Power distribution network load transfer strategy generation method, device and equipment and readable storage medium

Abstract

The application discloses a power distribution network load transfer strategy generation method, device and equipment and a readable storage medium. The method acquires runtime sequence data of a power distribution network; it is judged whether the runtime sequence data indicates that the power distribution network is in an abnormal period; if yes, a prediction model is acquired, and the prediction model is used to predict running state information of a next period based on the runtime sequence data; whether a transfer strategy needs to be generated is judged based on the running state information and the runtime sequence data; if yes, the prediction model is used to determine a total transfer period; a multi-objective transfer strategy model is established based on the total transfer period, with the target of maximizing the total amount of restored loads and minimizing the number of switch operations; the multi-objective transfer strategy model is solved, and a transfer strategy corresponding to the total transfer period is generated. It can be seen that the application generates a transfer strategy with the target of maximizing the total amount of restored loads and minimizing the number of switch operations, thereby improving safety and power supply reliability.

Description

Technical Field

[0001] This application relates to the field of power distribution network technology, and more specifically, to a method, apparatus, equipment, and readable storage medium for generating power distribution network load transfer strategies. Background Technology

[0002] With the advancement of the construction of new power systems, the complexity of power grid topology has increased significantly, making it more difficult to generate load transfer strategies.

[0003] Traditional load transfer strategies for distribution networks prioritize maximizing load restoration, neglecting to consider operational costs and equipment safety. Specifically, during load restoration, these methods often trigger frequent switching operations due to an excessive focus on load restoration rates. This not only accelerates the mechanical wear of critical equipment such as circuit breakers and disconnectors, shortening their lifespan, but also may lead to secondary problems like voltage dips and power flow exceeding limits due to improper timing, resulting in insufficient safety assurance for the distribution network.

[0004] Against this backdrop, the optimized dispatching of distribution networks faces the challenge of balancing load recovery efficiency and equipment safety operation. There is an urgent need to provide a new method for generating load transfer strategies for distribution networks to effectively improve the safety of power grid operation and the reliability of power supply. Summary of the Invention

[0005] In view of this, this application provides a method, apparatus, equipment and readable storage medium for generating load transfer strategies in a distribution network, which addresses the shortcomings of existing load transfer technologies in distribution networks that cannot coordinate load recovery efficiency and equipment safe operation.

[0006] To achieve the above objectives, the following solution is proposed:

[0007] A method for generating a load transfer strategy for a distribution network includes:

[0008] Obtain runtime sequence data of the power distribution network;

[0009] Determine whether the runtime sequence data indicates that the power distribution network is in an abnormal cycle;

[0010] If so, a prediction model is obtained, and the prediction model is used to predict the operating status information of the next cycle based on the runtime sequence data; based on the operating status information and the runtime sequence data, it is determined whether a transfer strategy needs to be generated; if so, the prediction model is used to determine the total transfer period; with the goal of maximizing the total restored load and minimizing the number of switching operations, a multi-objective transfer strategy model is established based on the total transfer period; the multi-objective transfer strategy model is solved to generate the transfer strategy corresponding to the total transfer period.

[0011] Optionally, determining whether the runtime sequence data indicates that the distribution network is in an abnormal cycle includes:

[0012] Determine the number of abnormal moments and the number of normal moments contained in the runtime sequence data;

[0013] Calculate the ratio of the number of abnormal moments to the number of normal moments;

[0014] When the ratio is not less than the abnormal threshold, the power distribution network is determined to be in an abnormal cycle;

[0015] When the ratio is less than the abnormal threshold, it is determined that the power distribution network is not in an abnormal cycle.

[0016] Optionally, the runtime timing data includes voltage deviation values ​​at multiple time points;

[0017] Determining the number of abnormal moments and the number of normal moments contained in the runtime sequence data includes:

[0018] Based on the voltage deviation value at each time point, the temperature value and temperature limit of the distribution network at different time points are used to calculate the operating status value at different time points;

[0019] Calculate the average state value for each operating state value;

[0020] When the absolute value of the difference between the running status value at a corresponding time point and the average status value is greater than a preset status threshold, that time point is regarded as an abnormal time point.

[0021] When the absolute value of the difference between the running status value at the corresponding time point and the average status value is not greater than the preset status threshold, the time point is regarded as the normal time point.

[0022] The total number of all abnormal time points is counted as the number of abnormal time points;

[0023] The total number of all normal time points is counted as the number of normal time points.

[0024] Optionally, obtaining the prediction model includes:

[0025] Acquire training data corresponding to multiple consecutive historical periods. Each historical period corresponds to two types of periods: one type is abnormal historical periods, and the other type is normal historical periods.

[0026] Construct an LSTM model;

[0027] A Transformer encoding layer is added to the LSTM model, and the forgetting layer of the LSTM model is converted into a dynamically enhanced forgetting layer to obtain the initial prediction model.

[0028] The initial prediction model is trained using various training data, and the parameters of the initial prediction model are adjusted until the initial prediction model meets the preset stopping conditions; the final initial prediction model is the prediction model.

[0029] Optionally, converting the forgetting layer of the LSTM model into a dynamically enhanced forgetting layer includes:

[0030] The forgetting layer output function of the LSTM model is replaced with a dynamic set of output functions to convert the forgetting layer of the LSTM model into a dynamically enhanced forgetting layer.

[0031] The dynamic output function set is as follows:

[0032]

[0033]

[0034]

[0035] In the formula, To dynamically enhance the output of the forgetting layer; This is the output of the forgetting layer; This is a dynamic adjustment coefficient; This is a dynamic forgetting weight matrix; Let Re be the historical prediction error vector, and Re be the modified linear unit activation function. Forgetting weight matrix, The table shows the hidden state at time i-1; This is the input at time i; It is the bias vector; P is the output function of the forgetting layer. i P represents the node power at time i; i-1 Let P represent the node power at time i-1; P is the node's rated power.

[0036] Optionally, determining whether a transfer strategy needs to be generated based on the running status information and the running sequence data includes:

[0037] Determine the number of abnormal moments contained in the runtime sequence data;

[0038] Determine the total number of abnormal moments contained in the operational status information;

[0039] Calculate the ratio of the total number of abnormal moments to the number of abnormal moments;

[0040] When the abnormal ratio is greater than a preset ratio threshold, it is determined that a transfer strategy needs to be generated.

[0041] If the abnormal ratio is not greater than a preset ratio threshold, it is determined that no transfer strategy needs to be generated.

[0042] Optionally, determining the total supply period using the prediction model includes:

[0043] The period corresponding to the aforementioned operating status information is taken as the total supply period;

[0044] Using the prediction model, based on the operating status information, predict the state time series information of the next cycle of the operating status information;

[0045] Determine whether the state timing information corresponds to an abnormal cycle;

[0046] If so, add the abnormal period corresponding to the status time sequence information after the total transfer period, and update the total transfer period; use the status time sequence information as the new operating status information, and return to execute the step of using the prediction model to predict the status time sequence information of the next period of the operating status information based on the operating status information;

[0047] If not, then determine the final total transfer period.

[0048] A power distribution network load transfer strategy generation device, comprising:

[0049] The acquisition module is used to acquire runtime sequence data of the power distribution network;

[0050] The judgment module is used to determine whether the runtime sequence data indicates that the power distribution network is in an abnormal cycle; if so, the generation module is called.

[0051] A generation module is used to acquire a prediction model and use the prediction model to predict the operating status information of the next cycle based on the runtime sequence data; based on the operating status information and the runtime sequence data, it determines whether a transfer strategy needs to be generated; if so, it uses the prediction model to determine the total transfer period; with the goal of maximizing the total restored load and minimizing the number of switching operations, it establishes a multi-objective transfer strategy model based on the total transfer period; it solves the multi-objective transfer strategy model to generate the transfer strategy corresponding to the total transfer period.

[0052] A power distribution network load transfer strategy generation device includes a memory and a processor;

[0053] The memory is used to store programs;

[0054] The processor is used to execute the program to implement each step of the above-described method for generating power distribution network load transfer strategies.

[0055] A readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the various steps of the above-described method for generating a power distribution network load transfer strategy.

[0056] As can be seen from the above technical solution, the distribution network load transfer strategy generation method provided in this application can acquire the operation sequence data of the distribution network; determine whether the operation sequence data indicates that the distribution network is in an abnormal cycle; based on this, this application can assess whether the current distribution network is in an abnormal period; if so, a prediction model is acquired, and the prediction model is used to predict the operation status information of the next cycle based on the operation sequence data; based on the operation status information and the operation sequence data, it is determined whether a transfer strategy needs to be generated; based on this, this application can proactively predict the severity of the abnormality in the next cycle by comprehensively predicting the operation status information and the current operation sequence data, and activate the transfer strategy when the abnormality is relatively severe to avoid resource waste caused by excessive transfer; if it is determined that a transfer strategy needs to be generated, the prediction model is used to predict the operation status information and the current operation sequence data. The model determines the total power transfer period. Based on this, the application can determine the time boundary of the power transfer strategy and determine the duration of power transfer based on the trend of operating status changes output by the prediction model. Compared with the fixed time period setting of traditional power transfer methods, this application can determine the time period dynamically, ensuring that the power transfer duration accurately matches the actual abnormal cycle, reducing the possibility of abnormal recurrence, and improving the timeliness and economy of the strategy. Subsequently, this application establishes a multi-objective power transfer strategy model based on the total power transfer period, with the goal of maximizing the total restored load and minimizing the number of switching operations. The multi-objective power transfer strategy model is solved to generate the power transfer strategy corresponding to the total power transfer period. Based on this, this application can coordinate the optimization of load restoration efficiency and equipment safe operation by generating a multi-objective power transfer strategy model, avoiding the power transfer strategy from being one-sided and ensuring that the power transfer strategy meets the multi-objective balance. It can be seen that this application can generate the power transfer strategy corresponding to the total power transfer period by constructing a multi-objective power transfer strategy model with the goal of maximizing the total restored load and minimizing the number of switching operations, thereby reducing equipment wear, avoiding secondary problems, reducing the possibility of abnormal recurrence, and improving safety and power supply reliability. Attached Figure Description

[0057] 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 embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0058] Figure 1This is a flowchart of a method for generating a power distribution network load transfer strategy, as disclosed in an embodiment of this application.

[0059] Figure 2 This is a structural block diagram of a power distribution network load transfer strategy generation device disclosed in an embodiment of this application;

[0060] Figure 3 This is a hardware structure block diagram of a power distribution network load transfer strategy generation device disclosed in an embodiment of this application. Detailed Implementation

[0061] 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 skilled in the art without creative effort are within the scope of protection of this application.

[0062] This application provides a method for generating a load transfer strategy for a distribution network. This method can be applied to various distribution network systems or power dispatching systems, as well as to various computer terminals or smart terminals. The executing entity can be the processor or server of the computer terminal or smart terminal.

[0063] Next, combine Figure 1 The method for generating the distribution network load transfer strategy in this application is described in detail, including the following steps:

[0064] Step S1: Obtain the operating sequence data of the distribution network.

[0065] Specifically, time points are set with a time interval of t, and data from the distribution network is collected every t time interval through SCADA systems and synchronous phasor measurement devices (PMUs) deployed at key nodes of the distribution network.

[0066] The SCADA system collects voltage, current, and power data from key nodes of the distribution network at second-level intervals, while the PMU device captures transient processes of the power grid and collects PMU data through high-precision synchronous phasor data at the millisecond level.

[0067] Among them, key nodes refer to equipment and / or feeders that participate in the operation of the distribution network.

[0068] Kalman filtering was used to eliminate measurement noise from the high-frequency sampled PMU data. Moving average filtering was used to smooth random fluctuations in the PMU data. Time series prediction was used to fill in missing data caused by communication interruption. The missing values ​​were filled in by combining data from adjacent nodes and spatial correlation analysis to obtain the node data.

[0069] By processing the voltage, current, power data and node data collected by the SCADA system, runtime sequence data containing real-time load rate, voltage deviation and node status at different times of different key nodes can be obtained.

[0070] Among them, the real-time load rate is the ratio of the actual transmitted active power to the rated allowable current carrying capacity;

[0071] Determine the difference between the actual measured voltage and the nominal voltage, and use the ratio of this difference to the nominal voltage as the voltage deviation.

[0072] The node state characterizes the switching state of the corresponding key node at different points in time.

[0073] Therefore, by constructing a comprehensive data acquisition network through SCADA and PMU, both steady-state monitoring and transient events can be covered, thereby improving the reliability of runtime sequence data.

[0074] Step S2: Determine whether the runtime sequence data indicates that the power distribution network is in an abnormal cycle.

[0075] Specifically, the severity of anomalies in runtime sequence data can be used to determine whether the distribution network is in an abnormal cycle.

[0076] Step S3: If yes, obtain the prediction model and use the prediction model to predict the operating status information of the next cycle based on the runtime sequence data; based on the operating status information and the runtime sequence data, determine whether a transfer strategy needs to be generated; if so, use the prediction model to determine the total transfer period; with the goal of maximizing the total restored load and minimizing the number of switching operations, establish a multi-objective transfer strategy model based on the total transfer period; solve the multi-objective transfer strategy model to generate the transfer strategy corresponding to the total transfer period.

[0077] Specifically, when it is determined that the distribution network is in an abnormal state, a trained prediction model can be obtained.

[0078] Using a predictive model, the operating status information for the next cycle is predicted based on runtime sequence data. This operating status information may include the real-time load rate, voltage deviation, and node status of different key nodes at different times in the next cycle.

[0079] By comparing the operating status information and the operating sequence data, it is determined whether the severity of the abnormal situation in the next cycle is higher than that in the current cycle. If so, it is determined that the power supply reliability of the distribution network is not high and cannot be restored on its own, requiring overload transfer.

[0080] In this case, iterative prediction using a forecasting model can be used to determine the total supply period.

[0081] The objective function is constructed to maximize the total restored load and minimize the number of switching operations;

[0082] Based on the total transfer time period and objective function, and considering topological connectivity constraints, node voltage deviation constraints, and line power constraints, a multi-objective transfer strategy model can be established.

[0083] The objective function can be expressed as follows:

[0084]

[0085] Where x refers to the x-th cycle; N is the number of cycles included in the total transfer period; D x The single operation time is the total time required from issuing the critical node switching command to completing the switching action; H x L represents the number of switching actions within the x-th period; x k1 represents the transferable load within the x-th period; k2 represents the weighting coefficient of the operation time; k3 represents the weighting coefficient of the number of switching actions; and k4 represents the weighting coefficient of the transferable load.

[0086] Topological connectivity constraints can be defined as follows:

[0087]

[0088] Where j represents the j-th critical node in the distribution network, This represents the number of connected power supply paths for the j-th critical node, ensuring that each critical node has at least one connected power supply path after the power transfer.

[0089] The node voltage deviation constraint can be defined as follows:

[0090]

[0091] m1>m2

[0092] in, This represents the voltage deviation value at critical node j. The voltage deviation threshold of the critical node j is represented by m1 and m2 represent the degree of constraint.

[0093] The line power constraint can be shown below:

[0094]

[0095] Where k represents the k-th route, This represents the actual active power transmitted by the k-th line. This represents the maximum allowed transmission power of the k-th line.

[0096] The Big M method is introduced to handle the logical relationship between switch states and topology. PCA is used to compress high-dimensional voltage and power data. Hierarchical control HAC is adopted to divide the switching operation into feeder level and segment level. Safety penalty terms such as immediate termination when the voltage exceeds the limit are added. The multi-objective power transfer strategy model is solved through weighted or Pareto front analysis to generate the power transfer strategy corresponding to the total power transfer time period.

[0097] Digital twin simulation technology is used to simulate the distribution network. The newly generated transfer strategy is applied to the simulated distribution network to preview the operational consequences and check whether it will cause a voltage drop. After confirming that the specific operation is correct, the transfer strategy can be pushed to the scheduling interface. The instructions to execute the distribution network transfer strategy are issued through the GOOSE protocol, and the operation log is saved to record the transfer process data for model iteration and optimization.

[0098] If an error occurs during the operation, the solution process will be returned until the specific operation is correct.

[0099] Therefore, by prioritizing the restoration of critical loads to maximize load recovery and improve power supply reliability, and by minimizing the number of operations and time to reduce the number of switching actions and transfer time, the impact of equipment wear and power outages on users is reduced. Through weighted or Pareto front analysis, suboptimal solutions caused by single-objective optimization are avoided. Based on topological connectivity constraints, it is ensured that each load node has at least one connected path after transfer to avoid islanded operation. Through voltage deviation thresholds and line power limits, equipment overload or voltage exceeding limits is prevented. This balances the restoration of load with the operating costs and improves the safety of the transfer strategy implementation process.

[0100] If the runtime sequence data does not indicate that the distribution network is in an abnormal cycle and / or that a transfer strategy does not need to be generated, no action is required.

[0101] As can be seen from the above technical solution, the distribution network load transfer strategy generation method provided in this application can acquire the operation sequence data of the distribution network; determine whether the operation sequence data indicates that the distribution network is in an abnormal cycle; based on this, this application can assess whether the current distribution network is in an abnormal period; if so, a prediction model is acquired, and the prediction model is used to predict the operation status information of the next cycle based on the operation sequence data; based on the operation status information and the operation sequence data, it is determined whether a transfer strategy needs to be generated; based on this, this application can proactively predict the severity of the abnormality in the next cycle by comprehensively predicting the operation status information and the current operation sequence data, and activate the transfer strategy when the abnormality is relatively severe to avoid resource waste caused by excessive transfer; if it is determined that a transfer strategy needs to be generated, the prediction model is used to predict the operation status information and the current operation sequence data. The model determines the total power transfer period. Based on this, the application can determine the time boundary of the power transfer strategy and determine the duration of power transfer based on the trend of operating status changes output by the prediction model. Compared with the fixed time period setting of traditional power transfer methods, this application can determine the time period dynamically, ensuring that the power transfer duration accurately matches the actual abnormal cycle, reducing the possibility of abnormal recurrence, and improving the timeliness and economy of the strategy. Subsequently, this application establishes a multi-objective power transfer strategy model based on the total power transfer period, with the goal of maximizing the total restored load and minimizing the number of switching operations. The multi-objective power transfer strategy model is solved to generate the power transfer strategy corresponding to the total power transfer period. Based on this, this application can coordinate the optimization of load restoration efficiency and equipment safe operation by generating a multi-objective power transfer strategy model, avoiding the power transfer strategy from being one-sided and ensuring that the power transfer strategy meets the multi-objective balance. It can be seen that this application can generate the power transfer strategy corresponding to the total power transfer period by constructing a multi-objective power transfer strategy model with the goal of maximizing the total restored load and minimizing the number of switching operations, thereby reducing equipment wear, avoiding secondary problems, reducing the possibility of abnormal recurrence, and improving safety and power supply reliability.

[0102] In some embodiments of this application, the process of step S2, determining whether the runtime timing data indicates that the distribution network is in an abnormal cycle, is described in detail below:

[0103] S20. Determine the number of abnormal moments and the number of normal moments contained in the runtime sequence data.

[0104] Specifically, the number of abnormal time points contained in the runtime sequence data can be determined as the number of abnormal time points;

[0105] The number of normal time points contained in the runtime sequence data can be determined as the number of normal time points.

[0106] S21. Calculate the ratio of the number of abnormal moments to the number of normal moments.

[0107] Specifically, the ratio between the number of abnormal moments and the number of normal moments can be calculated.

[0108] S22. When the ratio is not less than the abnormal threshold, the power distribution network is determined to be in an abnormal cycle.

[0109] Specifically, the ratio can be compared with an anomaly threshold.

[0110] If the ratio is not less than the abnormal threshold, it can be determined that the distribution network is in an abnormal cycle.

[0111] The abnormal threshold can be determined based on the severity of the distribution network as represented by different ratios in historical periods.

[0112] S23. When the ratio is less than the abnormal threshold, it is determined that the power distribution network is not in an abnormal cycle.

[0113] Specifically, if the ratio is less than the abnormal threshold, it can be determined that the distribution network is not in an abnormal cycle.

[0114] As can be seen from the above technical solution, this embodiment provides an optional method for determining whether the runtime sequence data indicates that the distribution network is in an abnormal cycle. This method enables dynamic anomaly determination, quantifies the state of a time period by the proportion of abnormal time points, avoids misjudgments of occasional anomalies, and improves the reliability of anomaly identification.

[0115] In some embodiments of this application, the process of step S20, determining the number of abnormal moments and the number of normal moments contained in the runtime sequence data, is described in detail as follows:

[0116] S200. Based on the voltage deviation value at each time point, the temperature value and temperature limit of the distribution network at different time points, calculate the operating status value at different time points.

[0117] Specifically, the voltage deviation value at each time point, the temperature value and temperature limit of the distribution network at different time points can be substituted into the state calculation function to calculate the operating state value at different time points.

[0118] The state calculation function can be shown below:

[0119]

[0120] In the formula, S(i) is the operating state value at the i-th time point; Dy(i) represents the voltage deviation at the i-th time point; I[] represents the condition function, which is applied when the condition is met. hour, The value equals 1, when the condition is not met. hour, =0; h(i) is the temperature value at the i-th time point; O is the temperature limit.

[0121] S201. Calculate the average state value of each running state value.

[0122] Specifically, the average value of each operating state value can be calculated to obtain the average state value.

[0123] S202. When the absolute value of the difference between the running status value at the corresponding time point and the average status value is greater than the preset status threshold, the time point is regarded as an abnormal time point.

[0124] Specifically, the difference between the operating status value at each time point and the average status value can be calculated;

[0125] When the absolute value of the difference is greater than the preset state threshold, the corresponding time point is regarded as an abnormal time point.

[0126] S203. When the absolute value of the difference between the running status value at the corresponding time point and the average status value is not greater than the preset status threshold, the time point is taken as the normal time point.

[0127] Specifically, the difference between the operating status value at each time point and the average status value can be calculated;

[0128] When the absolute value of the difference is not greater than the preset state threshold, the corresponding time point is taken as the normal time point.

[0129] S204. Count the total number of all abnormal time points as the number of abnormal time points.

[0130] Specifically, the total number of all abnormal time points in the runtime sequence data can be calculated as the number of abnormal time points.

[0131] S205. Count the total number of all normal time points as the number of normal time points.

[0132] Specifically, the total number of all normal time points in the runtime sequence data can be calculated as the number of normal time points.

[0133] As can be seen from the above technical solution, this embodiment provides an optional method for determining the number of abnormal moments and the number of normal moments contained in the runtime sequence data. This method can comprehensively calculate the operating status value by combining voltage deviation and temperature values, avoiding the limitations of a single indicator and improving the accuracy of anomaly identification.

[0134] In some embodiments of this application, the process of obtaining the prediction model in step S3 is described in detail, and the steps are as follows:

[0135] S30. Obtain training data corresponding to multiple consecutive historical periods. Each historical period corresponds to two types of periods: one type is abnormal historical periods, and the other type is normal historical periods.

[0136] Specifically, multiple training run data sets can be obtained. Each training run data set comes from multiple consecutive historical periods.

[0137] Within each historical cycle, some historical cycles are abnormal, while others are normal.

[0138] Normal historical cycles account for more than 70% of all historical cycles.

[0139] Data separability is verified by calculating the KL divergence between training data from normal historical cycles and training data from abnormal historical cycles, keeping the divergence value below 0.3. The time series characteristics are verified by examining the autocorrelation of the distribution network training data. When the autocorrelation coefficient lagging by multiple time steps is greater than the threshold, it indicates that the data has sufficient time series dependence for predictive modeling.

[0140] S31. Construct the LSTM model.

[0141] Specifically, an LSTM model containing a forgetting layer can be constructed.

[0142] S32. Use the hidden state sequence output by LSTM as the input of the Transformer encoding layer to add the Transformer encoding layer to the LSTM model, and convert the forgetting layer of the LSTM model into a dynamically enhanced forgetting layer to obtain the initial prediction model.

[0143] Specifically, an initial prediction model can be obtained by constructing an LSTM model that includes a forgetting layer and updating the forgetting layer of the LSTM model to a dynamically enhanced forgetting layer.

[0144] S33. Train the initial prediction model using various training data, adjust the parameters of the initial prediction model until the initial prediction model meets the preset stopping condition; the final initial prediction model is the prediction model.

[0145] Specifically, the LSTM is responsible for capturing temporal dependencies, with 64 hidden units and 8 input features. An enhanced dynamic forgetting layer automatically increases the historical information forgetting rate during load abrupt changes. The Transformer encoding layer models the global relationships of temporal features using 4 attention heads. The fully connected layer outputs the probability distributions of two categories: normal and abnormal. When the absolute value of the error between the predicted value and the true value at a certain time point is less than a preset error threshold, the predicted value is considered to match the true value; otherwise, it is considered not to match. If the matching degree between the predicted value and the true value in a historical period is less than or equal to 70% (i.e., the ratio of the predicted value matching the true value to the total number of time points in that historical period is less than or equal to 70%), it is determined that the update method of the LSTM's internal state needs to be adjusted. This is achieved by modifying the bias weight coefficients and input weight coefficients in the LSTM, improving the matching degree between the predicted value and the true value. Focal Loss is used as the loss function to alleviate the class imbalance problem. The optimizer is NAdam, with an initial learning rate preset to 0.001. An early stopping mechanism is implemented, terminating training when the validation set loss does not decrease for 5 consecutive rounds.

[0146] As can be seen from the above technical solution, this embodiment provides an optional method for obtaining a prediction model. This method allows for the training of an initial prediction model using training data with varying degrees of anomalies. The initial prediction model includes a dynamically enhanced forgetting layer and a Transformer encoding layer. Therefore, it retains the ability of LSTM to capture temporal dependencies while leveraging the global modeling advantages of Transformer and the dynamically enhanced forgetting layer to improve feature selection accuracy and enhance the prediction reliability of the prediction model.

[0147] In some embodiments of this application, the process of converting the forgetting layer of the LSTM model into a dynamically enhanced forgetting layer in step S32 is described in detail, and the steps are as follows:

[0148] S320. Replace the output function of the forgetting layer of the LSTM model with a dynamic output function set to convert the forgetting layer of the LSTM model into a dynamically enhanced forgetting layer.

[0149] Specifically, the dynamic output function group is as follows:

[0150]

[0151]

[0152]

[0153] In the formula, To dynamically enhance the output of the forgetting layer; This is the output of the forgetting layer; This is a dynamic adjustment coefficient; This is the dynamic forgetting weight matrix, used to map the error vector to the forgetting gate adjustment space; Let Re be the historical prediction error vector, and Re be the modified linear unit activation function. Forgetting weight matrix, The table shows the hidden state at time i-1; This is the input at time i; It is the bias vector; P is the output function of the forgetting layer. i P represents the node power at time i; i-1 Let P represent the node power at time i-1; P is the node's rated power.

[0154] As can be seen from the above technical solution, this embodiment provides an optional method to convert the forgetting layer of an LSTM model into a dynamically enhanced forgetting layer. This method dynamically adjusts the forgetting rate based on the power change rate and rated power, enabling the prediction model to quickly forget historical information during load abrupt changes, avoiding error accumulation and significantly improving prediction accuracy in abrupt change scenarios. Through the dynamic forgetting mechanism, hybrid architecture design, and data quality assurance, the accuracy, robustness, and real-time adaptability of distribution network operation status prediction are significantly improved.

[0155] In some embodiments of this application, the process of determining whether a transfer strategy needs to be generated based on the running status information and the running sequence data in step S3 is described in detail, and the steps are as follows:

[0156] S30. Determine the number of abnormal moments contained in the runtime sequence data.

[0157] Specifically, the running state value at each time point in the runtime sequence data can be calculated by combining the state calculation function; the average state value of each running state value can be calculated; when the absolute value of the difference between the running state value at the corresponding time point and the average state value is greater than a preset state threshold, the time point is regarded as an abnormal time point; and the total number of abnormal time points contained in the runtime sequence data is counted as the number of abnormal time points.

[0158] S31. Determine the total number of abnormal moments contained in the running status information.

[0159] Specifically, the state calculation function can be used to calculate the running state value at each time point in the running state information; calculate the average state value of each running state value in the running state information; when the absolute value of the difference between the running state value at the corresponding time point and the average state value is greater than a preset state threshold, the time point is regarded as an abnormal time point in the running state information; and count the total number of abnormal time points contained in the running state information as the total number of abnormal time points.

[0160] S32. Calculate the ratio of the total number of abnormal moments to the number of abnormal moments.

[0161] Specifically, the ratio between the total number of abnormal moments and the number of abnormal moments can be calculated as the abnormality ratio.

[0162] S33. When the abnormal ratio is greater than the preset ratio threshold, it is determined that a transfer strategy needs to be generated.

[0163] Specifically, the ratio threshold can be determined based on the supply and demand situation of multiple consecutive historical cycles.

[0164] The abnormal ratio can be compared with the ratio threshold;

[0165] If the abnormal ratio is greater than the preset ratio threshold, it can be determined that a transfer strategy needs to be generated.

[0166] S34. When the abnormal ratio is not greater than the preset ratio threshold, it is determined that no transfer strategy needs to be generated.

[0167] Specifically, if the abnormal ratio is not greater than the preset ratio threshold, it can be determined that no transfer strategy needs to be generated.

[0168] As can be seen from the above technical solution, this embodiment provides an optional method for determining whether a transfer strategy needs to be generated based on the operating status information and the operating sequence data. Through this method, the ratio of the total number of abnormal moments in the next cycle to the number of abnormal moments in the previous cycle can be calculated to assess whether the operating situation in the next cycle has improved and whether the abnormal operating conditions meet expectations. This allows for an assessment of whether a transfer strategy needs to be generated, reducing the excessive generation of transfer strategies and avoiding unnecessary resource consumption.

[0169] In some embodiments of this application, the process of determining the total supply period using the prediction model in step S3 is described in detail, and the steps are as follows:

[0170] S30. The period corresponding to the operating status information is taken as the total time period for power transfer.

[0171] Specifically, the total supply period can be used as a variable to initialize the total supply period, and the cycle corresponding to the operating status information can be written into the total supply period.

[0172] S31. Using the prediction model, based on the operating status information, predict the state time sequence information of the next cycle of the operating status information.

[0173] Specifically, the latest operating status information can be input into the prediction model to obtain the state time series information for the next cycle of the latest operating status information.

[0174] S32. Determine whether the state timing information corresponds to an abnormal cycle; if yes, proceed to step S33; if no, proceed to step S34.

[0175] Specifically, step S33 can be executed when the ratio of the number of abnormal moments contained in the state timing information to the number of normal moments contained in the state timing information is not less than the abnormal threshold.

[0176] When the ratio of the number of abnormal moments in the state timing information to the number of normal moments in the state timing information is less than the abnormal threshold, step S34 is executed.

[0177] S33. Add the abnormal period corresponding to the status timing information after the total transfer time period, and update the total transfer time period; use the status timing information as the new running status information, and return to the execution step S31.

[0178] Specifically, when it is determined that the status timing information corresponds to an abnormal period, the abnormal period of the status timing information can be added to the total transfer time period to complete the update of the total transfer time period.

[0179] Subsequently, the status timing information can be used as the new running status information, and the process can return to step S31.

[0180] S34. Determine the final total transfer period.

[0181] Specifically, when the timing information of the determined state does not correspond to an abnormal period, the latest total transfer period can be determined as the final total transfer period.

[0182] As can be seen from the above technical solution, this embodiment provides an optional method for determining the total supply transfer period using the aforementioned prediction model. Through this method, the total supply transfer period can be determined by iteratively predicting and integrating abnormal cycles.

[0183] Next, we will combine Figure 2 The distribution network load transfer strategy generation device provided in this application is described in detail. The distribution network load transfer strategy generation device described below can be compared with the distribution network load transfer strategy generation method described above.

[0184] See Figure 2 It can be observed that the distribution network load transfer strategy generation device may include:

[0185] The acquisition module 10 is used to acquire the runtime sequence data of the power distribution network;

[0186] The judgment module 20 is used to determine whether the runtime sequence data indicates that the power distribution network is in an abnormal cycle; if so, the generation module 30 is called.

[0187] The generation module 30 is used to acquire a prediction model and use the prediction model to predict the operating status information of the next cycle based on the runtime sequence data; based on the operating status information and the runtime sequence data, it determines whether a transfer strategy needs to be generated; if so, it uses the prediction model to determine the total transfer period; with the goal of maximizing the total restored load and minimizing the number of switching operations, it establishes a multi-objective transfer strategy model based on the total transfer period; it solves the multi-objective transfer strategy model to generate the transfer strategy corresponding to the total transfer period.

[0188] Furthermore, the judgment module 20 may include:

[0189] The normal time quantity determination unit is used to determine the number of abnormal times and the number of normal times contained in the runtime sequence data;

[0190] A ratio calculation unit is used to calculate the ratio of the number of abnormal moments to the number of normal moments;

[0191] A ratio comparison unit is used to determine that the power distribution network is in an abnormal cycle when the ratio is not less than an abnormal threshold.

[0192] An abnormal threshold comparison unit is used to determine that the distribution network is not in an abnormal cycle when the ratio is less than an abnormal threshold.

[0193] Furthermore, the normal time quantity determination unit may include:

[0194] The first normal time quantity determination subunit is used to calculate the operating status value at different time points based on the voltage deviation value at each time point, the temperature value and temperature limit of the distribution network at different time points;

[0195] The second normal time quantity determination subunit is used to calculate the average state value of each operating state value;

[0196] The third normal time quantity determination subunit is used to determine the abnormal time point when the absolute value of the difference between the running state value and the average state value at the corresponding time point is greater than the preset state threshold.

[0197] The fourth normal time point quantity determination subunit is used to determine the normal time point when the absolute value of the difference between the running state value and the average state value at the corresponding time point is not greater than a preset state threshold.

[0198] The fifth normal time quantity determination subunit is used to count the total number of all abnormal time points as the abnormal time quantity.

[0199] The sixth normal time quantity determination subunit is used to count the total number of all normal time points as the normal time quantity.

[0200] Furthermore, the generation module 30 may include:

[0201] The training run data acquisition unit is used to acquire training run data corresponding to multiple consecutive historical periods. Each historical period corresponds to two types of periods: one type of historical period is an abnormal historical period, and the other type of historical period is a normal historical period.

[0202] LSTM model building unit, used to build LSTM models;

[0203] The initial prediction model generation unit is used to add a Transformer encoding layer to the LSTM model and convert the forgetting layer of the LSTM model into a dynamically enhanced forgetting layer to obtain the initial prediction model.

[0204] The initial prediction model training unit is used to train the initial prediction model using various training run data, adjust the parameters of the initial prediction model, until the initial prediction model meets the preset stopping condition; the final initial prediction model is the prediction model.

[0205] Furthermore, the initial prediction model generation unit may include:

[0206] The output function replacement subunit is used to replace the output function of the forgetting layer of the LSTM model with a dynamic output function set, so as to convert the forgetting layer of the LSTM model into a dynamically enhanced forgetting layer.

[0207] The dynamic output function set is as follows:

[0208]

[0209]

[0210]

[0211] In the formula, To dynamically enhance the output of the forgetting layer; This is the output of the forgetting layer; This is a dynamic adjustment coefficient; This is a dynamic forgetting weight matrix; Let Re be the historical prediction error vector, and Re be the modified linear unit activation function. Forgetting weight matrix, The table shows the hidden state at time i-1; This is the input at time i; It is the bias vector; P is the output function of the forgetting layer. i P represents the node power at time i; i-1Let P represent the node power at time i-1; P is the node's rated power.

[0212] Furthermore, the generation module 30 may include:

[0213] The first generation unit is used to determine the number of abnormal moments contained in the runtime sequence data;

[0214] The second generation unit is used to determine the total number of abnormal moments contained in the running status information;

[0215] The third generation unit is used to calculate the abnormal ratio of the total number of abnormal moments to the number of abnormal moments.

[0216] The fourth generation unit is used to determine that a transfer strategy needs to be generated when the abnormal ratio is greater than a preset ratio threshold.

[0217] The fifth generation unit is used to determine that no transfer strategy needs to be generated when the abnormal ratio is not greater than a preset ratio threshold.

[0218] Furthermore, the generation module 30 may include:

[0219] The total supply time period update unit is used to take the period corresponding to the operating status information as the total supply time period;

[0220] A state timing information prediction unit is used to predict the state timing information of the next cycle of the operating state information based on the operating state information using the prediction model.

[0221] The status timing information judgment unit is used to determine whether the status timing information corresponds to an abnormal period; if yes, it calls the running status information update unit; if no, it calls the total supply period determination unit.

[0222] The operation status information update unit is used to add the abnormal period corresponding to the status time sequence information after the total transfer period, update the total transfer period, and use the status time sequence information as the new operation status information to return to the execution step of using the prediction model to predict the status time sequence information of the next period of the operation status information based on the operation status information.

[0223] The total transfer time period determination unit is used to determine the final total transfer time period.

[0224] The distribution network load transfer strategy generation device provided in this application embodiment can be applied to distribution network load transfer strategy generation equipment, such as PC terminals, cloud platforms, servers, and server clusters. Optionally, Figure 3 The hardware structure block diagram of the distribution network load transfer strategy generation device is shown, with reference to... Figure 3The hardware structure of the power distribution network load transfer strategy generation device may include: at least one processor 1, at least one communication interface 2, at least one memory 3 and at least one communication bus 4;

[0225] In this embodiment of the application, the number of processor 1, communication interface 2, memory 3, and communication bus 4 is at least one, and processor 1, communication interface 2, and memory 3 communicate with each other through communication bus 4;

[0226] Processor 1 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention.

[0227] Memory 3 may include high-speed RAM, and may also include non-volatile memory, such as at least one disk storage device;

[0228] The memory stores a program, which the processor can call. The program is used for:

[0229] Obtain runtime sequence data of the power distribution network;

[0230] Determine whether the runtime sequence data indicates that the power distribution network is in an abnormal cycle;

[0231] If so, a prediction model is obtained, and the prediction model is used to predict the operating status information of the next cycle based on the runtime sequence data; based on the operating status information and the runtime sequence data, it is determined whether a transfer strategy needs to be generated; if so, the prediction model is used to determine the total transfer period; with the goal of maximizing the total restored load and minimizing the number of switching operations, a multi-objective transfer strategy model is established based on the total transfer period; the multi-objective transfer strategy model is solved to generate the transfer strategy corresponding to the total transfer period.

[0232] Optionally, the refined and extended functions of the program can be referred to the above description.

[0233] This application embodiment also provides a readable storage medium that can store a program suitable for execution by a processor, the program being used for:

[0234] Obtain runtime sequence data of the power distribution network;

[0235] Determine whether the runtime sequence data indicates that the power distribution network is in an abnormal cycle;

[0236] If so, a prediction model is obtained, and the prediction model is used to predict the operating status information of the next cycle based on the runtime sequence data; based on the operating status information and the runtime sequence data, it is determined whether a transfer strategy needs to be generated; if so, the prediction model is used to determine the total transfer period; with the goal of maximizing the total restored load and minimizing the number of switching operations, a multi-objective transfer strategy model is established based on the total transfer period; the multi-objective transfer strategy model is solved to generate the transfer strategy corresponding to the total transfer period.

[0237] Optionally, the refined and extended functions of the program can be referred to the above description.

[0238] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0239] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0240] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. The various embodiments of this application can be combined with each other. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for generating a load transfer strategy for a distribution network, characterized in that, include: Obtain runtime sequence data of the power distribution network; Determine whether the runtime sequence data indicates that the power distribution network is in an abnormal cycle; If so, then obtain the prediction model and use the prediction model to predict the running status information of the next cycle based on the running sequence data; Based on the running status information and the running sequence data, determine whether a transfer strategy needs to be generated; If necessary, the period corresponding to the operating status information shall be used as the total supply period; Using the prediction model, based on the operating status information, predict the state time series information of the next cycle of the operating status information; Determine whether the state timing information corresponds to an abnormal cycle; If so, add the abnormal period corresponding to the status timing information after the total transfer period, and update the total transfer period; The state time series information is used as new running state information, and the process is returned to execute the step of using the prediction model to predict the state time series information of the next cycle of the running state information based on the running state information. If not, then determine the final total transfer period; With the goal of maximizing the total restored load and minimizing the number of switching operations, a multi-objective power transfer strategy model is established based on the total power transfer period. Solve the multi-objective transfer strategy model to generate the transfer strategy corresponding to the total transfer time period; The acquisition of the prediction model includes: Acquire training data corresponding to multiple consecutive historical periods. Each historical period corresponds to two types of periods: one type is abnormal historical periods, and the other type is normal historical periods. Construct an LSTM model; A Transformer encoding layer is added to the LSTM model, and the forgetting layer of the LSTM model is converted into a dynamically enhanced forgetting layer to obtain the initial prediction model. The initial prediction model is trained using various training data, and the parameters of the initial prediction model are adjusted until the initial prediction model meets the preset stopping conditions; the final initial prediction model is the prediction model.

2. The method for generating a power distribution network load transfer strategy according to claim 1, characterized in that, The step of determining whether the runtime sequence data indicates that the distribution network is in an abnormal cycle includes: Determine the number of abnormal moments and the number of normal moments contained in the runtime sequence data; Calculate the ratio of the number of abnormal moments to the number of normal moments; When the ratio is not less than the abnormal threshold, the power distribution network is determined to be in an abnormal cycle; When the ratio is less than the abnormal threshold, it is determined that the power distribution network is not in an abnormal cycle.

3. The method for generating a power distribution network load transfer strategy according to claim 2, characterized in that, The runtime sequence data includes voltage deviation values ​​at multiple time points; Determining the number of abnormal moments and the number of normal moments contained in the runtime sequence data includes: Based on the voltage deviation value at each time point, the temperature value and temperature limit of the distribution network at different time points are used to calculate the operating status value at different time points; Calculate the average state value for each operating state value; When the absolute value of the difference between the running status value at a corresponding time point and the average status value is greater than a preset status threshold, that time point is regarded as an abnormal time point. When the absolute value of the difference between the running status value at the corresponding time point and the average status value is not greater than the preset status threshold, the time point is regarded as the normal time point. The total number of all abnormal time points is counted as the number of abnormal time points; The total number of all normal time points is counted as the number of normal time points.

4. The method for generating a power distribution network load transfer strategy according to claim 1, characterized in that, The process of converting the forgetting layer of the LSTM model into a dynamically enhanced forgetting layer includes: The forgetting layer output function of the LSTM model is replaced with a dynamic set of output functions to convert the forgetting layer of the LSTM model into a dynamically enhanced forgetting layer. The dynamic output function set is as follows: ； ； ； In the formula, To dynamically enhance the output of the forgetting layer; This is the output of the forgetting layer; This is a dynamic adjustment coefficient; This is a dynamic forgetting weight matrix; Let Re be the historical prediction error vector, and Re be the modified linear unit activation function. Forgetting weight matrix, The hidden state at time i-1; This is the input at time i; It is the bias vector; P is the output function of the forgetting layer. i P represents the node power at time i; i-1 Let P represent the node power at time i-1; P is the node's rated power.

5. The method for generating a power distribution network load transfer strategy according to claim 1, characterized in that, The step of determining whether a transfer strategy needs to be generated based on the running status information and the running sequence data includes: Determine the number of abnormal moments contained in the runtime sequence data; Determine the total number of abnormal moments contained in the operational status information; Calculate the ratio of the total number of abnormal moments to the number of abnormal moments; When the abnormal ratio is greater than a preset ratio threshold, it is determined that a transfer strategy needs to be generated. If the abnormal ratio is not greater than a preset ratio threshold, it is determined that no transfer strategy needs to be generated.

6. A distribution network load transfer strategy generation device, characterized in that, include: The acquisition module is used to acquire runtime sequence data of the power distribution network; The judgment module is used to determine whether the runtime sequence data indicates that the power distribution network is in an abnormal cycle; If so, then call the generation module; A generation module is used to obtain a prediction model and use the prediction model to predict the running status information of the next cycle based on the runtime sequence data; Based on the running status information and the running sequence data, determine whether a transfer strategy needs to be generated; If necessary, the period corresponding to the operating status information shall be used as the total supply period; Using the prediction model, based on the operating status information, predict the state time series information of the next cycle of the operating status information; Determine whether the state timing information corresponds to an abnormal cycle; If so, add the abnormal period corresponding to the status timing information after the total transfer period, and update the total transfer period; The state time series information is used as new running state information, and the process is returned to execute the step of using the prediction model to predict the state time series information of the next cycle of the running state information based on the running state information. If not, then determine the final total transfer period; With the goal of maximizing the total restored load and minimizing the number of switching operations, a multi-objective power transfer strategy model is established based on the total power transfer period. Solve the multi-objective transfer strategy model to generate the transfer strategy corresponding to the total transfer time period; The generation module includes: The training run data acquisition unit is used to acquire training run data corresponding to multiple consecutive historical periods. Each historical period corresponds to two types of periods: one type of historical period is an abnormal historical period, and the other type of historical period is a normal historical period. LSTM model building unit, used to build LSTM models; The initial prediction model generation unit is used to add a Transformer encoding layer to the LSTM model and convert the forgetting layer of the LSTM model into a dynamically enhanced forgetting layer to obtain the initial prediction model. The initial prediction model training unit is used to train the initial prediction model using various training run data, adjust the parameters of the initial prediction model, until the initial prediction model meets the preset stopping condition; the final initial prediction model is the prediction model.

7. A power distribution network load transfer strategy generation device, characterized in that, Including memory and processor; The memory is used to store programs; The processor is used to execute the program to implement each step of the distribution network load transfer strategy generation method as described in any one of claims 1-5.

8. A readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements each step of the distribution network load transfer strategy generation method as described in any one of claims 1-5.

Images