Multi-source load intelligent optimization control method for power distribution network

Abstract

The invention discloses a power distribution network multi-source load intelligent optimization control method, and relates to the technical field of power system power distribution networks, and the method comprises the following components: S1, building a digital twin model, S2, optimization control strategy preview, S3, strategy verification and evaluation, S4, strategy adjustment and optimization, and S5, practical application and feedback. According to the method, intelligent optimization control over the multi-source loads of the power distribution network is achieved by building the digital twin model and optimizing control strategy rehearsal, the topological structure, the electrical parameters and the real-time operation state of the physical power distribution network can be accurately mapped, the optimization control strategy is formulated according to the load priority and the collaborative weight of distributed energy, and the optimization control over the multi-source loads of the power distribution network is achieved. The operation efficiency and stability of the power distribution network are improved, the reliability and safety of load supply are ensured, and potential problems can be found and solved in time through simulation operation and strategy verification of the digital twin model, so that risks and faults possibly occurring in the operation process of the power distribution network are effectively avoided.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of power distribution networks of power systems, in particular to a multi-source load intelligent optimization control method of a power distribution network. BACKGROUND

[0002] With the continuous development of power systems and the improvement of intelligent level, the power distribution network, as an important link directly connected with users in the power system, has important significance for guaranteeing the reliability and safety of power supply in terms of operation efficiency and stability. However, with the wide access of distributed power sources, the diversification of load types and the rapid growth of emerging loads such as electric vehicles, the power distribution network is facing unprecedented complexity and uncertainty.

[0003] The traditional technology has the following deficiencies: firstly, the traditional digital twin model often ignores the uncertainty and noise interference of the operation data of the power distribution network in the construction process, resulting in insufficient model accuracy and difficulty in accurately reflecting the real-time operation state of the physical power distribution network; secondly, the traditional method lacks flexibility in formulating the optimization control strategy and fails to fully consider the load priority and the synergistic effect of distributed energy, so that the control strategy has poor effect in actual application; in addition, the traditional simulation and verification process is time-consuming and difficult to find and solve potential problems in time, which brings hidden dangers to the safe operation of the power distribution network.

[0004] In summary, the traditional multi-source load optimization control method of the power distribution network is not competent when facing the complex and variable operation environment of the power distribution network, therefore, it is particularly important to develop a multi-source load intelligent optimization control method of the power distribution network. SUMMARY

[0005] The purpose of the present application is to make up for the deficiencies of the prior art, and to provide a multi-source load intelligent optimization control method of a power distribution network, which can realize intelligent optimization control of the multi-source load of the power distribution network by constructing a more accurate digital twin model and optimization control strategy, improve the operation efficiency and stability of the power distribution network, and ensure the reliability and safety of load supply.

[0006] To solve the above technical problems, the application provides the following technical scheme: a multi-source load intelligent optimization control method of a power distribution network, and the specific steps of the method are as follows:

[0007] S1, constructing a digital twin model

[0008] Deploy sensors in the physical power distribution network to collect node voltage, line current, power of various types of loads and power of distributed power sources, and use a deep learning algorithm to construct a multi-layer neural network digital twin model, wherein the input layer is the collected data, the hidden layer is connected through an activation function Feature extraction and nonlinear transformation are performed, where α and β are parameters determined by genetic algorithm optimization according to historical data and operating characteristics of the power distribution network, and the output layer is the predicted operating state of each node by minimizing the mean square error of the predicted value and the actual value Train the model, where N is the total number of nodes, S n is the actual operating state of node n, and through continuous iteration training, the digital twin model can accurately map the topology, electrical parameters and real-time operating state of the physical power distribution network

[0009] S2, optimization control strategy rehearsal

[0010] Based on the load priority and distributed energy, the optimization control strategy is formulated, and the load priority weight w is determined by the analytic hierarchy process Lk A hierarchical structure model is constructed, the target layer is set as the load priority evaluation, the criterion layer includes the impact on production and the impact factors on life, and the scheme layer is each type of load. Through expert scoring and pairwise comparison, a judgment matrix is constructed, and after consistency test, the priority weight w of each load type is calculated Lk At the same time, considering the power generation characteristics and cost of distributed energy, the coordination weight w of distributed energy is determined through cost-benefit analysis Gj That is, considering the power generation cost C Gj , environmental benefit E Gj and other factors, the calculation is as follows:

[0011]

[0012] Where J is the total number of distributed power sources. Based on these weights, the optimization control strategy is formulated, such as reducing the load with low priority first during the load peak, and adjusting the output of distributed energy according to the coordination weight. The strategy is input into the digital twin model, and the model simulates the operating state changes of the power distribution network under different control strategies according to real-time operating data and preset strategies

[0013] S3, strategy verification and evaluation

[0014] In the digital twin model, the system power balance index Where P Gn , Q Gn are the active and reactive power of the distributed power source at node n, P Ln , Q Ln are the active and reactive power of the load at node n, and the voltage stability index Where V rated is the rated voltage, and the network loss index Evaluate the strategy, where R mnR is the resistance of line mn, simulation is run, analysis index change judges strategy feasibility and effect;

[0015] S4, strategy adjustment and optimization

[0016] According to the evaluation results, if the strategy has problems, the particle swarm optimization algorithm is used to adjust the control parameters in the strategy, the position and speed of the particle are updated iteratively, and the control parameter combination that minimizes the objective function is found, so as to adjust and optimize the control strategy;

[0017] S5, actual application and feedback

[0018] The optimized control strategy after verification is applied to the actual distribution network, and the operation data is fed back to the digital twin model by real-time monitoring system, and the model adjusts the parameters online to adapt to the actual changes.

[0019] Further, the deployment position of the sensor in the step of constructing the digital twin model is optimized and selected, considering the topological structure and electrical characteristics of the distribution network, and a graph theory-based method is used to determine the key nodes and key lines. For radial distribution networks, the topological graph of the distribution network is traversed by using a depth-first search algorithm, and the betweenness centrality index BC n of each node is calculated.

[0020]

[0021] Where σ st is the number of shortest paths from node s to node t, and σ s t(n) is the number of shortest paths from node s to node t through node n. The nodes with betweenness centrality index greater than a set threshold BC threshold are regarded as key nodes, and sensors are deployed on these key nodes and lines connecting the key nodes to ensure that key data reflecting the overall operation state of the distribution network can be accurately collected. At the same time, for ring distribution networks, sensors are deployed near the tie switches of the ring network and at key load nodes and power nodes in the ring network to monitor ring network switching and power flow. Through this optimized deployment method, the number of sensors can be reduced and the cost can be reduced while ensuring data accuracy.

[0022] Further, in the step of pre-rehearsing the optimized control strategy, the control strategy of the energy storage system is also introduced. The energy storage system plays a role in peak shaving and valley filling in the distribution network to improve the stability and reliability of the system. First, according to the charge and discharge power limits P c,max , P d,max and the state of charge range SOC min , SOC maxDetermine the operating constraints of the energy storage system, when the system power is surplus, preferentially store the excess power into the energy storage system, the charging power P c Determine according to the following formula:

[0023] P c = min (P excess , P c,max , (SOC max -SOC) x C / Δt)

[0024] Where P excess is the system surplus power, SOC is the current state of charge of the energy storage system, C is the capacity of the energy storage system, and Δt is the time interval, when the system power is insufficient, the energy storage system discharges, the discharge power P d Determine according to the following formula:

[0025] P d = min (P deficit , P d,max , SOC x C / Δt)

[0026] Where P deficit is the system power shortage, by combining the control strategy of the energy storage system with the load priority and the distributed energy coordination strategy, the control effect of the multi-source load of the distribution network is further optimized.

[0027] Further, in the strategy verification and evaluation step, the load satisfaction index LS is also considered, which is used to measure the user's satisfaction with the power supply quality, and is calculated according to the outage loss and outage time of different types of loads. For industrial loads, the outage loss includes economic loss caused by production stagnation and equipment restart cost, and for residential loads, the outage loss mainly considers the impact of inconvenience, let the outage loss of the kth load be L Lk , the outage time be T Lk , then the load satisfaction index is:

[0028]

[0029] Where and are the maximum outage loss and the longest outage time that the kth load can bear, respectively, and w Lk is the load priority weight determined by the analytic hierarchy process. By comprehensively considering the load satisfaction index, the influence of the optimized control strategy on the user can be more comprehensively evaluated, and the optimized control strategy can ensure the stable operation of the power grid while meeting the user's demand to the greatest extent.

[0030] Further, the strategy adjustment and optimization utilizes a particle swarm optimization algorithm to adjust the control parameters in the strategy, with the voltage deviation in the voltage out-of-limit area being minimized as the objective function F:

[0031]

[0032] Each particle in the particle swarm represents a set of control parameters, and the velocity v i and position x i of the particle are updated according to the following formula:

[0033] v i (t+1)=ωv i (t)+c1r1(pbest i (t)-x i (t))+c2r2(gbest(t)-x i (t))

[0034] x i (t+1)=x i (t)+v i (t+1)

[0035] where ω is the inertia weight, c1 and c2 are learning factors, r1 and r2 are random numbers between 0 and 1, pbest i (t) is the historical optimal position of particle i at time t, and gbest(t) is the global optimal position of the entire particle swarm at time t.

[0036] Further, in the strategy adjustment and optimization step, when adjusting the strategy parameters by using the particle swarm optimization algorithm, in order to prevent the particle swarm from falling into a local optimum, an adaptive inertia weight adjustment strategy is introduced, and the inertia weight ω changes dynamically with the increase of the iteration number t, and the calculation formula is as follows:

[0037]

[0038] where ω max is the initial inertia weight, ω min is the minimum inertia weight, and T maxFor the maximum number of iterations, at the beginning of the iteration, the larger inertia weight enables the particles to search the solution space in a larger range, which is beneficial to global search. With the increase of the number of iterations, the inertia weight gradually decreases, so that the particles focus on local search, improve the convergence accuracy of the algorithm, and dynamically adjust the learning factors c1 and c2. At the beginning of the iteration, c1 is larger, encouraging particles to search for their own historical optimal position to fully utilize individual experience. With the iteration, c2 gradually increases, prompting particles to search for the global optimal position, accelerating the convergence speed of the algorithm. Through this adaptive adjustment strategy, the performance of the particle swarm optimization algorithm can be effectively improved, and the optimal control strategy parameters can be found more quickly and accurately.

[0039] Further, in the actual application and feedback step, in order to improve the response speed of the digital twin model to the changes of the actual running environment, a distributed computing architecture is adopted, which distributes different parts of the digital twin model on multiple computing nodes for parallel computing. Each computing node is responsible for processing a part of data and computing tasks, and data interaction and synchronization are performed through high-speed network. In the data processing module, different computing nodes are responsible for collecting sensor data in different areas and performing preliminary preprocessing. In the model calculation module, different layers of the neural network are calculated in parallel by each computing node, greatly improving the computing efficiency of the model. At the same time, a distributed database is used to store real-time running data and model parameters, ensuring data consistency and scalability. Through this distributed computing architecture, a large amount of real-time data can be processed quickly, and the digital twin model can be adjusted and optimized in time to better adapt to the complex and variable operating environment of the actual power distribution network.

[0040] Further, in the step of constructing the digital twin model, considering the uncertainty and noise interference of the power distribution network operation data, data fusion and noise reduction technology is adopted. First, feature extraction is performed on data from different types of sensors, and then a data fusion method based on Bayesian estimation is used to fuse the feature data of different sensors to improve the accuracy and reliability of the data. In the fusion process, different weights w s are assigned to each sensor data according to the accuracy and reliability of the sensor, and the fused data is calculated by the Bayesian formula:

[0041]

[0042] where P(X|D) is the probability distribution of the fused data, P(D|X) is the probability distribution of X given the data D, and P(X) is the prior probability distribution of X. In order to reduce noise interference, an adaptive filtering algorithm is used to perform noise reduction processing on the fused data. The adaptive filtering algorithm automatically adjusts the parameters of the filter according to the statistical characteristics of the data, and the filter coefficients w iMinimizing the mean square error between the filter output and the desired signal:

[0043] w i (n+1)=w i (n)+2μe(n)x(n-i)

[0044] Where μ is the step factor, e(n) is the error signal, x(n) is the input signal, through data fusion and noise reduction technology, the quality of the input data of the digital twin model can be improved, thereby improving the simulation accuracy of the model.

[0045] Further, in the optimization control strategy pre-rehearsal step, considering the future development trend of the power distribution network, the electric vehicle charging pile load is finely modeled and controlled, and the electric vehicle charging pile load has randomness and aggregation, and its charging behavior is affected by various factors such as user travel habits and electricity price policy. First, the user travel data of the electric vehicle is analyzed, a user travel behavior model based on Markov chain is established, the arrival time, departure time and charging demand of the electric vehicle are predicted, then according to the electricity price policy and the power grid load, a time-of-use charging strategy of the electric vehicle charging pile is formulated, the electric vehicle is encouraged to charge during the low electricity price period and the power grid load valley period, the user is guided to adjust the charging behavior through the price incentive mechanism, and the electricity price p(t), the power grid load L(t) and the electric vehicle charging demand P(t) are set. The charging control strategy is as follows: ev

[0046]

[0047] Wherein, p threshold is the electricity price threshold, L threshoid is the power grid load threshold, and α is the charging power adjustment coefficient adjusted according to the actual situation, and the value range is (0, 1). Through the fine modeling and control of the electric vehicle charging pile load, it is included in the overall optimization control strategy, which can better balance the load of the power distribution network, improve the adaptability of the power distribution network to new loads, further optimize the control effect of the power distribution network multi-source load, reduce the impact on the power grid caused by large-scale charging of electric vehicles, and ensure the stable operation of the power grid. At the same time, combined with the load priority, distributed energy cooperation and energy storage system control strategy, a more comprehensive and efficient power distribution network multi-source load optimization control system is formed to adapt to the complex and changeable power distribution network operation environment.

[0048] Compared with the prior art, the power distribution network multi-source load intelligent optimization control method has the following beneficial effects:

[0049] ​The present application realizes intelligent optimization control of multi-source load of the power distribution network by constructing a digital twin model and optimizing a control strategy, which can accurately map the topology structure, electrical parameters and real-time operating state of the physical power distribution network, and formulate an optimization control strategy according to the load priority and the collaborative weight of the distributed energy, which not only improves the operation efficiency and stability of the power distribution network, but also ensures the reliability and safety of the load supply, and through the simulation operation and strategy verification of the digital twin model, potential problems can be found and solved in time, so that risks and failures that may occur in the operation process of the power distribution network can be effectively avoided.

[0050] The present application introduces a load satisfaction index in the strategy verification and evaluation stage, and considers the fine modeling and control of the electric vehicle charging pile load, which not only can more comprehensively evaluate the influence of the optimization control strategy on the user, ensure that the strategy can guarantee the stable operation of the power grid while meeting the user demand to the greatest extent, but also can better adapt to the development trend of the future power distribution network, and further optimize the control effect of the multi-source load of the power distribution network by fine management of the electric vehicle charging pile load, reduce the impact on the power grid caused by large-scale charging of electric vehicles, and ensure the stable operation of the power grid, in addition, the present application also adopts a distributed computing architecture and data fusion and noise reduction technology, improves the calculation efficiency and input data quality of the digital twin model, and thus improves the overall optimization control effect.

[0051] Other advantages, objects, and features of the present application will be in part apparent and in part pointed out hereinafter in the specification, and in part will be observed by persons skilled in the art upon examination of the following specification, or can be learned by practice of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creating any creative labor.

[0053] Figure 1 Flow chart for operation of power distribution network multi-source load intelligent optimization control method;

[0054] Figure 2 Flow chart for operation of power distribution network multi-source load intelligent optimization control method. DETAILED DESCRIPTION

[0055] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0056] Example 1

[0057] This embodiment describes a power distribution network in a small town with 50 nodes, a certain number of distributed power sources, and various types of loads, such as residential loads, small commercial loads, and a small number of industrial loads. Due to urban development, electricity demand has increased, and the distributed power sources are affected by weather and have unstable power generation, resulting in frequent voltage fluctuations and power imbalances in the power distribution network. During the peak summer electricity consumption period, residential air conditioning loads increase significantly, and distributed photovoltaic power generation is reduced due to cloud cover, causing voltages in some areas to be below the normal range, affecting the normal operation of residential and commercial electrical equipment.

[0058] Key nodes and lines are identified using graph theory-based methods. The distribution network topology is traversed using a depth-first search algorithm, and the betweenness centrality index of each node is calculated. Assuming a threshold BC is set threshold =0.2. After calculation, 10 key nodes were identified. These nodes are distributed in areas with concentrated loads and near power access points. Sensors were deployed on these key nodes and connecting lines to collect real-time data on node voltage, line current, power of various types of loads, and power operation of distributed power sources.

[0059] Based on the strategy of coordinating load priority and distributed energy resources, the load priority weight w is determined by the analytic hierarchy process. Lk A hierarchical model was constructed, with the target layer for load priority assessment, the criteria layer including factors affecting production and daily life, and the alternative layer representing various load types. Relevant experts were organized to conduct pairwise comparisons and scoring of different load types under the criteria layer factors. The analytic hierarchy process (AHP) was used to calculate the priority weights w for residential load, small commercial load, and industrial load. L1 =0.3, w L2 =0.35, w L3 =0.35.

[0060] In the digital twin model, based on the system power balance index Voltage stability index Network loss indicators and load satisfaction indicators The evaluation strategy involved simulating the system for 100 time steps, each lasting 15 minutes, and recording the changes in various indicators. Under the initial strategy, the system power balance indicator P... balanceThe voltage stability index V stab exceeds the allowed range during peak electricity consumption periods, with a maximum of 150 kW. The voltage deviation of some nodes exceeds 5%, and the network loss index L loss is high, reaching 30 kW. The load satisfaction index LS is 0.75, indicating that some users are not satisfied with the power supply quality.

[0061] If the evaluation finds problems with the strategy, the particle swarm optimization algorithm is used to minimize the voltage deviation in the voltage out-of-limit area, with the objective function F = ∑ n∈越限区域 |V n -V rated | adjustment control parameters, with each particle in the particle swarm representing a set of control parameters. The number of particles is set to 30, the initial inertia weight ω max = 0.9, the minimum inertia weight ω min = 0.4, the maximum number of iterations T max = 50, the learning factor c1 = c2 = 2, and the particle velocity and position update formula is v i (t+1) = ωv i (t) + c1r1(pbest i (t) - x i (t)) + c2r2(gbest(t) - x i (t)), x i (t+1) = x i (t) + v i (t+1), an adaptive inertia weight adjustment strategy is introduced, and the learning factors c1 and c2 are dynamically adjusted. At the beginning of iteration, c1 = 2.5, encouraging particles to search for their own historical optimal positions. As iteration progresses, c2 gradually increases, and after the 30th iteration, c2 = 2.5, prompting particles to search for global optimal positions. After 50 iterations, the control parameter combination that minimizes the objective function is found. The optimized strategy significantly reduces the voltage deviation in the voltage out-of-limit area.

[0062] The verified strategy is applied to the actual distribution network. Real-time monitoring system obtains operation data feedback to digital twin model adopts distributed computing architecture, different parts of digital twin model, such as data acquisition and processing module, model calculation module, result analysis module, are distributed on 5 computing nodes for parallel computing, each computing node is responsible for processing part of data and calculation task, data interaction and synchronization are carried out through high-speed network, at the same time, distributed database is used to store real-time operation data and model parameters, ensuring data security and reliability. In actual operation process, according to real-time feedback data, digital twin model automatically adjusts parameters once an hour to adapt to actual changes. After a period of operation, the system power balance index P balanceThe voltage stability index V is stabilized within 50 kW during the peak power consumption period stab The display node voltage deviation is within 3%, and the network loss index L loss is reduced to 20 kW, the load satisfaction index LS is improved to 0.9, and the operation of the distribution network is effectively improved.

[0063] Example Two

[0064] This example describes a certain industrial park with many industrial enterprises, diverse production equipment and complex power demand. At the same time, the park is equipped with distributed power and energy storage systems. Different enterprises have different production importance and sensitivity to power outage.

[0065] Based on the topology of the park distribution network, the key nodes are determined using the depth-first search algorithm. Sensors are deployed at these nodes and related lines to collect node voltage, line current, enterprise load power, and distributed power data. A multi-layer neural network digital twin model is constructed, and the activation function parameters are optimized and determined using a genetic algorithm. The model is continuously iteratively trained to accurately map the operation state of the distribution network.

[0066] Using the analytic hierarchy process, the priority weight of each enterprise load is determined considering the influencing factors on production. The distributed energy coordination weight is determined by considering the generation cost and environmental benefits. The energy storage system control strategy is introduced. When the system power is surplus, the charging power is determined to store excess power according to the formula. When the power is insufficient, the discharging power is determined to supplement the power. Considering future development, the electric vehicle charging pile load in the park is modeled, and a time-of-use charging strategy is developed based on user travel data and electricity price policies.

[0067] In the digital twin model, the system power balance, voltage stability, network loss, and load satisfaction index evaluation strategies are used. For example, if a certain enterprise experiences a large economic loss due to power outage in a critical production link, the load satisfaction index will be highlighted in the calculation. If the strategy results in excessive power outage time for the enterprise, it will be determined that the strategy is not feasible.

[0068] If the evaluation finds voltage out-of-limit problems, the particle swarm optimization algorithm is used to adjust the control parameters with the minimum voltage deviation in the voltage out-of-limit area as the objective function. The adaptive inertia weight adjustment strategy is used to avoid the algorithm falling into local optimum.

[0069] The optimization strategy is applied to the park distribution network, and real-time monitoring data is fed back to the digital twin model. Distributed computing architecture and distributed database are used to improve the response speed of the model and realize online adjustment of the parameters of the model.

[0070] The above merely describes the preferred embodiments of the present application, and is not intended to limit the present application in any form. Although the present application has been disclosed with the preferred embodiments as above, it is not intended to limit the present application. Any person skilled in the art can make some changes or modifications to the above disclosed technical content to obtain equivalent embodiments with equivalent changes, as long as the changes or modifications do not deviate from the technical solution of the present application. Any modification, change, equivalent change and modification of the above embodiments made according to the technical essence of the present application still belong to the scope of the technical solution of the present application.

Claims

1. A method for intelligent optimization control of multi-source load of a power distribution network, characterized in that, The specific steps of the method are: S1, constructing a digital twin model In the physical power distribution network, sensors are deployed to collect node voltage, line current, various types of load power, and distributed power operation data. A multi-layer neural network digital twin model is constructed using deep learning algorithms. The input layer is the collected data, the hidden layer is transformed through an activation function with feature extraction and nonlinearity, where α and β are parameters optimized by genetic algorithms based on historical data and operating characteristics of the power distribution network, and the output layer is the predicted operating state of each node The mean square error between the predicted value and the actual value is minimized The model is trained, where N is the total number of nodes, S n is the actual operating state of node n. Through continuous iteration and training, the digital twin model can accurately map the topology, electrical parameters, and real-time operating state of the physical power distribution network. S2, optimizing control strategy pre-play Based on the load priority and distributed energy collaborative development of optimization control strategy, through the analytic hierarchy process to determine the load priority weight w Lk , build a hierarchical model, the target layer is set as load priority evaluation, the criterion layer includes the impact on production, the impact factors on life, the scheme layer is each type of load, while considering the distributed energy generation characteristics and cost, through cost-benefit analysis to determine the collaborative weight w Gj of distributed energy, that is, comprehensive consideration of power generation cost C Gj , environmental benefits E Gj Factors, calculation: Where J is the total number of distributed power sources, based on these weights, the optimization control strategy is formulated, the strategy is input into the digital twin model, and the model simulates the operating state changes of the power distribution network under different control strategies according to real-time operating data and preset strategies; S3, strategy verification and evaluation In the digital twin model, according to the system power balance index Where P Gn , Q Gn are the active and reactive power of the distributed power supply at node n, P Ln , Q Ln are the active and reactive power of the load at node n, the voltage stability index Where V rated is the rated voltage, the network loss index The evaluation strategy, where R mn is the resistance of the line mn, the simulation operation, the analysis index change judges the feasibility and effect of the strategy; S4, strategy adjustment and optimization According to the evaluation results, if problems are found in the strategy, the particle swarm optimization algorithm is used to adjust the control parameters in the strategy, the position and speed of the particles are updated iteratively to find the control parameter combination that minimizes the objective function, and thus the control strategy is adjusted and optimized; S5, actual application and feedback The verified and optimized control strategy is applied to the actual power distribution network, and the real-time monitoring system obtains operating data feedback to the digital twin model, and the model adjusts the parameters online to adapt to actual changes. 2.The power distribution network multi-source load intelligent optimization control method according to claim 1, characterized in that, The deployment position of the sensor in the step of constructing the digital twin model is selected by optimization, considering the topological structure and electrical characteristics of the power distribution network, and a graph theory-based method is used to determine the key nodes and key lines. For a radial power distribution network, the topological graph of the power distribution network is traversed by a depth-first search algorithm, and the betweenness centrality index BC of each node is calculated n : where σ st is the number of shortest paths from node s to node t, σ s t(n) is the number of shortest paths from node s to node t through node n, and nodes with betweenness centrality index greater than a set threshold BC threshold are regarded as key nodes. Sensors are deployed on these key nodes and lines connecting the key nodes. For ring distribution networks, sensors are deployed near the tie switches of the ring network and at key load nodes and power supply nodes in the ring network to monitor ring network switching and power flow. 3.The power distribution network multi-source load intelligent optimization control method according to claim 1, characterized in that, The control strategy of the energy storage system is also introduced in the optimization control strategy pre-rehearsal step. The energy storage system plays a role in peak shaving in the distribution network. First, the operation constraints of the energy storage system are determined according to the charge and discharge power limits P c,max 、P d,max and the state of charge range SOC min 、SOC max When the control strategy is formulated, when the system power is excessive, the excess power is stored in the energy storage system in priority. The charging power P c is determined according to the following formula: P c = min(P excess ,P c,max ,(SOC max -SOC) x C / Δt) where P excess is the system excess power, SOC is the current state of charge of the energy storage system, C is the energy storage system capacity, Δt is the time interval, the energy storage system discharges when the system power is insufficient, the discharge power P d is determined according to the following formula: P d = min(P deficit ,P d,max ,SOC x C / Δt) where P deficit is the system power shortage, the control strategy of the energy storage system is combined with the load priority and the distributed energy coordination strategy to further optimize the control effect of the multi-source load of the distribution network.

4. The power distribution network multi-source load intelligent optimization control method according to claim 1, characterized in that, The load satisfaction index LS is considered in the policy verification and evaluation step, and is used to measure the satisfaction of users to the power supply quality. The load satisfaction index is calculated according to the power failure loss and power failure time of different types of loads. For industrial loads, the power failure loss includes economic loss caused by production stagnation and equipment restart cost. For residential loads, the power failure loss mainly considers the impact of inconvenience. Let the power failure loss of the kth type of load be L Lk , and the power failure time be T Lk , then the load satisfaction index is: wherein and respectively the maximum loss of load and the maximum outage time that the kth class of load can withstand, w Lk is the load priority weight determined by the analytic hierarchy process.

5. The power distribution network multi-source load intelligent optimization control method according to claim 1, characterized in that, In the strategy adjustment and optimization, the particle swarm optimization algorithm is used to adjust the control parameters in the strategy, and the minimum voltage deviation in the voltage out-of-limit area is taken as the objective function F: Each particle in the swarm represents a set of control parameters, the velocity v i and position x i The update equations are as follows: v i (t+1) = ωv i (t) + c1r1(pbest i (t) - x i (t)) + c2r2(gbest(t) - x i (t)) x i (t+1) = x i (t) + v i (t+1) where ω is the inertia weight, c1, c2 are learning factors, r1, r2 are random numbers between [0, 1], pbest i (t) is the historical optimal position of particle i at time t, gbest(t) is the global optimal position of the entire particle swarm at time t.

6. The power distribution network multi-source load intelligent optimization control method according to claim 1, characterized in that, In the strategy adjustment and optimization step, when the particle swarm optimization algorithm is used to adjust the strategy parameters, in order to avoid the particle swarm from falling into local optimization, an adaptive inertia weight adjustment strategy is introduced, and the inertia weight ω changes dynamically with the increase of the iteration number t, and its calculation formula is: where ω max is the initial inertia weight, ω min is the minimum inertia weight, T max is the maximum number of iterations. At the beginning of the iteration, the larger inertia weight enables the particles to search the solution space in a larger range, which is beneficial to global search. As the number of iterations increases, the inertia weight gradually decreases, which makes the particles focus on local search. Meanwhile, the learning factors c1 and c2 are also dynamically adjusted. At the beginning of the iteration, c1 is larger, which encourages the particles to search the historical optimal position of themselves to make full use of individual experience. As the iteration proceeds, c2 gradually increases, which prompts the particles to search the global optimal position to accelerate the convergence speed of the algorithm.

7. The power distribution network multi-source load intelligent optimization control method according to claim 1, characterized in that, In the actual application and feedback step, in order to improve the response speed of the digital twin model to the changes of the actual operating environment, a distributed computing architecture is adopted, different parts of the digital twin model are distributed on multiple computing nodes for parallel computing, each computing node is responsible for processing a part of data and computing tasks, data interaction and synchronization are carried out through high-speed network, and a distributed database is used to store real-time operating data and model parameters. 8.The power distribution network multi-source load intelligent optimization control method of claim 1, wherein, In the step of constructing the digital twin model, considering the uncertainty and noise interference of the power distribution network operation data, data fusion and noise reduction technology is adopted, first, the data from different types of sensors are extracted, then the data fusion method based on Bayesian estimation is used to fuse the feature data of different sensors, in the fusion process, according to the accuracy and reliability of the sensor, different weights w are assigned to each sensor data s , the fused data is calculated by Bayesian formula: Wherein P(X|D) is the fused data probability distribution, P(D|X) is the probability distribution of X under the condition of known data D, P(X) is the prior probability distribution of X, and in order to reduce noise interference, an adaptive filtering algorithm is used for noise reduction processing on the fused data, the adaptive filtering algorithm automatically adjusts the parameters of the filter according to the statistical characteristics of the data, and the filter coefficient w i is continuously adjusted, so that the mean square error between the filter output and the expected signal is minimized: w i (n+1) = w i (n) + 2μe(n)x(n-i) Where μ is the step factor, e(n) is the error signal, and x(n) is the input signal. 9.The power distribution network multi-source load intelligent optimization control method of claim 1, wherein, In the optimization control strategy pre-rehearsal step, the future power distribution network development trend is considered, the electric vehicle charging pile load is finely modeled and controlled, the electric vehicle charging pile load has randomness and aggregation, the charging behavior is influenced by various factors such as user travel habits and electricity price policy, firstly, the user travel data of the electric vehicle is analyzed, a user travel behavior model based on Markov chain is established, the arrival time, departure time and charging demand of the electric vehicle are predicted, then according to the electricity price policy and the power grid load condition, the time-of-use charging strategy of the electric vehicle charging pile is formulated, the electric vehicle is encouraged to charge in the low electricity price period and the power grid load low valley period, the user is guided to adjust the charging behavior through the price incentive mechanism, the electricity price p(t), the power grid load L(t) and the electric vehicle charging demand P ev (t) are set, and the charging control strategy is as follows: wherein p threshold is the electricity price threshold, L threshoid is the grid load threshold, and a is a charging power adjustment coefficient adjusted according to actual conditions. By fine modeling and control of the electric vehicle charging pile load, the electric vehicle charging pile load is incorporated into the overall optimization control strategy, so that the load of the distribution network can be better balanced, the adaptability of the distribution network to new loads can be improved, the control effect of the distribution network multi-source load can be further optimized, the impact of large-scale electric vehicle charging on the power grid can be reduced, the stable operation of the power grid can be ensured, and a more comprehensive and efficient distribution network multi-source load optimization control system can be formed by combining the load priority, distributed energy coordination and energy storage system control strategy to adapt to the complex and changeable distribution network operation environment in the future.

Images