithms. The first part constructs optimization models for the upper planning layer and the lower operation layer. The upper planning layer aims to minimize annual investment and operating costs, comprehensively considering key variable constraints such as energy storage capacity, wind power capacity, photovoltaic capacity, and adjustable load aggregation capacity to build an optimization model and generate multiple candidate planning schemes. The lower operation layer aims to minimize daily operating costs, constructing an optimization model based on the upper-layer planning schemes, considering equipment operation constraints, tie-line transmission power constraints, regulation command tracking constraints, power flow constraints, and grid security constraints. The second part uses a deep function-based outlier detection method to identify the differences between candidate schemes and training samples. When the outlier level is large, a complete operation simulation is used; when the outlier level is small, a neural network mapping is used to quickly provide result predictions. A dynamic retraining mechanism continuously improves the accuracy of the neural network, thereby achieving an organic integration of operation simulation and intelligent acceleration. The third part uses a non-dominated sorting genetic algorithm as an external search framework to find the optimal balance between investment cost and operating cost, thereby optimizing the planning scheme.

[0021] The specific steps are as follows: Step 1: Obtain the network topology and resource parameters of the rural integrated energy system, construct a two-layer planning model with embedded time-series operation simulation. The two-layer planning model includes an upper planning layer model and a lower operation layer model. The planning layer model obtains the preliminary planning scheme and investment cost based on the network topology and resource parameters of the rural integrated energy system.

[0022] In one specific implementation, the resource parameters include boundary parameters and cost parameters for the planning of the rural integrated energy system, and equipment parameters for the rural integrated energy system. The boundary parameters and cost parameters are used to construct the planning layer model, and the equipment parameters are used to construct the operation layer model.

[0023] The specific steps for constructing a two-layer programming model with embedded time-series simulation are as follows: First, set constraints based on various resource capacity limitations to construct a multi-objective function for the upper-layer planning model, aiming to minimize investment and operating costs. Second, set constraints based on the operational process limitations of the rural integrated energy system to construct an objective function for the lower-layer operation model, aiming to minimize the total daily operating cost. Then, construct optimization models for the upper-layer planning and lower-layer operation layers, such as... Figure 2 As shown, it includes: Step 1.1 Planning layer model construction.

[0024] Step 1.1.1 Investigate the planning boundary parameters and cost parameters of the rural integrated energy system.

[0025] This embodiment analyzes the resource types of the energy system, including but not limited to wind power, photovoltaics, energy storage, general loads such as lighting and television, electric heating loads, air conditioning loads, gas loads, water heater loads, etc., and records the various types of resources included in the list.

[0026] The upper and lower limits of wind power installed capacity for each node are determined based on local wind energy resources, the exploitable area of ​​wind farms, and grid connection conditions. The upper and lower limits of photovoltaic installed capacity at each node are determined based on local solar energy resources, the scale of available rooftop and ground resources, and grid connection and absorption capacity. Based on industry regulations regarding energy storage configuration ratios and regional grid absorption capacity, the upper and lower limits of rated energy storage power for each node are determined. Based on the characteristics of rural energy supply and consumption, the upper and lower limits of the aggregated capacity of adjustable load at each node were statistically analyzed, mainly considering the upper and lower limits of the aggregated capacity of rural domestic loads such as water heaters. , Upper and lower limits of air conditioning aggregate capacity , Upper and lower limits of electric heating aggregate capacity , Upper and lower limits of gas load (P2G) aggregation capacity , Among them, the upper and lower limits of the production load of greenhouse polymerization capacity are the same as the upper and lower limits of the electric heating polymerization capacity.

[0027] Research the various equipment models released by major manufacturers to determine the discount rate for the planning period. Unit investment cost parameters and the expected lifespan of various equipment .

[0028] Step 1.1.2 Constructing constraints based on various resource capacity limitations.

[0029] Taking into account the upper and lower limits of the exploitable capacity of various resources in the rural integrated energy system, a capacity constraint is constructed: (1).

[0030] In the formula, for Type resources in nodes The planned capacity on the above, , They are respectively Type resources in nodes Upper and lower limits of capacity.

[0031] Step 1.1.3 Construct a multi-objective function that minimizes investment and operating costs.

[0032] In the process of constructing the multi-objective function of the upper-level planning layer model, the initial investment cost of various resources is evenly distributed to each year according to the life cycle, and the annual investment cost function of each type of resource is constructed respectively: (2).

[0033] In the formula, The parameter is the unit investment cost. For the expected lifespan of various types of equipment, The discount rate during the planning period. for Annual investment cost of different types of resources For inclusion A collection of nodes for a type of resource. for Type resources in nodes The planned capacity on the above, These represent wind power installed capacity, photovoltaic installed capacity, rated energy storage power, aggregated capacity of electric heating, aggregated capacity of air conditioning, aggregated capacity of gas load, and aggregated capacity of water heater, respectively.

[0034] The total annual investment cost consists of the investment costs of various resources, and the function for the total annual investment cost is constructed as follows: (3).

[0035] In the formula, This represents the total annual investment cost.

[0036] With the goal of minimizing the total annual investment cost and total operating cost, a multi-objective function for the planning layer is formed by simultaneously solving all constraints: (4).

[0037] in, The total annual operating cost, derived from operational simulations at the lower operational levels, reflects the actual economic efficiency of the planned scheme throughout the year. The total annual investment cost, for Type resources in nodes The planned capacity on the above, , They are respectively Type resources in nodes Upper and lower capacity limits These represent wind power installed capacity, photovoltaic installed capacity, rated energy storage power, aggregated capacity of electric heating, aggregated capacity of air conditioning, aggregated capacity of gas load, and aggregated capacity of water heater, respectively.

[0038] Step 1.2 Building the runtime model.

[0039] In this embodiment, the constraints based on the operation process of the rural integrated energy system include equipment operation constraints based on equipment regulation characteristics, tie-line transmission power constraints based on power balance principles and tie-line transmission power requirements, regulation command tracking constraints based on grid regulation command requirements, power flow constraints based on grid power flow calculation, and grid security constraints.

[0040] Step 1.2.1 Investigate the network topology and equipment parameters of the rural integrated energy system.

[0041] This embodiment investigates the power grid topology of a rural integrated energy system and records relevant network parameters, mainly including the total number of nodes. Total number of branch roads Branch node association matrix Conductivity and susceptance of branch circuits and In the nodal admittance matrix OK The real and imaginary parts of a column and Apparent power limit of branch circuits Nodes during normal operation of the power system Minimum and maximum permissible voltage amplitude and Maximum transmission power of the tie line In this embodiment, , Both represent network node indices, and the row and column indices in the node admittance matrix are consistent with the network node numbers.

[0042] Investigate the interaction price between this power grid and the main grid. If the composition of the power system has not changed significantly, the interactive electricity price published by the power trading agency in the previous year can be used.

[0043] Investigate the operating parameters of various types of equipment: If the equipment is a new energy unit, the main parameter is the normalized maximum wind power output curve. Normalized photovoltaic maximum output curve .

[0044] If the equipment is an energy storage device, the main parameter is the maximum charge / discharge rate. Charge and discharge efficiency Rated energy storage capacity .

[0045] If the equipment is a general load device, then the annual general load curve for each node is formed by combining the typical daily load curve and the annual maximum daily load curve released by the Ministry of Electric Power last year. .

[0046] If the equipment is electric heating, the main parameter is heating efficiency. Thermal resistance of the hot water storage tank to air and the wall surface in contact with the ground and Supply and return water temperatures and Heat capacity and total mass of water stored User behavior parameters are user water consumption. Permissible upper and lower limits of water temperature and Meteorological parameters are typical curves of outdoor air temperature and ground temperature. and .

[0047] If the equipment is an air conditioner, the main parameter is the cooling efficiency of the air conditioning system. Room thermal resistance Room thermal resistance User behavior parameters include the user-set temperature. Permissible upper and lower limits of indoor temperature and Meteorological parameters are outdoor temperature curves. .

[0048] If the equipment is powered by gas, the main parameter is the higher calorific value of natural gas. Conversion efficiency Gas storage tank self-consumption rate Input and output conversion efficiency and Maximum gas storage capacity of the gas storage tank Maximum and minimum permissible gas storage ratio and User behavior parameters include gas consumption. .

[0049] If the device is a water heater, the main parameter is heating efficiency. The thermal resistance of the water heater Inlet and outlet water temperatures and Heat capacity With the total mass of water stored User behavior parameters refer to the quality of water used by the user. Permissible upper and lower limits of water temperature and Meteorological parameters are typical indoor air temperature curves. .

[0050] Research on the cost of penalties for curtailing renewable energy and unit adjustment costs of various load equipment .

[0051] Step 1.2.2 Construct equipment operation constraints based on equipment adjustment characteristics.

[0052] Considering the regulation characteristics and energy requirements of various equipment, construct the operating constraints for each type of equipment: If the equipment is a new energy unit, the constraints for its upper and lower output limits are as follows: (5).

[0053] In the formula, and For nodes wind power and photovoltaic Efforts are made at all times.

[0054] If the device is an energy storage device, the constraints are constructed considering its energy storage characteristics, energy upper and lower limits, and energy storage power upper and lower limits: (6).

[0055] In the formula, For nodes Upper energy storage The energy of a moment; For nodes Upper energy storage Power at any given moment (positive value for charging, negative value for discharging).

[0056] If the equipment is electric heating, the constraints should be constructed considering its heating characteristics, upper and lower limits of water temperature, and upper and lower limits of heating power: (7). In the formula, For electric heating The water temperature at any given time; For electric heating Power at any given moment.

[0057] If the equipment is an air conditioner, the constraints are constructed considering its cooling characteristics, the upper and lower limits of indoor temperature, and the upper and lower limits of cooling power: (8).

[0058] In the formula, This refers to the air conditioning temperature setting. For air conditioning Power at any moment This represents the simulation time step from time t to time t+1.

[0059] If the equipment is gas-fired, considering its gas storage characteristics, upper and lower limits of gas storage capacity, and upper and lower limits of gas power, the adjustment constraints are as follows: (9).

[0060] In the formula, This refers to the gas storage capacity; For gas Power at any given moment.

[0061] If the device is a water heater, the constraints should be constructed considering its heating characteristics, upper and lower limits of water temperature, and upper and lower limits of heating power: (10).

[0062] In the formula, For water heater The water temperature at any given time; For water heater Power at any given moment.

[0063] Step 1.2.3 Based on the power balance principle and the power transmission requirements of the tie line, construct the tie line power transmission constraint.

[0064] The total load of this system consists of energy storage capacity, general load, electric heating load, air conditioning load, gas load, and water heater load. The total output consists of wind power and photovoltaic power. The power balance constraint is as follows: (11), (12).

[0065] In the formula, For nodes Total load on; This represents the total system load. For nodes Total output of new energy sources; It contributes to the total new energy output of the system.

[0066] Considering that the main network input power should meet the tie-line transmission power requirements, the tie-line transmission power constraint is constructed as follows: (13).

[0067] In the formula, This represents the power transmitted through the tie line (positive for electricity purchased, negative for electricity sold).

[0068] Step 1.2.4 Based on the requirements of power grid regulation commands, construct regulation command tracking constraints. Command decomposition is based on the overall control commands issued by the main network. The sum of the power changes of each resource must be constrained by tracking this instruction, and is constructed as follows: (14).

[0069] In the formula, For nodes superior Baseline power of the resource type; This refers to the change in load. The change in the amount of energy contributed to new energy sources.

[0070] Step 1.2.5 Construct power flow constraints based on power grid power flow calculation.

[0071] The distribution network operates in accordance with power flow constraints and is constructed as follows: (15).

[0072] In the formula, and They are nodes The active power injection and reactive power injection; For nodes exist Voltage amplitude at any given moment; branch road The voltage phase angle difference at both ends.

[0073] The power flow constraints for distribution network branches are as follows: (16).

[0074] In the formula, For connecting nodes and nodes Branch number; and Branch roads The active and reactive power. Ignoring the branch-to-ground admittance, substituting equation (16) into equation (15), the constraints can be obtained according to the definition of the nodal admittance matrix: (17).

[0075] The node injected power consists of tie-line transmission power, total renewable energy output, and total load, with the following constraints: (18).

[0076] In the formula, , These are the reactive power transmitted via tie lines and the reactive power of general loads, respectively.

[0077] Step 1.2.6 Construct power grid security constraints.

[0078] Power grid safety constraints include preventing heavy overloads on all lines and ensuring that the voltage amplitude at each network node remains within specified limits. These constraints are constructed as follows: (19).

[0079] Step 1.2.7 Construction of the objective function The total daily operating cost includes the cost of purchasing and selling electricity, regulation costs, and the cost of curtailment penalties. The daily operating cost function is constructed as follows: (20).

[0080] In the formula, The power of the resource of type i on node i. This refers to the interaction price between the power grid and the main grid. For nodes superior Baseline power of the type of resource, To transmit power to the tie line, This represents the simulation time step from time t to time t+1. For the unit adjustment cost of various load equipment, This is the normalized maximum output curve. This represents the total number of time periods per day. These are the costs of purchasing and selling electricity, regulation costs, and penalties for curtailing electricity.

[0081] Using minimizing the total daily operating cost as the objective function, a system of simultaneous constraints is established to form the operational-level optimization model: (twenty one).

[0082] Step 2: The lower-level operation layer model uses outlier detection to identify and evaluate outliers in the preliminary planning scheme, obtains the operating cost, and uses neural networks to accelerate or refine the time-series simulation to solve the operation layer model. Specifically, during the outlier detection process, for schemes with high outlier scores, a time-series simulation with embedded power flow constraints, voltage quality indicators, and equipment physical characteristics is performed; for conventional schemes with low outlier scores, a pre-trained neural network model is directly called to perform millisecond-level performance indicator mapping.

[0083] In one specific implementation, this embodiment addresses the problem of large computational load and long processing time in refined operation simulation by combining outlier detection oriented towards energy system characteristics with neural network mapping oriented towards the planning search process, constructing a hybrid evaluation system with continuous learning capabilities. In integrated energy planning, the "outlier" nature of a scheme essentially reflects the degree of structural mismatch between its resource allocation and local wind and solar resources and load demand. This embodiment uses an autoencoder to learn the structural feature distribution of "reasonable" schemes in a historical database, and its outlier score directly reflects the degree of deviation between the new scheme and the historical set of "reasonable" schemes. When the outlier score is small, it indicates that its resource allocation structure is highly similar to known, simulation-verified feasible schemes, and its operating cost is likely to follow known patterns, making it suitable for rapid prediction using neural networks. When the outlier score is large, it indicates that the scheme has an unprecedented configuration structure (e.g., extremely high photovoltaic penetration rate coupled with extremely small energy storage capacity). Such structurally innovative schemes may have operational risks and must be evaluated safely and reliably using precise physical simulation. Through a dynamic retraining mechanism, the model accuracy is continuously improved, ultimately forming a closed-loop acceleration system of "outlier detection—neural network prediction / refined operation simulation—retraining". The process of building a high-efficiency simulation computing engine that integrates intelligent algorithms, such as... Figure 3 As shown, it includes: Step 2.1: Set an outlier score threshold based on the principle of adaptive statistical distribution, introduce a depth function to detect outliers in the planning scheme, obtain outlier scores, and determine whether a sample is an outlier based on the comparison results of the outlier score threshold.

[0084] In one specific implementation, an outlier score threshold is set based on the principle of adaptive statistical distribution. This means that the depth value is automatically determined based on the distribution characteristics of the training samples in the depth function space. Specifically, after the initial training of the neural network is completed, the depth value distribution of all training samples is calculated, and a threshold is set according to its confidence interval or quantile criterion. Typically, the 5th percentile of the sample depth value or a range based on the mean ± 2 standard deviations is taken as the distribution boundary. When the depth value of a candidate solution is lower than this threshold, it is determined to be an outlier sample.

[0085] Obtain the planning layer scheme and construct the input vector: (twenty two).

[0086] An autoencoder is used as the depth function, based on the encoder. and decoder Construct the input vector The low-dimensional dense representation of the data, and the data reconstruction as lossless as possible, involves the following encoding and decoding process: (twenty three).

[0087] In the formula, A low-dimensional encoding vector in the latent space; For reconstructing vectors; For activation functions; , , , For historical database The parameters of the autoencoder obtained during training.

[0088] Calculate the input vector and reconstructed vector The mean squared error is used as the outlier score: (twenty four).

[0089] In the formula, Scoring for outliers; is the dimension of the input vector.

[0090] This embodiment employs a depth function for outlier detection, aiming to determine whether candidate planning schemes fall within the distribution range of the neural network training samples. Compared to traditional outlier detection methods based on distance or density, the depth function reflects representativeness by measuring the "centrality" of samples within the overall distribution, better adapting to the high-dimensional, nonlinear, and multi-feature coupled feature space structure of rural integrated energy planning schemes. This method requires no additional model training or parameter tuning, has low computational cost, and is suitable for real-time application in the rapid evaluation of multiple schemes using genetic algorithms. Furthermore, the depth function can be dynamically updated with the accumulation of simulated samples, facilitating an adaptive "outlier detection-retraining" loop, which aligns perfectly with the intelligent acceleration mechanism of this invention.

[0091] Step 2.2: Select a solution scheme based on the outlier sample judgment results.

[0092] In one specific implementation, for non-outlier samples, a neural network model pre-trained based on a historical database is directly invoked. By establishing a nonlinear mapping relationship between the planning configuration scheme and the total annual operating cost, the evaluation index can be quickly predicted. For outlier samples, the objective function of the lower-level operating layer model is programmed into a function model, and a commercial solver is used to solve it to obtain the total daily system operating cost. The solution is continuously rolled over day by day, traversing all days of a typical year, and the daily operating costs for the whole year are accumulated to obtain the total annual operating cost.

[0093] Specifically, if the planning scheme is a non-outlier sample, that is Then directly use the historical database. The neural network mapping obtained from the training results from the planning scheme to the total annual operating cost: (25). In the formula, For neural network parameters, The output result is the input vector. Substitute these values ​​into the equation to obtain the total annual operating cost, and then conclude the solution for the operating layer.

[0094] If the planning scheme is an outlier, that is... Then, the runtime optimization model (21) is programmed into a function model, and the total daily system operating cost is obtained by using mature commercial solvers (such as Knitro and BARON). The solution is continuously rolled over day by day, traversing all the days of a typical year. The total annual operating cost is obtained by summing up the daily operating costs throughout the year: .

[0095] At the same time, new data pairs are formed. This data is added to a buffer database composed of such outliers. .

[0096] Step 2.3: Update the parameters of the deep function and the neural network model through retraining.

[0097] In one specific implementation, outlier data generated from all time-series simulations are fed back in real time and dynamically enriched into the training library as new samples. When the accumulation of new samples reaches a preset trigger threshold, incremental retraining of the neural network is automatically executed. In the early stages of optimization, the genetic algorithm population may focus on searching a local region. As it evolves (crossover, mutation), the algorithm explores new solution spaces that may offer better performance. The initial intelligent acceleration model is trained on data from the old search region and cannot accurately evaluate solutions in these newly explored regions. The core purpose of retraining is to keep the capabilities of the intelligent acceleration model synchronized with the search frontier of the genetic algorithm, ensuring efficient and accurate evaluation throughout the entire optimization process, whether for solutions initially searched in a concentrated manner or for innovative solutions discovered later.

[0098] The specific steps are as follows: Step 2.3.1: Set the threshold for the number of retraining samples. ,like If so, the runtime solution will end.

[0099] Step 2.3.2: If the number of "new exploration area" schemes accumulated through accurate simulation... Then, a retraining database containing historical and buffer databases is constructed. Essentially, it expands the experience base of intelligent algorithms to encompass the latest planning and search solutions. Simultaneously, it clears the buffer database.

[0100] Step 2.3.3: Retrain the autoencoder with the expanded database, update the system's cognitive boundaries of "reasonable scheme structure", and incorporate newly discovered feasible configurations into the regular range.

[0101] use All scheme vectors The autoencoder parameters are updated by minimizing the overall reconstruction error. , , , : (26).

[0102] Step 2.3.4: Retrain the neural network mapper with the expanded database so that the neural network can learn a new "capacity-cost" mapping relationship in the new search area, thereby providing more accurate and faster predictions in the new search area.

[0103] use All data pairs Supervised learning is performed on the neural network of equation (25), and the training objective is to find the optimal parameters. The loss function that minimizes the difference between the predicted and actual simulated values ​​is expressed using the mean squared error loss as follows: (28).

[0104] In the formula, For data pair sequence numbers; This represents the number of training samples; From the retraining database ; The L2 regularization coefficient is used to prevent overfitting. Parameters are iteratively updated using gradient descent methods such as backpropagation and adaptive moment estimator optimizer. : (29).

[0105] In the formula, The learning rate; and The bias correction term is used for estimating the first and second moments of the gradient; Small constants used to maintain numerical stability.

[0106] Step 2.3.5: Return the updated autoencoder parameters to equation (23); return the updated neural network parameters to equation (25); return to step 2.1 to continue running the simulation.

[0107] Step 3: Using a genetic algorithm as the global search framework, the optimal planning scheme is selected from multiple candidate schemes to obtain the final rural integrated energy planning scheme.

[0108] In one specific implementation, the evaluation index of the candidate scheme is obtained by combining the investment cost calculated by the upper-level planning layer model and the operating cost calculated by the lower-level operation layer model after outlier detection. This embodiment uses a non-dominated sorting genetic algorithm as the external search framework to find the optimal balance between investment cost and operating cost, thereby optimizing the planning scheme. First, the algorithm randomly generates a set of initial planning schemes; second, it performs a two-level evaluation on each scheme in the population to obtain two target values ​​for each scheme; third, it generates new offspring schemes through selection, crossover, and mutation operations, and merges the parent and offspring generations; then, it selects elite individuals to form the next generation population using two criteria: rank and crowding; finally, it iterates the above process until a preset maximum number of generations is reached, and the final output set of rank-schemes is the optimal planning scheme set. The optimization process of the genetic algorithm for planning schemes is as follows: Figure 4 As shown, it includes: Step 3.1: Set the algorithm parameters and initialize the population.

[0109] Step 3.1.1: Set the running parameters of the genetic algorithm, including the population size. Crossover probability Probability of mutation Maximum number of iterations .

[0110] Step 3.1.2: Within the constraints of equation (1), randomly generate... The initial planning scheme constitutes the parental population. The decision variables to be optimized for each population are: ; Step 3.1.3: Encode each decision variable using fixed-length binary encoding, specifying the bit length for each variable. This linearly maps the range of values ​​of a variable to the range of integers that the binary bit string can represent. .

[0111] Step 3.2: Evaluate the operational plan and construct the corresponding target vector.

[0112] Step 3.2.1: For each individual in the population First, substitute the values ​​into equations (2) and (3) to calculate the total annual investment cost. .

[0113] Step 3.2.2: Then, using the efficient simulation calculation based on the integrated intelligent algorithm from Step 2, the annual operating cost under this scheme is obtained. .

[0114] Step 3.2.3: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require the full context.] Return to planning layer and Construct the target vector .

[0115] Step 3.3: Construct a child population based on the target vector, and merge the child population with the parent population to obtain a merged population.

[0116] Step 3.3.1: Based on the target vector, use the tournament selection method from... Select individuals, and cross over the selected parents with a given probability. Randomly select two intersection points to exchange the middle segment of the decision variable encoding, and generate offspring.

[0117] Step 3.3.2: Then, for each newly generated offspring, iterate through each bit of the encoding and, with the given mutation probability... Perform a bit flip.

[0118] Step 3.3.3: If a variable violates the constraint of equation (1) after decoding, then project the variable back to the nearest feasible value, i.e. the constraint boundary.

[0119] Step 3.3.4: Transfer the parent generation With offspring Merging to form a merged population .

[0120] Step 3.4: Perform non-dominated sorting and elite retention operations on the merged population.

[0121] Step 3.4.1: For Perform a non-dominated ranking on all individuals and calculate the dominant individuals. The number of individuals and the record of individuals A dominant set of individuals, thereby Divided into a series of Pareto levels , This is the Pareto front currently found.

[0122] Step 3.4.2: For each level For individuals within the same area, calculate their crowding distance: (30) In the formula, Distance based on congestion level and In order to achieve the goal After sorting The adjacent individuals.

[0123] Step 3.4.3: Among different levels, prioritize individuals with higher Pareto rank; within the same level, prioritize individuals with greater crowding distance; place individuals in the appropriate positions in sequence. until the number of individuals equals .

[0124] Step 3.5: Obtain the final planning scheme through iteration.

[0125] make ,like If so, return to step 3.3; if Then output the final population. The Pareto level is The solution set, i.e., the set of all programming schemes on the Pareto front. .

[0126] To verify the effectiveness of the rural integrated energy planning method based on operation simulation and intelligent acceleration described in this embodiment, taking the IEEE 33-node distribution network as an example, we conduct integrated energy planning for distributed photovoltaic and electric heating in this distribution network based on the method of this embodiment. The specific process is as follows: Firstly, regarding Figure 5 The IEEE 33-node distribution network shown in the figure underwent detailed resource boundary surveys and parameter initialization. The numbers in the figure represent the 33 nodes. The system base voltage was set at 12.66kV, and based on local meteorological data and rural energy supply characteristics, the total active power load of the entire network was determined to be 3.715MW. Based on this, the planning layer model clarified the access restrictions for 8 distributed photovoltaic (PV) units and 5 electric heating units. The PV installation locations were selected based on the abundance of solar resources at each node, and the upper and lower limits of the installed capacity were set based on the available roof area at each node, with the maximum installed capacity for a single node set at 30MW. Electric heating units were located at nodes with higher loads, and their maximum conversion capacity was strictly constrained to 50% of the node's original load. Simultaneously, the survey obtained key economic parameters, including a discount rate of 5%, a PV unit investment cost of 3,500,000 yuan / MW, and an electric heating unit investment cost of 150,000 yuan / MW. These initial investments were then evenly distributed over each year of the equipment's lifespan using a recovery factor.

[0127] In constructing the operational model, the physical operating parameters of various types of equipment were investigated. Parameters for electric heating equipment included a rated power of 0.5MW, a heating efficiency of 0.946, an air thermal resistance of 1820℃ / MW, and a user-allowed water temperature range of 40-90℃. Power prediction for photovoltaic equipment was based on normalized output curves. The curtailment penalty cost was set at 100 yuan / MWh, and the regulation cost was set at 1000 yuan / MWh.

[0128] The initial population size was set at 30 individuals, and the decision variables for each individual covered the access capacity of 8 photovoltaic units and 5 electric heating units. Fixed-length binary encoding was used, assigning an 8-bit length to each capacity variable, linearly mapping its value range to [0, 2]. 8 The integer range is defined as [-1]. The crossover probability is set to 0.9, the mutation probability to 0.3, and the maximum number of iterations to 15 generations.

[0129] For each planning individual in the population, its annual equivalent investment cost is first calculated based on the surveyed equipment unit price. Then, the scheme configuration vector is input into the computational engine fused with intelligent acceleration: the system uses an autoencoder to calculate the outlier score of the scheme; if the score is below a threshold, a pre-trained neural network mapper directly predicts the total annual operating cost within milliseconds; if the score is high, a mature commercial solver is invoked to perform refined time-series simulations, obtaining accurate daily operating costs and accumulating them. Through this "prediction-simulation" coupling mechanism, a multi-objective vector containing both investment and operating costs is constructed.

[0130] A tournament selection method is used to select individuals from the parent generation. Two-point crossover and positional inversion mutations are performed on the selected parents to generate new offspring schemes. If the decoded variables violate the constraints of the photovoltaic installation limit or the 50% load replacement rate for electric heating, they are projected back to the nearest constraint boundary. The parents and newly generated offspring are merged, and a non-dominated sorting algorithm is used to divide the merged population into different Pareto tiers. Among different tiers, those with higher tiers are prioritized; within the same tier, selection is based on crowding distance, retaining the top 10 elite individuals directly into the next generation. During this process, if the number of outlier samples accumulated through refined simulation reaches a threshold, a dynamic retraining mechanism is initiated to update the autoencoder and neural network parameters, ensuring that the accelerated model can capture the potential optimal features discovered by the genetic algorithm in the new search space in real time.

[0131] Finally, within the external search framework of the aforementioned non-dominated sorting genetic algorithm, after 10 generations of evolutionary iterations, the algorithm successfully searched for the Pareto optimal front between investment cost and operating cost. The iterative process is as follows: Figure 6 and Figure 7As shown in Table 1, the investment and operating costs of the Pareto optimal frontier scheme are as follows. The output first-level scheme set, under the premise of satisfying the safety constraints of branch power limits and node voltage, provides the optimal access locations and rated capacity configurations for photovoltaic (PV) and electric heating. The results show that the optimal scheme involves installing PV capacities of 2.3MW, 3.6MW, 1MW, 4.1MW, 4.3MW, 7.3MW, and 1.6MW at nodes 4, 11, 14, 22, 23, 30, and 31 respectively, and installing electric heating capacities of 0.01MW, 0.05MW, 0.01MW, 0.01MW, and 0.03MW at nodes 7, 8, 24, 25, and 32 respectively. The investment cost is 6,861,558.818 yuan, and the operating cost is 1,793,152.305 yuan. The scheme effectively balances equipment investment costs with the actual economic efficiency of annual operation, providing reliable technical support for the scientific allocation of high-proportion new energy sources in rural areas.

[0132] Table 1. Investment and operating costs of the Pareto optimal frontier solution PV1(MW) PV2(MW) PV3(MW) PV4(MW) PV5(MW) PV6(MW) PV7(MW) PV8(MW) EH1(MW) EH2(MW) EH3(MW) EH4(MW) EH5(MW) Operating costs (RMB) Investment cost (RMB) 0 0 0 0 0 0 0 0 0 0 0 0 0 14564066.47 0 17.5 0 2.8 2.5 2.4 5.7 1.4 9.2 0.36 0.18 0.01 0.01 0.01 673906.9 11774131 17.5 0 2.8 2.5 2.4 5.7 1.4 9.2 0.36 0.18 0.01 0.01 0.01 673906.9 11774131 0.1 0.1 2.8 2.5 2.4 0.1 0.1 0.1 0.01 0.18 0.01 0.07 0.01 9460884 2329706 0.1 0.1 0.1 0.1 0.1 5.7 1.4 9.2 0.36 0.32 0.01 0.01 0.05 4562574 4776482 0.1 4.3 0.1 0.1 0.1 0.1 0.1 0.3 0.01 0.01 0.01 0.17 0.01 11235248 1478005 0.1 0 0.1 2.4 2.7 1 0.1 0.1 0.34 0.36 0.07 0.01 0.01 9834247 1857753 12.6 0.1 3.4 0 2.2 8.9 0.1 5.9 0.36 0.18 0.01 0.05 0 688676.5 9422102 2.3 3.6 1 4.1 4.3 0 7.3 1.6 0.01 0.01 0.03 0.05 0.01 1793152 6861559 0.1 0 11.8 0.1 0.1 0.4 0.1 0.1 0.01 0.01 0.01 0.1 0.01 6919250 3602499 0.1 1.7 0.9 2 8.8 5.6 4.6 3.5 0.01 0.04 0 0.01 0 1341833 7710929 17.5 0 2.3 2.6 0.1 5.7 0.1 9.2 0.29 0.23 0.01 0.01 0.01 685349.7 10639954 0 0.1 0.1 0.4 12.3 0.1 1.6 0.1 0.01 0.01 0.09 0.02 0.05 6300695 4170170 0.7 0.1 0.1 0.1 0.2 2.4 0.1 0.1 0.01 0.36 0.05 0.03 0 11859335 1085841 0.1 0.1 5.6 0.1 0 0.1 9.2 0.1 0.22 0.03 0.08 0.01 0.01 5408345 4343541 4 4.4 0.3 3.9 7 0.4 10.4 1.7 0.07 0.01 0.07 0.02 0.01 928130.4 9102151 17.5 0 2.8 2.5 2.4 5.7 1.4 9.2 0.36 0.18 0.01 0.01 0.01 673906.9 11774131 17.5 0 2.3 2.6 0.1 5.7 0.1 9.2 0.29 0.23 0.01 0.01 0.01 685349.7 10639954 3.9 10.8 0.1 8 0.1 0.1 0.1 0.3 0.01 0.01 0.01 0.14 0.06 2760228 6637132 9 10.8 0.1 0.1 0.1 0.1 0 0.3 0.01 0.01 0.05 0.14 0 3128945 5814747 0.1 0.1 0.1 2.5 0.2 0.1 0.1 0.1 0.01 0.18 0.01 0.01 0.01 12363395 939649.4 4.5 10.8 0.1 0.1 0.1 1.9 0 0.1 0.01 0.01 0.05 0.14 0.02 4224475 4993138 0.1 0.1 4.6 0 0.1 0 5 0.9 0.01 0.11 0 0.15 0 8024219 3066475 0 1 0 3.7 0 4.8 0.1 0.1 0.01 0 0 0.03 0.04 8767325 2750992 0.1 1.4 0.1 0.1 10.5 7.1 4.5 0.1 0.01 0.05 0 0.01 0.06 1921292 6776913 0 0.1 0.1 0.6 0.1 0.1 0 0 0 0.01 0.05 0.16 0 13721176 287720.8 2.5 4.4 0.3 3.7 5.6 0.4 10.2 1.7 0.07 0.01 0.07 0.05 0.01 1051245 8167358 0.1 0 0.1 2.5 2.4 0.1 0.1 0.1 0.36 0.36 0.05 0.01 0.01 10458164 1545961 0 0.1 0.4 0.3 1 0.4 0 0 0 0.01 0.06 0.17 0 12994421 628246 3.9 10.8 0.1 4.3 0.1 0.1 0.1 0.3 0.01 0.01 0.03 0.15 0.06 3435987 5588960 This invention constructs a two-layer optimization planning model. The upper-layer model, serving as the planning layer, is primarily responsible for the initial formulation and global optimization of resource allocation schemes, generating multi-dimensional decision schemes encompassing wind power, photovoltaics, energy storage, and load aggregation capacity through a genetic algorithm. The lower-layer model, serving as the operation evaluation layer, focuses on evaluating the schemes proposed by the upper layer, verifying their economic efficiency and reliability under actual grid constraints.

[0133] The core innovation of this invention is mainly applied to the lower-level operation evaluation stage, which integrates refined time-series operation simulation and neural network acceleration, aiming to overcome the insufficient computational efficiency caused by relying solely on time-series simulation and the accuracy risk of relying solely on neural network prediction.

[0134] This invention constructs a self-evolutionary hybrid evaluation system based on outlier detection path diversion and incremental retraining in the lower-level model. In the process of multi-objective optimization using genetic algorithms, this invention does not employ traditional single-time-consuming time-series simulations or easily distorted pure neural network models. Instead, it introduces a deep function in the lower-level evaluation stage to perform real-time spatial location representation and outlier degree measurement of the evaluated schemes. By quantifying the distribution characteristics of the schemes in the currently known sample space, the system can autonomously identify whether the planning scheme is in the blind spot of the model's cognition: for schemes with high outlier scores, it performs refined time-series simulations embedding complex power grid flow constraints, voltage quality indicators, and equipment physical characteristics. Daily continuous rolling simulations ensure that rigorous mechanistic constraints and evaluation accuracy are maintained when exploring unknown planning areas; for conventional schemes with low outlier scores, it directly calls the pre-trained neural network model for millisecond-level performance indicator mapping, thereby significantly reducing the computational overhead of the genetic algorithm in large-scale iteration processes. In addition, all outlier data generated by the time-series simulation are transmitted back in real time and dynamically enriched into the training library as new samples. When the accumulation of new samples reaches the preset trigger threshold, the neural network is automatically retrained incrementally, driving the effective cognitive boundary of the neural network acceleration engine to dynamically expand as the upper-layer search range expands.

[0135] Example 2: Embodiment 2 of the present invention provides a rural integrated energy planning system based on operational simulation and intelligent acceleration, comprising: The model building module is configured to obtain the network topology and resource parameters of the rural integrated energy system, construct a two-layer planning model with embedded time-series operation simulation, the two-layer planning model includes an upper planning layer model and a lower operation layer model, and the planning layer model obtains the preliminary planning scheme and investment cost based on the network topology and resource parameters of the rural integrated energy system. The model simulation and optimization module is configured as the lower-level runtime model to identify and evaluate outlier samples in the preliminary planning scheme through outlier detection, thereby obtaining the operating cost. Specifically, during the outlier detection process, for schemes with high outlier scores, time-series simulations with embedded power flow constraints, voltage quality indicators, and equipment physical characteristics are executed; for conventional schemes with low outlier scores, a pre-trained neural network model is directly called to perform millisecond-level performance indicator mapping; all outlier data generated by the time-series simulations are transmitted back in real time and dynamically enriched into the training library as new samples. When the accumulation of new samples reaches a preset trigger threshold, incremental retraining of the neural network is automatically executed. The planning and decision-making module is configured to use a genetic algorithm as a global search framework to select the optimal planning scheme from multiple candidate schemes and obtain the final rural integrated energy planning scheme. The evaluation index of the candidate scheme is obtained by combining the investment cost calculated by the upper planning layer model and the operating cost calculated after outlier detection by the lower operation layer model.

[0136] Example 3: Embodiment 3 of the present invention provides a computer-readable storage medium storing a computer program adapted for loading by a processor and executing the steps in the rural integrated energy planning method based on operational simulation and intelligent acceleration as described in Embodiment 1 of the present invention.

[0137] Example 4: Embodiment 4 of the present invention provides a computer device, the device comprising: A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the steps in the rural integrated energy planning method based on operational simulation and intelligent acceleration as described in Embodiment 1 of the present invention.

[0138] The steps and methods involved in Examples 2, 3 and 4 above correspond to those in Example 1. For specific implementation details, please refer to the relevant description section of Example 1.

[0139] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0140] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The computer-readable storage medium can be any available medium that a computer can access or a data processing device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium, an optical medium, or a semiconductor medium, etc.

[0141] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A rural integrated energy planning method based on operational simulation and intelligent acceleration, characterized in that, Includes the following steps: The network topology and resource parameters of the rural integrated energy system are obtained, and a two-layer planning model with embedded time-series operation simulation is constructed. The two-layer planning model includes an upper planning layer model and a lower operation layer model. The planning layer model obtains the preliminary planning scheme and investment cost based on the network topology and resource parameters of the rural integrated energy system. The lower-level operation layer model uses outlier detection to identify and evaluate outlier samples in the initial planning scheme, thereby obtaining the operating cost. Specifically, during the outlier detection process, for schemes with high outlier scores, time-series operation simulations with embedded power flow constraints, voltage quality indicators, and equipment physical characteristics are performed; for conventional schemes with low outlier scores, a pre-trained neural network model is directly called to perform millisecond-level performance indicator mapping; all outlier data generated by the time-series operation simulations are transmitted back in real time and dynamically enriched into the training library as new samples. When the accumulation of new samples reaches a preset trigger threshold, incremental retraining of the neural network is automatically performed. A genetic algorithm is used as a global search framework to select the optimal planning scheme from multiple candidate schemes, and the final rural integrated energy planning scheme is obtained. The evaluation index of the candidate scheme is obtained by combining the investment cost calculated by the upper planning layer model and the operating cost calculated by the lower operation layer model after outlier detection.

2. The rural integrated energy planning method based on operational simulation and intelligent acceleration as described in claim 1, characterized in that, Resource parameters include boundary parameters and cost parameters for rural integrated energy system planning, and equipment parameters for rural integrated energy system. Boundary parameters and cost parameters are used to construct the planning layer model, while equipment parameters are used to construct the operation layer model.

3. The rural integrated energy planning method based on operational simulation and intelligent acceleration as described in claim 1, characterized in that, The specific steps for constructing a two-level programming model with embedded time-series simulation are as follows: Set constraints based on various resource capacity limitations, and construct a multi-objective function for the upper-level planning layer model with the goal of minimizing investment and operating costs; Constraints based on the operational constraints of the rural integrated energy system are set, with the goal of minimizing the total daily operating cost. The objective function of the lower-level operation layer model is constructed. The constraints based on the operational constraints of the rural integrated energy system include equipment operation constraints based on equipment regulation characteristics, tie-line transmission power constraints based on power balance principles and tie-line transmission power requirements, regulation command tracking constraints based on grid regulation command requirements, power flow constraints based on grid power flow calculation, and grid security constraints.

4. The rural integrated energy planning method based on operational simulation and intelligent acceleration as described in claim 3, characterized in that, With the goal of minimizing the total annual investment cost and total operating cost, a multi-objective function for the planning layer is formed by simultaneously solving all constraints: ， in, The total annual operating cost is derived from operational simulations at the lower-level operating layer. The total annual investment cost, for Type resources in nodes The planned capacity on the above, , They are respectively Type resources in nodes Upper and lower capacity limits These represent wind power installed capacity, photovoltaic installed capacity, rated energy storage power, aggregated capacity of electric heating, aggregated capacity of air conditioning, aggregated capacity of gas load, and aggregated capacity of water heater, respectively.

5. The rural integrated energy planning method based on operational simulation and intelligent acceleration as described in claim 4, characterized in that, The total daily operating cost includes the cost of purchasing and selling electricity, regulation costs, and the cost of curtailment penalties. The daily operating cost function is constructed as follows: ， in, The power of the resource of type i on node i. This refers to the interaction price between the power grid and the main grid. For nodes superior Baseline power of the type of resource, To transmit power to the tie line, This represents the simulation time step from time t to time t+1. For the unit adjustment cost of various load equipment, This is the normalized maximum output curve. This represents the total number of time periods per day. These are the costs of purchasing and selling electricity, regulation costs, and penalties for curtailing electricity.

6. The rural integrated energy planning method based on operational simulation and intelligent acceleration as described in claim 1, characterized in that, The specific steps of the lower-level operational model to identify outliers in the initial planning scheme through outlier detection are as follows: Based on the principle of adaptive statistical distribution, an outlier score threshold is set, a depth function is introduced to detect outliers in the planning scheme, an outlier score is obtained, and the comparison result of the outlier score threshold is used to determine whether it is an outlier sample. The solution scheme is selected based on the outlier sample judgment results. For non-outlier samples, a neural network model pre-trained based on historical database is directly called. By establishing a nonlinear mapping relationship between the planning configuration scheme and the annual total operating cost, the evaluation index can be quickly estimated. For outlier samples, the objective function of the lower-level operating layer model is programmed into a function model and a commercial solver is used to solve for the daily total system operating cost. The solution is continuously rolled over day by day, traversing all days of a typical year, and the daily operating costs for the whole year are accumulated to obtain the annual total operating cost.

7. The rural integrated energy planning method based on operational simulation and intelligent acceleration as described in claim 1, characterized in that, The specific steps for using a genetic algorithm as a global search framework to select the optimal planning scheme from multiple candidate schemes and obtain the final rural integrated energy planning scheme are as follows: Set the algorithm parameters and initialize the population; Evaluate the operational plan and construct the corresponding target vector; Construct a child population based on the target vector, and merge the child population with the parent population to obtain a merged population. Perform non-dominated sorting and elite retention operations on the merged population; The final planning scheme is obtained through iteration.

8. A rural integrated energy planning system based on operational simulation and intelligent acceleration, characterized in that, include: The model building module is configured to obtain the network topology and resource parameters of the rural integrated energy system, construct a two-layer planning model with embedded time-series operation simulation, the two-layer planning model includes an upper planning layer model and a lower operation layer model, and the planning layer model obtains the preliminary planning scheme and investment cost based on the network topology and resource parameters of the rural integrated energy system. The model simulation and optimization module is configured as the lower-level runtime model to identify and evaluate outlier samples in the preliminary planning scheme through outlier detection, thereby obtaining the operating cost. Specifically, during the outlier detection process, for schemes with high outlier scores, time-series simulations with embedded power flow constraints, voltage quality indicators, and equipment physical characteristics are executed; for conventional schemes with low outlier scores, a pre-trained neural network model is directly called to perform millisecond-level performance indicator mapping; all outlier data generated by the time-series simulations are transmitted back in real time and dynamically enriched into the training library as new samples. When the accumulation of new samples reaches a preset trigger threshold, incremental retraining of the neural network is automatically executed. The planning and decision-making module is configured to use a genetic algorithm as a global search framework to select the optimal planning scheme from multiple candidate schemes and obtain the final rural integrated energy planning scheme. The evaluation index of the candidate scheme is obtained by combining the investment cost calculated by the upper planning layer model and the operating cost calculated after outlier detection by the lower operation layer model.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted to be loaded by a processor and executed as described in any one of claims 1-7: a rural integrated energy planning method based on operational simulation and intelligent acceleration.

10. A computer device, characterized in that, include: A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the rural integrated energy planning method based on operational simulation and intelligent acceleration as described in any one of claims 1-7.

Images