Circuit breaker control system and method

Abstract

The invention relates to a circuit breaker control system and method. The method comprises the following steps: a data acquisition module acquires predicted closing time, operation parameters and actual closing time, and uploads data to a cloud computing module and a mechanism control module; judging whether a prediction strategy needs to be adjusted currently; if not, the mechanism control module responds and predicts the closing time of the next closing action according to the current operation parameters; if yes, the cloud computing module trains a prediction model; and based on the trained prediction model, outputting prediction data of each node, and sending the prediction data to a mechanism control module for updating. The mechanism control module considers the real-time operation parameters to predict the closing time of the next closing action, can adapt to the current actual working condition to adjust the closing action time, improves the stability of the action of the circuit breaker, and further improves the accuracy and reliability of the closing action. The cloud computing module optimizes the prediction mechanism of the mechanism control module through model training, and improves the accuracy of the switch-on time prediction.

Description

Technical Field

[0001] The present application relates to the technical field of electrical equipment, and in particular to a circuit breaker control system and method. Background Art

[0002] High-voltage circuit breakers are the most important switching devices in power systems. They can open or close normal lines, and can also open or close faulty lines. In actual engineering applications, the action time of circuit breakers will be affected by various external factors such as temperature, operating voltage, etc.

[0003] Phase-controlled switch technology requires that the circuit breaker switch opening and closing speed is as fast as possible and the dispersion of the operation time is small enough. At present, the circuit breaker control method has poor flexibility, the operating mechanism works in a harsh environment, and its action time stability is poor when factors such as external temperature and voltage change. Therefore, the stability of circuit breaker action time is still a key issue in the field of phase-controlled switches.

[0004] At present, the benchmark is generally selected through experiments, and the influencing factors of external variables are replaced one by one to calculate the impact of external variables on the circuit breaker action time. While ensuring that other external variables remain unchanged, change the value of a certain external variable, record the circuit breaker action time and calculate its deviation from the benchmark. Using the interval interpolation method, the relationship curve between the circuit breaker closing time and each variable is drawn according to the experimental data, which is used as a reference for the circuit breaker switch. However, due to the discrepancy between the experimental conditions and the actual working conditions, this method cannot take into account the impact of multi-variable coupling factors on the action time. Especially for circuit breakers that work for a long time, changes in the external environment have a greater impact on the action characteristics of the circuit breaker. The compensation curve obtained according to the experiment will gradually deviate from the normal situation, resulting in low accuracy of circuit breaker control and difficulty in achieving compensation effect. Summary of the invention

[0005] Based on this, it is necessary to provide a circuit breaker control system and method to address the above technical problems, which can adapt to the current actual working conditions to adjust the closing action time, improve the stability of the circuit breaker action time, and thereby improve the accuracy and reliability of the closing action.

[0006] In a first aspect, the present application provides a circuit breaker control system, the system comprising a data acquisition module, a mechanism control module, a cloud computing module and a circuit breaker; wherein:

[0007] The data acquisition module is used to collect the predicted closing time, operating parameters and actual closing time each time the circuit breaker performs a closing action; upload the collected predicted closing time, operating parameters and actual closing time to the cloud computing module; and send the collected operating parameters to the mechanism control module;

[0008] The data acquisition module is also used to determine whether the prediction strategy needs to be adjusted at present according to the predicted closing time and the actual closing time of the historical N closing actions; if not, a first signal is sent to the mechanism control module; if so, a second signal is sent to the cloud computing module;

[0009] A mechanism control module, configured to respond to the first signal and predict the closing time of the next closing action according to the current operating parameters and the prediction data stored locally;

[0010] The cloud computing module is used to respond to the second signal, train the random forest neural network model according to the predicted closing time, operating parameters and actual closing time of each closing action, and obtain a trained prediction model; output the prediction data of each node based on the trained prediction model; send the prediction data of each node to the mechanism control module to instruct the mechanism control module to update the locally stored prediction data; wherein each node corresponds to a different value of the operating parameter.

[0011] In one embodiment, the cloud computing module is further used to:

[0012] During the t-th iteration training process, the internal parameters of the random forest neural network model are dynamically optimized according to the whale optimization algorithm;

[0013] According to the predicted closing time, operating parameters and actual closing time of each closing action, the random forest neural network model is trained to obtain the intermediate training results;

[0014] The model parameters of the whale optimization algorithm and the random forest neural network model are adjusted according to the intermediate training results, and the t+1th iteration training is continued until the current number of iterations reaches the maximum number of iterations or the prediction accuracy of the random forest neural network model meets the preset requirements, and a trained prediction model is obtained.

[0015] In one embodiment, the cloud computing module is further used to evaluate the prediction accuracy of the random forest neural network model through the coefficient of determination, mean absolute error, mean absolute percentage error, and square error.

[0016] In one embodiment, the cloud computing module is further used to:

[0017] During the t-th iteration training process, several candidate solutions are updated according to the position update strategy;

[0018] Calculate the fitness value of each candidate solution;

[0019] The candidate solution with the highest fitness value is selected as the target solution, and the internal parameters of the random forest neural network model are optimized based on the target solution.

[0020] In one embodiment, the cloud computing module is further used to:

[0021] For each node, multiple reference nodes of the current node in a target direction are determined; the target direction is any one of multiple preset directions;

[0022] Using the trained prediction model, the closing time of the current node and multiple reference nodes is predicted in turn to obtain multiple prediction results;

[0023] Based on multiple prediction results, the partial derivative value of the current node in the target direction is calculated using a preset formula;

[0024] The closing time of the current node and the partial derivative values ​​of the current node in multiple preset directions are used as the prediction data of the current node;

[0025] The preset formula is:

[0026] ;

[0027] In the formula, S is the partial derivative value of the current node in the target direction; is the closing time predicted by the prediction model for the i-th run; is the closing time predicted by the prediction model for the i+1th run; is the operating parameter input to the prediction model for the i-th time; is the operating parameter input to the prediction model for the i+1th time; n is the total number of prediction model runs.

[0028] In one embodiment, the mechanism control module is further used to:

[0029] According to the current operating parameters, multiple target nodes around the node to be predicted are determined; the current operating parameters correspond to the node to be predicted;

[0030] Obtain the closing time of each target node and the partial derivative value of each target node in each preset direction from the locally stored prediction data;

[0031] An interpolation calculation is performed based on the closing time of each target node and the partial derivative value of each target node in each preset direction to determine the closing time corresponding to the node to be predicted as the closing time of the next closing action.

[0032] In one of the embodiments, the mechanism control module is also used to perform Hermite interpolation calculation according to the closing time of each target node and the partial derivative value of each target node in each preset direction, and determine the closing time corresponding to the node to be predicted as the closing time of the next closing action.

[0033] In a second aspect, the present application further provides a circuit breaker control method, which is applied to the circuit breaker control system as described in the first aspect above; the method comprises:

[0034] The data acquisition module collects the predicted closing time, operating parameters and actual closing time each time the circuit breaker performs a closing action; uploads the collected predicted closing time, operating parameters and actual closing time to the cloud computing module; and sends the collected operating parameters to the mechanism control module;

[0035] The data acquisition module determines whether the prediction strategy needs to be adjusted based on the predicted closing time and the actual closing time of N historical closing actions; if not, a first signal is sent to the mechanism control module; if so, a second signal is sent to the cloud computing module;

[0036] The mechanism control module responds to the first signal and predicts the closing time of the next closing action according to the current operating parameters and the prediction data stored locally;

[0037] The cloud computing module responds to the second signal, and trains the random forest neural network model according to the predicted closing time, operating parameters and actual closing time of each closing action to obtain a trained prediction model; based on the trained prediction model, the prediction data of each node is output; the prediction data of each node is sent to the mechanism control module to instruct the mechanism control module to update the locally stored prediction data; wherein each node corresponds to a different value of the operating parameter.

[0038] In one embodiment, a random forest neural network model is trained according to the predicted closing time, operating parameters and actual closing time of each closing action to obtain a trained prediction model, including:

[0039] During the t-th iteration training process, the internal parameters of the random forest neural network model are dynamically optimized according to the whale optimization algorithm;

[0040] According to the predicted closing time, operating parameters and actual closing time of each closing action, the random forest neural network model is trained to obtain the intermediate training results;

[0041] The model parameters of the whale optimization algorithm and the random forest neural network model are adjusted according to the intermediate training results, and the t+1th iteration training is continued until the current number of iterations reaches the maximum number of iterations or the prediction accuracy of the random forest neural network model meets the preset requirements, and a trained prediction model is obtained.

[0042] In one embodiment, the method further comprises:

[0043] The prediction accuracy of the random forest neural network model was evaluated by the coefficient of determination, mean absolute error, mean absolute percentage error, and squared error.

[0044] In the above-mentioned circuit breaker control system and method, the data acquisition module collects the predicted closing time, operating parameters and actual closing time each time the circuit breaker performs a closing action; uploads the collected predicted closing time, operating parameters and actual closing time to the cloud computing module; and sends the collected operating parameters to the mechanism control module; the data acquisition module determines whether it is necessary to adjust the prediction strategy at present based on the predicted closing time and actual closing time of N historical closing actions; if not, sends a first signal to the mechanism control module; if so, sends a second signal to the cloud computing module; the mechanism control module responds to the first signal and predicts the closing time of the next closing action based on the current operating parameters and the locally stored prediction data; the cloud computing module responds to the second signal and trains the random forest neural network model based on the predicted closing time, operating parameters and actual closing time of each closing action to obtain a trained prediction model; based on the trained prediction model, outputs the prediction data of each node; sends the prediction data of each node to the mechanism control module to instruct the mechanism control module to update the locally stored prediction data; wherein each node corresponds to a different value of the operating parameter. Through the above method, the mechanism control module considers the operating parameters collected in real time by the data acquisition module to predict the closing time of the next closing action, and can adjust the closing action time to adapt to the current actual working conditions, reduce the inconsistency or volatility of the action time of the circuit breaker during the closing process, and improve the stability of the circuit breaker action, thereby improving the accuracy and reliability of the closing action. In the case of unreliable prediction strategies, the cloud computing module performs model training based on the data collected multiple times by the data acquisition module, optimizes the prediction mechanism of the mechanism control module, further reduces the dispersion of the circuit breaker closing action time, and improves the accuracy of the closing time prediction. By decoupling model training from closing control, it is possible to avoid the impact of model training on the stability of closing control, thereby improving the flexibility, scalability and stability of the circuit breaker control system. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 A schematic diagram of the structure of a circuit breaker control system in one embodiment;

[0047] Figure 2 is a schematic diagram of node positions in an embodiment;

[0048] Figure 3 A schematic flow chart of a circuit breaker control method in one embodiment;

[0049] Figure 4 FIG. 4 is a flow chart of a circuit breaker control method in another embodiment. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0051] In an exemplary embodiment, Figure 1 As shown, a circuit breaker control system is provided, the system comprising a data acquisition module, a mechanism control module, a cloud computing module and a circuit breaker; wherein:

[0052] The data acquisition module is used to collect the predicted closing time, operating parameters and actual closing time each time the circuit breaker performs a closing action; upload the collected predicted closing time, operating parameters and actual closing time to the cloud computing module; and send the collected operating parameters to the mechanism control module.

[0053] It can be understood that the closing time refers to the time interval from the moment the circuit breaker receives the closing command to the moment the closing action is completed. The operating parameters are external parameters that affect the operation of the circuit breaker, including electrical parameters and environmental parameters. Optionally, the data acquisition module includes a temperature sensor, an oil pressure sensor, and a voltage transformer, and the collected operating parameters include temperature, oil pressure, and voltage.

[0054] The data acquisition module monitors the size of external operating parameters in real time when the closing command is issued, and records the actual closing time of each closing action. The data acquisition module determines the predicted closing time of each closing command through interaction with the mechanism control module. The collected predicted closing time, operating parameters and actual closing time are transmitted to the cloud computing module, and the collected current operating parameters are sent to the mechanism control module.

[0055] The data acquisition module is also used to determine whether the prediction strategy needs to be adjusted at present based on the predicted closing time and actual closing time of N historical closing actions; if not, a first signal is sent to the mechanism control module; if so, a second signal is sent to the cloud computing module.

[0056] Among them, N is an integer greater than 1, which is a parameter set in advance according to actual needs. Specifically, the data acquisition module regularly determines whether the historical N closing time predictions are reliable, so as to determine whether the prediction strategy needs to be adjusted at present. If the historical N closing time predictions are reliable, the prediction data stored locally in the mechanism control module is kept unchanged; if the historical N closing time predictions are unreliable, the cloud computing module optimizes the prediction data and updates it to the mechanism control module. Exemplarily, N=10, based on the predicted closing time and actual closing time of the historical 10 closing actions, the error average of the historical 10 closing actions is determined. If the error average is greater than the set value (for example, 0.5ms), it is determined that the prediction strategy needs to be adjusted at present; if the error average is less than or equal to the set value, it is determined that the prediction strategy does not need to be adjusted at present.

[0057] The first signal is used to instruct the mechanism control module to perform subsequent operations, and the second signal is used to instruct the cloud computing module to perform subsequent operations, wherein "first" and "second" are only used to distinguish each other and do not constitute a limitation on the order or type of the signals.

[0058] The mechanism control module is used to respond to the first signal and predict the closing time of the next closing action based on the current operating parameters and the prediction data stored locally.

[0059] Among them, the prediction data stored locally in the mechanism control module can be presented in the form of a compensation correction database, and the compensation correction database stores prediction data corresponding to several nodes. Optionally, the mechanism control module responds to the first signal, determines the node corresponding to the current operating parameter, queries the compensation correction database to determine the corresponding prediction data, and predicts the closing time of the next closing action based on the prediction data. Specifically, when the mechanism control module receives the closing instruction next time, it performs closing control in combination with the predicted closing time.

[0060] The cloud computing module is used to respond to the second signal, train the random forest neural network model according to the predicted closing time, operating parameters and actual closing time of each closing action, and obtain a trained prediction model; output the prediction data of each node based on the trained prediction model; send the prediction data of each node to the mechanism control module to instruct the mechanism control module to update the locally stored prediction data; wherein each node corresponds to a different value of the operating parameter.

[0061] Among them, the cloud computing module uses historical data, that is, the predicted closing time, operating parameters and actual closing time of each closing action uploaded by the data acquisition module, to perform big data training and build a closing action time prediction model affected by operating parameters. Specifically, the cloud computing module performs big data training based on the random forest algorithm to generate a trained prediction model. The random forest neural network is a multi-layer forward neural network. The model randomly samples from the original data set to form n different sample data sets, builds n different decision tree models based on these data sets, and obtains the final result based on the average value of these decision tree models. Based on the trained prediction model, the closing time of each node in the compensation correction database is predicted, the prediction data of each node is output, and the compensation correction database is established and sent to the mechanism control module. The mechanism control module updates the locally stored prediction data, and then performs closing control based on the updated prediction data.

[0062] It is understandable that since the training of neural networks and the establishment of databases require a large amount of computing resources, deploying neural networks on the mechanism control side will affect the circuit breaker control process and make it difficult to deploy in actual working conditions. In the above-mentioned circuit breaker control system, the data acquisition module collects the predicted closing time, operating parameters and actual closing time each time the circuit breaker performs a closing action; uploads the collected predicted closing time, operating parameters and actual closing time to the cloud computing module; and sends the collected operating parameters to the mechanism control module; the data acquisition module determines whether it is necessary to adjust the prediction strategy at present based on the predicted closing time and actual closing time of N historical closing actions; if not, sends a first signal to the mechanism control module; if so, sends a second signal to the cloud computing module; the mechanism control module responds to the first signal and predicts the closing time of the next closing action based on the current operating parameters and the locally stored prediction data; the cloud computing module responds to the second signal and trains the random forest neural network model based on the predicted closing time, operating parameters and actual closing time of each closing action to obtain a trained prediction model; based on the trained prediction model, outputs the prediction data of each node; sends the prediction data of each node to the mechanism control module to instruct the mechanism control module to update the locally stored prediction data; wherein each node corresponds to a different value of the operating parameter. Through the above method, the mechanism control module considers the operating parameters collected in real time by the data acquisition module to predict the closing time of the next closing action, and can adjust the closing action time to adapt to the current actual working conditions, reduce the inconsistency or volatility of the action time of the circuit breaker during the closing process, and improve the stability of the circuit breaker action, thereby improving the accuracy and reliability of the closing action. In the case of unreliable prediction strategies, the cloud computing module performs model training based on the data collected multiple times by the data acquisition module, optimizes the prediction mechanism of the mechanism control module, further reduces the dispersion of the circuit breaker closing action time, and improves the accuracy of the closing time prediction. By decoupling model training from closing control, it is possible to avoid the impact of model training on the stability of closing control, thereby improving the flexibility, scalability and stability of the circuit breaker control system.

[0063] In an exemplary embodiment, the cloud computing module is also used to: dynamically optimize the internal parameters of the random forest neural network model according to the whale optimization algorithm during the t-th iterative training process; train the random forest neural network model according to the predicted closing time, operating parameters and actual closing time of each closing action to obtain intermediate training results; adjust the model parameters of the whale optimization algorithm and the random forest neural network model according to the intermediate training results, and continue the t+1-th iterative training until the current number of iterations reaches the maximum number of iterations or the prediction accuracy of the random forest neural network model meets the preset requirements, thereby obtaining a trained prediction model.

[0064] The cloud computing module performs T (maximum number of iterations) iterations. In any iteration process, the internal parameters of the random forest neural network model are dynamically optimized according to the Whale Optimization Algorithm (WOA) to improve the algorithm performance. The internal parameters can be the number of decision trees, the number of cotyledons, the feature selection method, the growth method of the decision tree, etc., which are not limited in this embodiment.

[0065] In an optional implementation, this embodiment uses an improved whale optimization algorithm to dynamically optimize the internal parameters of the model. Specifically, in order to solve the problem of uneven initial distribution of the population, the improved whale optimization algorithm uses a strategy based on Circle chaotic mapping to initialize the population. Chaotic mapping is highly unpredictable and can generate complex random sequences. Chaotic mapping can be expressed as:

[0066] ;

[0067] Among them, a and b are control parameters, commonly used values ​​are 0.5 and 0.2, and mod is the remainder function.

[0068] In the optimization process of the whale optimization algorithm, the convergence factor a decreases linearly from 2 to 0 with the increase of the number of iterations, and the global exploration ability of the algorithm decreases linearly, which makes the algorithm insufficient in the early search and the later iteration speed is relatively slow. For this problem, an improved nonlinear adjustment strategy is proposed to optimize the impact of the convergence factor on the algorithm performance, which can be expressed by the following formula:

[0069] ;

[0070] Among them, t is the current iteration number and T is the total iteration number.

[0071] In order to further improve the convergence speed, the adaptive weight w is considered when updating the position. In the early stage of the operation, the influence of poor data can be eliminated more quickly. The adaptive weight is expressed as:

[0072] ;

[0073] Among them, t is the current iteration number and T is the total iteration number.

[0074] The location update strategy can be expressed as:

[0075] ;

[0076] in, is the optimal solution for the t-1th iteration; is the candidate solution after the position is updated; is the candidate solution before the position update; A and C are coefficients, A is related to the convergence factor a; is the adaptive weight.

[0077] In this embodiment, the random forest neural network model is optimized in combination with the whale optimization algorithm, and the random forest neural network model is trained using historical data to generate a trained prediction model. A compensation correction database is established based on the trained prediction model to improve the accuracy of closing time prediction.

[0078] In an exemplary embodiment, the cloud computing module is also used to evaluate the prediction accuracy of the random forest neural network model through the coefficient of determination, mean absolute error, mean absolute percentage error, and square error.

[0079] Among them, the predicted closing time of each node based on the random forest neural network model is compared with the actual historical closing time, and the prediction accuracy is evaluated by four indicators: determination coefficient (R2), mean absolute error (MAE), mean absolute percentage error (MAPE), and sum of squared error (SSE).

[0080] In an optional implementation, referring to Table 1, based on the same sample data, the improved WOA-RF algorithm adopted in this embodiment is compared with the traditional WOA-RF algorithm, the traditional RF (random forest) algorithm, and the ELM (Extreme Learning Machine) algorithm. In terms of the four indicators of determination coefficient (R2), mean absolute error (MAE), mean absolute percentage error (MAPE), and sum of square error (SSE), the improved WOA-RF algorithm in this embodiment is obviously superior to other algorithms.

[0081] Table 1:

[0082]

[0083] In an exemplary embodiment, the cloud computing module is also used to: update several candidate solutions according to the position update strategy during the t-th iterative training process; calculate the fitness value of each candidate solution; select the candidate solution with the highest fitness value as the target solution, and optimize the internal parameters of the random forest neural network model based on the target solution.

[0084] Each candidate solution corresponds to an optional internal parameter. During the initialization process, a set of multiple candidate solutions is randomly generated. During each iteration, the positions of several previous candidate solutions are updated based on the position update strategy, and the fitness value of each candidate solution is calculated to determine the optimal solution for each iteration, that is, the target solution. The internal parameters of the random forest neural network model are optimized based on the determined target solution.

[0085] In an optional implementation, an adaptive weight is set in the location update strategy, and the target weight increases as the number of current iterations increases; the adaptive weight can be expressed by the following formula:

[0086] ;

[0087] The location update strategy can adopt the rule of approaching the optimal solution of the previous iteration. This process can be expressed by the following formula:

[0088] ;

[0089] The position update strategy can also randomly explore the rules of new solutions in the search space. This process can be expressed by the following formula:

[0090] ;

[0091] in, is the optimal solution for the t-1th iteration; is a randomly selected solution; is the candidate solution after the position is updated; is the candidate solution before the position update; A and C are coefficients, A is related to the convergence factor a; is the adaptive weight.

[0092] In an exemplary embodiment, the cloud computing module is further used to: for each node, determine multiple reference nodes of the current node in a target direction; the target direction is any one of multiple preset directions; use the trained prediction model to predict the closing time of the current node and multiple reference nodes in turn to obtain multiple prediction results; based on the multiple prediction results, use a preset formula to calculate the partial derivative value of the current node in the target direction; use the closing time of the current node and the partial derivative values ​​of the current node in multiple preset directions as the prediction data of the current node;

[0093] The preset formula is:

[0094] ;

[0095] In the formula, S is the partial derivative value of the current node in the target direction; is the closing time predicted by the prediction model for the i-th run; is the closing time predicted by the prediction model for the i+1th run; is the operating parameter input to the prediction model for the i-th time; is the operating parameter input to the prediction model for the i+1th time; n is the total number of prediction model runs.

[0096] The partial derivative is calculated by the finite difference method. The preset directions are the four directions of the current node in the reference coordinate system. For each direction of each node, multiple reference nodes are selected, and the closing time of each node is predicted using the prediction model. Optionally, multiple reference nodes are determined by using the mode of random step change of the independent variable, combined with the above preset formula, and Random selection, but selection needs to be based on the following conditions: , ensuring that the step size changes within a reasonable range.

[0097] It can be understood that based on the partial derivative values ​​of each node in multiple preset directions, a compensation correction database is established and fed back to the mechanism control module.

[0098] In an exemplary embodiment, the mechanism control module is also used to: determine multiple target nodes around the node to be predicted based on current operating parameters; the current operating parameters correspond to the node to be predicted; obtain the closing time of each target node and the partial derivative value of each target node in each preset direction from the locally stored prediction data; perform interpolation calculation based on the closing time of each target node and the partial derivative value of each target node in each preset direction, and determine the closing time corresponding to the node to be predicted as the closing time of the next closing action.

[0099] It can be understood that the cloud computing module can perform interpolation calculations based on the partial derivative values ​​of each node in each preset direction, enrich the content of the compensation correction database, and provide the compensation correction database with a larger amount of information to the mechanism control module, so that the mechanism control module can query the closing time of the corresponding node for each operating parameter, thereby performing closing control.

[0100] Among them, refer to Figure 2 , Figure 2The example assumes that the closing time of the circuit breaker is only affected by voltage and temperature. The coordinate axes shown represent voltage and temperature. In the specific implementation, other coordinate axes can be added to represent additional external parameters according to different operating parameters. P is the node to be predicted, Q1, Q2, Q3, and Q4 are four target nodes selected at equal intervals, and the closing time stored in these four target nodes in the compensation correction database (predicted by the prediction model in the cloud). According to the partial derivative values ​​of Q1, Q2, Q1 in the direction of Q2, and the partial derivative values ​​of Q2 in the direction of Q1, the closing time of R1 is determined. Similarly, according to the partial derivative values ​​of Q3, Q4, Q3 in the direction of Q4, and the partial derivative values ​​of Q4 in the direction of Q3, the closing time of R2 is determined. The partial derivative value of R1 toward R2 is determined by weighted summation of the partial derivative value of Q1 toward Q3 and the partial derivative value of Q2 toward Q4; similarly, the partial derivative value of R2 toward R1 is determined by weighted summation of the partial derivative value of Q3 toward Q1 and the partial derivative value of Q4 toward Q2. The closing time of P is determined by interpolation calculation based on the partial derivative values ​​of R1, R2, R1 toward R2, and R2 toward R1, that is, the closing time of the next closing action is predicted.

[0101] In an exemplary embodiment, the mechanism control module is also used to perform Hermite interpolation calculation based on the closing time of each target node and the partial derivative value of each target node in each preset direction, and determine the closing time corresponding to the node to be predicted as the closing time of the next closing action.

[0102] Among them, combined Figure 2 , the partial derivative value of Q1 toward Q2 is recorded as . as well as Construct the Hermite equations and solve the predicted value at point R1, denoted as R1. Similarly, solve Q3 and Q4 to obtain the predicted value at point R2, denoted as R2. and Weighted ;Depend on and Weighted Finally, through R1, R2, , The Hermite equations are constructed to calculate the closing time of the node P to be predicted.

[0103] It can be understood that the mechanism control module predicts the current closing time through Hermite interpolation. Two-point cubic Hermite interpolation is a form of Hermite interpolation, which achieves a smooth interpolation process by defining a function value and its derivative at two data points.

[0104] Among them, the construction of the Hermite equations can be expressed by the following formula:

[0105] ;

[0106] in, , , , These are the four basis functions constructed in advance. Each basis function is a cubic algebraic polynomial and has the following base properties:

[0107] ;

[0108] ;

[0109] Determine the above cubic Hermite interpolation polynomial The required interpolation conditions are met.

[0110] To determine the basis functions , , , According to the properties of polynomial roots, yes The single root is and The root of yes The single root of and The repeated roots of , so the four basis functions can be set as:

[0111] ;

[0112] Among them, A, B, C, D, E, and F are all unknown constants. The values ​​of A, B, C, D, E, and F can be solved by using the property formula of the above basis, and the expressions of their functions are:

[0113] ;

[0114] The Hermite interpolation calculation can be completed by substituting the parameters of the two nodes that need to be interpolated into the aforementioned Hermite equations.

[0115] It should be noted that, taking a 550kv spring hydraulic operating mechanism circuit breaker with a rated closing time of 60ms as an example, the circuit breaker control system and method provided in the embodiment of the application are used for closing control. Compared with the traditional linear interpolation prediction method, the closing time deviation can be reduced by about 0.2ms.

[0116] Based on the same inventive concept, the embodiment of the present application also provides a circuit breaker control method applied to the circuit breaker control system involved above. The implementation scheme for solving the problem provided by the method is similar to the implementation scheme recorded in the above system, so the specific limitations in one or more circuit breaker control method embodiments provided below can refer to the limitations on the circuit breaker control system above, and will not be repeated here.

[0117] In an exemplary embodiment, in combination Figure 1 , refer to Figure 3 A circuit breaker control method is provided, which is applied to Figure 1 The circuit breaker control system shown in the figure is used as an example for explanation, including:

[0118] Step 302, the data acquisition module collects the predicted closing time, operating parameters and actual closing time each time the circuit breaker performs a closing action; uploads the collected predicted closing time, operating parameters and actual closing time to the cloud computing module; and sends the collected operating parameters to the mechanism control module.

[0119] Step 304, the data acquisition module determines whether the prediction strategy needs to be adjusted based on the predicted closing time and the actual closing time of N historical closing actions; if not, a first signal is sent to the mechanism control module; if so, a second signal is sent to the cloud computing module.

[0120] Step 306: The mechanism control module responds to the first signal and predicts the closing time of the next closing action according to the current operating parameters and the prediction data stored locally.

[0121] Step 308, the cloud computing module responds to the second signal, and trains the random forest neural network model according to the predicted closing time, operating parameters and actual closing time of each closing action to obtain a trained prediction model; based on the trained prediction model, the prediction data of each node is output; the prediction data of each node is sent to the mechanism control module to instruct the mechanism control module to update the locally stored prediction data; wherein each node corresponds to a different value of the operating parameter.

[0122] In the above circuit breaker control method, the data acquisition module collects the predicted closing time, operating parameters and actual closing time each time the circuit breaker performs a closing action; uploads the collected predicted closing time, operating parameters and actual closing time to the cloud computing module; and sends the collected operating parameters to the mechanism control module; the data acquisition module determines whether it is necessary to adjust the prediction strategy at present based on the predicted closing time and actual closing time of N historical closing actions; if not, sends a first signal to the mechanism control module; if so, sends a second signal to the cloud computing module; the mechanism control module responds to the first signal and predicts the closing time of the next closing action based on the current operating parameters and the locally stored prediction data; the cloud computing module responds to the second signal and trains the random forest neural network model based on the predicted closing time, operating parameters and actual closing time of each closing action to obtain a trained prediction model; based on the trained prediction model, outputs the prediction data of each node; sends the prediction data of each node to the mechanism control module to instruct the mechanism control module to update the locally stored prediction data; wherein each node corresponds to a different value of the operating parameter. Through the above method, the mechanism control module considers the operating parameters collected in real time by the data acquisition module to predict the closing time of the next closing action, and can adjust the closing action time to adapt to the current actual working conditions, reduce the inconsistency or volatility of the circuit breaker action time during the closing process, and improve the stability of the circuit breaker action, thereby improving the accuracy and reliability of the closing action. In the case of unreliable prediction strategies, the cloud computing module performs model training based on the data collected multiple times by the data acquisition module, optimizes the prediction mechanism of the mechanism control module, and further improves the accuracy of the closing time prediction. By decoupling model training from closing control, the impact of model training on the stability of closing control can be avoided, and the flexibility, scalability and stability of the circuit breaker control system can be improved.

[0123] In an exemplary embodiment, a random forest neural network model is trained according to the predicted closing time, operating parameters and actual closing time of each closing action to obtain a trained prediction model, including: in the t-th iterative training process, dynamically optimizing the internal parameters of the random forest neural network model according to the whale optimization algorithm; training the random forest neural network model according to the predicted closing time, operating parameters and actual closing time of each closing action to obtain intermediate training results; adjusting the model parameters of the whale optimization algorithm and the random forest neural network model according to the intermediate training results, and continuing the t+1-th iterative training until the current number of iterations reaches the maximum number of iterations or the prediction accuracy of the random forest neural network model meets the preset requirements, thereby obtaining a trained prediction model.

[0124] In an exemplary embodiment, the method further includes: evaluating the prediction accuracy of the random forest neural network model by means of the coefficient of determination, the mean absolute error, the mean absolute percentage error, and the square error.

[0125] In an exemplary embodiment, during the t-th iterative training process, the internal parameters of the random forest neural network model are dynamically optimized according to the whale optimization algorithm, including: during the t-th iterative training process, updating several candidate solutions according to the position update strategy; calculating the fitness value of each candidate solution; selecting the candidate solution with the highest fitness value as the target solution, and optimizing the internal parameters of the random forest neural network model based on the target solution.

[0126] In an exemplary embodiment, based on a trained prediction model, prediction data of each node is output, including: for each node, multiple reference nodes of the current node in a target direction are determined; the target direction is any one of multiple preset directions; using the trained prediction model, the closing time of the current node and multiple reference nodes are predicted in turn to obtain multiple prediction results; based on the multiple prediction results, the partial derivative value of the current node in the target direction is calculated using a preset formula; the closing time of the current node and the partial derivative values ​​of the current node in multiple preset directions are used as the prediction data of the current node;

[0127] The preset formula is:

[0128] ;

[0129] In the formula, S is the partial derivative value of the current node in the target direction; is the closing time predicted by the prediction model for the i-th run; is the closing time predicted by the prediction model for the i+1th run; is the operating parameter input to the prediction model for the i-th time; is the operating parameter input to the prediction model for the i+1th time; n is the total number of prediction model runs.

[0130] In an exemplary embodiment, the closing time of the next closing action is predicted based on the current operating parameters and the locally stored prediction data, including: determining multiple target nodes around the node to be predicted based on the current operating parameters; the current operating parameters correspond to the node to be predicted; obtaining the closing time of each target node and the partial derivative value of each target node in each preset direction from the locally stored prediction data; performing interpolation calculation based on the closing time of each target node and the partial derivative value of each target node in each preset direction, and determining the closing time corresponding to the node to be predicted as the closing time of the next closing action.

[0131] In an exemplary embodiment, interpolation calculation is performed according to the closing time of each target node and the partial derivative value of each target node in each preset direction, and the closing time corresponding to the node to be predicted is determined as the closing time of the next closing action, including: Hermite interpolation calculation is performed according to the closing time of each target node and the partial derivative value of each target node in each preset direction, and the closing time corresponding to the node to be predicted is determined as the closing time of the next closing action.

[0132] In an exemplary embodiment, referring to Figure 4 , a circuit breaker control method is provided, comprising the following steps: a data acquisition module determines whether the past N closing predictions are reliable. If not, the cloud computing module counts the size of external influencing factors and closing time during the past closing actions, trains the prediction model, builds a database, selects nodes to calculate their partial derivative values, and inputs the updated information into the mechanism control module for on-site prediction. If reliable, after receiving the closing command, the mechanism control module performs closing control in combination with the current external parameters.

[0133] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0134] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the memory, database or other medium mentioned in each embodiment provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0135] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0136] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be construed as limiting the scope of the present application. It should be noted that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A circuit breaker control system, characterized in that: The system includes a data acquisition module, a mechanism control module, a cloud computing module and a circuit breaker; wherein: The data acquisition module is used to collect the predicted closing time, operating parameters and actual closing time each time the circuit breaker performs a closing action; upload the collected predicted closing time, operating parameters and actual closing time to the cloud computing module; and send the collected operating parameters to the mechanism control module; The data acquisition module is further used to determine whether the prediction strategy needs to be adjusted at present according to the predicted closing time and the actual closing time of the historical N closing actions; if not, a first signal is sent to the mechanism control module; if so, a second signal is sent to the cloud computing module; The mechanism control module is used to respond to the first signal and predict the closing time of the next closing action according to the current operating parameters and the prediction data stored locally; The cloud computing module is used to respond to the second signal, train the random forest neural network model according to the predicted closing time, operating parameters and actual closing time of each closing action, and obtain a trained prediction model; output the prediction data of each node based on the trained prediction model; send the prediction data of each node to the mechanism control module to instruct the mechanism control module to update the locally stored prediction data; wherein each node corresponds to a different value of the operating parameter.

2. The system according to claim 1, characterized in that The cloud computing module is also used for: During the t-th iteration training process, the internal parameters of the random forest neural network model are dynamically optimized according to the whale optimization algorithm; According to the predicted closing time, operating parameters and actual closing time of each closing action, the random forest neural network model is trained to obtain the intermediate training results; The model parameters of the whale optimization algorithm and the random forest neural network model are adjusted according to the intermediate training results, and the t+1th iteration training is continued until the current number of iterations reaches the maximum number of iterations or the prediction accuracy of the random forest neural network model meets the preset requirements, thereby obtaining a trained prediction model.

3. The system according to claim 2, characterized in that The cloud computing module is also used to evaluate the prediction accuracy of the random forest neural network model through the determination coefficient, mean absolute error, mean absolute percentage error, and square error.

4. The system according to claim 2, characterized in that The cloud computing module is also used for: During the t-th iteration training process, several candidate solutions are updated according to the position update strategy; Calculate the fitness value of each candidate solution; The candidate solution with the highest fitness value is selected as the target solution, and the internal parameters of the random forest neural network model are optimized based on the target solution.

5. The system according to claim 1, characterized in that The cloud computing module is also used for: For each node, determine multiple reference nodes of the current node in a target direction; the target direction is any one of multiple preset directions; Using the trained prediction model, predicting the closing time of the current node and the multiple reference nodes in sequence to obtain multiple prediction results; Based on the multiple prediction results, a preset formula is used to calculate the partial derivative value of the current node in the target direction; Using the closing time of the current node and the partial derivative values ​​of the current node in the multiple preset directions as the prediction data of the current node; Wherein, the preset formula is: ； Where S is the partial derivative value of the current node in the target direction; is the closing time predicted by the prediction model for the i-th run; is the closing time predicted by the prediction model for the i+1th run; is the operating parameter input to the prediction model for the i-th time; is the operating parameter input to the prediction model for the i+1th time; n is the total number of prediction model runs.

6. The system according to claim 5, characterized in that The mechanism control module is also used for: Determine a plurality of target nodes around the node to be predicted according to current operating parameters, wherein the current operating parameters correspond to the node to be predicted; Obtain the closing time of each target node and the partial derivative value of each target node in each preset direction from the locally stored prediction data; An interpolation calculation is performed based on the closing time of each target node and the partial derivative value of each target node in each preset direction, so as to determine the closing time corresponding to the node to be predicted as the closing time of the next closing action.

7. The system according to claim 6, characterized in that The mechanism control module is also used to perform Hermite interpolation calculation according to the closing time of each target node and the partial derivative value of each target node in each preset direction, and determine the closing time corresponding to the node to be predicted as the closing time of the next closing action.

8. A circuit breaker control method, characterized in that: The method is applied to the circuit breaker control system according to any one of claims 1 to 7; the method comprises: The data acquisition module collects predicted closing time, operating parameters and actual closing time each time the circuit breaker performs a closing action; uploads the collected predicted closing time, operating parameters and actual closing time to the cloud computing module; and sends the collected operating parameters to the mechanism control module; The data acquisition module determines whether the prediction strategy needs to be adjusted at present according to the predicted closing time and the actual closing time of the historical N closing actions; if not, a first signal is sent to the mechanism control module; if so, a second signal is sent to the cloud computing module; The mechanism control module responds to the first signal and predicts the closing time of the next closing action according to the current operating parameters and the prediction data stored locally; The cloud computing module responds to the second signal, trains the random forest neural network model according to the predicted closing time, operating parameters and actual closing time of each closing action, and obtains a trained prediction model; outputs the prediction data of each node based on the trained prediction model; sends the prediction data of each node to the mechanism control module to instruct the mechanism control module to update the locally stored prediction data; wherein each node corresponds to a different value of the operating parameter.

9. The method according to claim 8, characterized in that According to the predicted closing time, operating parameters and actual closing time of each closing action, the random forest neural network model is trained to obtain a trained prediction model, including: During the t-th iteration training process, the internal parameters of the random forest neural network model are dynamically optimized according to the whale optimization algorithm; According to the predicted closing time, operating parameters and actual closing time of each closing action, the random forest neural network model is trained to obtain the intermediate training results; The model parameters of the whale optimization algorithm and the random forest neural network model are adjusted according to the intermediate training results, and the t+1th iteration training is continued until the current number of iterations reaches the maximum number of iterations or the prediction accuracy of the random forest neural network model meets the preset requirements, thereby obtaining a trained prediction model.

10. The method according to claim 9, characterized in that The method further comprises: The prediction accuracy of the random forest neural network model was evaluated by the coefficient of determination, mean absolute error, mean absolute percentage error, and squared error.

Images