An intelligent control method and system based on pole-mounted circuit breaker equipment

Abstract

The present invention discloses an intelligent control method and system based on pole-mounted circuit breaker equipment, which relates to the technical field of power grid control. The method includes collecting environmental data of the pole-mounted circuit breaker, using the Kalman filter fusion algorithm to output a fused state vector, constructing time series data, and using a time series neural network accelerated by photon computing to map the extracted features in the complex domain and perform optical phase shift modulation to predict future current values. In the method of the present invention, the features extracted by the time convolution layer are mapped in the complex domain to enhance the expression ability of non-linear features. The introduction of optical phase shift modulation further adjusts the phase and amplitude information, improving the non-linear modeling ability of time series features. By combining multi-modal data fusion with multi-scale consistency verification of fractal dimensions, the time scale deviation of different data sources is eliminated, improving the effectiveness and consistency of multi-dimensional feature fusion.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid control, and in particular to an intelligent control method and system based on a pole-mounted circuit breaker device. Background Art

[0002] With the rapid development of power systems, the automation and intelligence of distribution networks have become important development directions for power grid operation and management. Pole-mounted circuit breakers, as key equipment in distribution networks, are widely used for fault isolation, load control, and power protection in medium-voltage distribution systems. Traditional control methods for pole-mounted circuit breakers mainly rely on scheduled inspections, overcurrent protection, and simple remote control. These methods have played a certain role in ensuring power grid stability.

[0003] However, due to the limitations of traditional technical means, the control method of pole-mounted circuit breakers has poor adaptability to complex environmental changes and short-term current disturbances, and it is difficult to meet the needs of modern smart grids for efficient, accurate, and intelligent circuit breaker control. Traditional circuit breaker current prediction methods are mostly based on simple timing analysis or empirical models, which make it difficult to accurately predict future short-term current fluctuations, resulting in a high false triggering rate and affecting power supply stability. Secondly, the existing technology lacks in-depth analysis of current fractal characteristics and cannot accurately characterize the nonlinear characteristics of current fluctuations, which in turn affects the accuracy of circuit breaker control strategies. Moreover, although there are optimization strategies based on machine learning, most methods rely on fixed rules or historical data training and cannot dynamically adjust the control strategy according to real-time environmental changes, resulting in poor system adaptability. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an intelligent control method based on a pole-mounted circuit breaker device to solve the problem that most current prediction methods of traditional circuit breakers are based on simple timing analysis or empirical models, which makes it difficult to accurately predict future short-term current fluctuations, resulting in a high false triggering rate and affecting power supply stability. Secondly, the existing technology lacks in-depth analysis of current fractal characteristics and cannot accurately characterize the nonlinear characteristics of current fluctuations, which in turn affects the accuracy of the circuit breaker control strategy. Moreover, although there are optimization strategies based on machine learning, most methods rely on fixed rules or historical data training and cannot dynamically adjust the control strategy according to real-time environmental changes, resulting in poor system adaptability.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides an intelligent control method based on a pole-mounted circuit breaker device, comprising:

[0008] Collect the environmental data of the pole-mounted circuit breaker, use the Kalman filter fusion algorithm to output the fusion state vector, construct time series data, use the time series neural network accelerated by photon computing to map the extracted features in the complex domain and perform optical phase shift modulation to predict the future current value;

[0009] Perform multi-scale partitioning on the predicted current time series, calculate the fractal dimension of the current, use the fractional Brownian motion algorithm to establish a non-linear perturbation model of current fluctuations to calculate the current fluctuation value, and verify the consistency of the fractal dimension through linear regression calculation, calculate the adjustment factor and correct the predicted value;

[0010] Calculate the current trend correction factor, dimension anomaly index and magnetic field perturbation index, determine the optimal strategy through the reinforcement learning algorithm and calculate the short-term perturbation value, and perform decision delay verification;

[0011] Monitor the state data of the circuit breaker, calculate the thermal anomaly index and give an early warning, analyze the current stability value, and perform closing operations.

[0012] As a preferred solution of the intelligent control method based on the pole-mounted circuit breaker device of the present invention, wherein: the collection of the environmental data of the pole-mounted circuit breaker and the use of the Kalman filter fusion algorithm to output the fusion state vector refer to the collection of environmental data through sensors based on the pole-mounted circuit breaker device, including temperature data, humidity data, current data and leakage current data;

[0013] Perform data synchronization on the collected environmental data using linear interpolation method and perform noise reduction using adaptive Wiener filter;

[0014] Using the Kalman filter fusion algorithm, for the environmental vector composed of environmental data, predict the current state and predict the covariance, calculate the Kalman gain, correct the state vector using the measurement data and update the covariance matrix, output the fusion state vector, and construct time series data;

[0015] Use the time series neural network accelerated by photon computing to construct a time series model, including an input layer, a time convolutional layer, a photon computing layer and an output layer;

[0016] The input layer inputs the fusion state vector data after normalization processing. The time convolutional layer uses the time convolutional network TCN to extract time series features, map the features extracted by the time convolutional layer in the complex domain, perform optical phase shift modulation, and finally predict the future current value through the photon computing layer;

[0017] Use a calibrated training set to train the model. Select the cross-entropy loss function to calculate the computational loss between the output predicted by the model and the actual label. Use the chain rule to calculate the gradients of the loss with respect to the parameters of each layer for backpropagation calculation. Use the Adam optimizer to update the parameters. Stop updating when the loss of the model no longer decreases significantly during continuous updates, and output the model parameters.

[0018] Use the newly collected data to predict the current data through a time series model.

[0019] As a preferred solution of the intelligent control method based on the pole-mounted circuit breaker device of the present invention, wherein: the multi-scale partitioning of the predicted current time series, calculating the current fractal dimension refers to performing a fractal dimension correction FDA on the predicted current data, forming sequence data from the predicted current data, and setting the scale parameter ;

[0020] Through the multi-scale Box-Counting algorithm at each scale Perform grid partitioning on the predicted current time series, count the number of intervals covering the entire current series, calculate the number of scale intervals according to the box counting method, and calculate the current fractal dimension through linear regression fitting;

[0021] Statistical current fractal dimension data and calculate the sum of the mean and standard deviation as the fractal threshold, determine the minimum threshold limit value based on historical data, and use the maximum value of the minimum threshold limit value and the fractal threshold as the dimension threshold. If the calculated current fractal dimension is greater than or equal to the dimension threshold, it is judged that correction is required;

[0022] Use the fractional Brownian motion algorithm fBM to establish a non-linear perturbation model of current fluctuations, and determine the current fluctuation value through the Hurst exponent and perturbation amplitude value of the current fluctuations ;

[0023] Based on the current fluctuation value Verify, and perform multi-scale segmentation on through the multi-scale Box-Counting algorithm according to the scale parameter , perform the calculation of the box technique, and calculate the corresponding verification fractal dimension through linear regression ;

[0024] Verify the difference between the verification fractal dimension and . Determine the minimum value as the verification threshold based on historical data. If the fractal difference is less than the verification threshold, it means Maintain consistency at multiple scales and the verification passes;

[0025] Based on the verified Set an adjustment factor and correct the predicted value according to the adjustment factor.

[0026] As a preferred solution of the intelligent control method based on the pole-mounted circuit breaker device described in the present invention, wherein: calculating the current trend correction factor, the dimension anomaly index, and the magnetic field disturbance index, determining the optimal strategy through the reinforcement learning algorithm, and defining the current trend correction factor based on the relative change rate between the corrected current predicted value and the over-predicted value;

[0027] Based on the difference between the current fractal dimension and the mean value of the fractal dimension to determine the dimension anomaly index ;

[0028] Based on the finite element simulation FEM to determine the magnetic induction intensity, calculate the electromagnetic torque of the circuit breaker at the current time t, and calculate the relative change rate of the electromagnetic torque at different times as the magnetic field disturbance index;

[0029] Use the curriculum learning mechanism to dynamically adjust the difficulty of the reinforcement learning task based on the current trend correction factor, the dimension anomaly index, and the magnetic field disturbance index, and determine the task complexity ;

[0030] Based on As a reward correction term, perform Q-learning update through the reinforcement learning algorithm and dynamically adjust the learning rate;

[0031] Based on the current predicted value, the current fractal dimension, the magnetic field disturbance index, and the task complexity to form a state space, based on the adjustment, closing, and holding states of the circuit breaker to form an action space, obtain the Q-value table through Q-learning update, including the corresponding Q-values of different state spaces and action spaces, and obtain the optimal strategy and determine the state of the circuit breaker in different states.

[0032] As a preferred solution of the intelligent control method based on the pole-mounted circuit breaker device described in the present invention, wherein: calculating the short-time disturbance value, performing decision delay verification, calculating the short-time disturbance based on the optimal strategy, wherein defining a disturbance time window and calculating the short-time disturbance according to the difference of the corrected predicted value;

[0033] Based on the sum of the historical mean and twice the standard deviation of the short-time disturbance index as the delay threshold, if the short-time disturbance index is less than or equal to the delay threshold, it is judged as non-delay and directly execute the optimal strategy, if the short-time disturbance index is greater than the delay threshold, then perform decision delay;

[0034] Perform time delay based on the decision delay value and re-obtain the Q-value table through Q-learning update to determine the optimal strategy after delay.

[0035] As a preferred solution of the intelligent control method based on the pole-mounted circuit breaker device of the present invention, wherein: the circuit breaker status data is monitored, the thermal anomaly index is calculated and early warning is carried out, and the action implementation based on the optimal strategy is carried out. The circuit breaker status data is monitored, the circuit breaker temperature data is collected, and the thermal anomaly index is calculated according to the difference between the ambient temperature and the collected temperature;

[0036] Based on the sum of the rated temperature of the circuit breaker and the allowable temperature rise value of the circuit breaker as the temperature threshold, if the thermal anomaly index is greater than or equal to the temperature threshold, it is determined that the circuit breaker temperature is abnormal, and an alarm is issued through the warning device.

[0037] As a preferred solution of the intelligent control method based on the pole-mounted circuit breaker device of the present invention, wherein: the analysis of the current stability value and the closing operation are based on the action implementation of the optimal strategy. If the circuit breaker is in the open state, the mean value of the corrected current prediction value within the acquisition window is used as the current stability value, and the sum of the mean value and the standard deviation of the current stability is used as the current threshold. If the current stability value is less than the current threshold, it is determined that the closing condition is met and the closing operation is carried out.

[0038] In a second aspect, the present invention provides a system for an intelligent control method based on a pole-mounted circuit breaker device, including,

[0039] A data acquisition and processing module that acquires the ambient data of the pole-mounted circuit breaker and performs time synchronization and noise reduction processing on the multi-modal data;

[0040] A time series construction module that uses the Kalman filter algorithm to fuse the multi-modal ambient data and constructs multi-dimensional time series data based on the fused state vector;

[0041] A time series neural network module that uses a TCN network accelerated by photon computing to perform multi-scale time series feature extraction on the time series of the fused state vector;

[0042] A non-linear correction module that calculates the current fractal dimension of the predicted current time series and performs consistency verification, and non-linearly corrects the current prediction value based on the adjustment factor calculated from the fractal dimension;

[0043] A reinforcement learning adaptive control module that calculates the current trend correction factor, the dimension anomaly index, and the magnetic field perturbation index, dynamically adjusts the complexity of the reinforcement learning task, updates the Q-value table through Q-learning, dynamically calculates the optimal strategy, and performs decision delay verification;

[0044] An intelligent warning module that acquires the circuit breaker status data and performs status monitoring and fault warning.

[0045] In a third aspect, the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and: when the computer program is executed by the processor, any step of the intelligent control method based on the pole-mounted circuit breaker device as described in the first aspect of the present invention is implemented.

[0046] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and: when the computer program is executed by the processor, any step of the intelligent control method based on the pole-mounted circuit breaker device as described in the first aspect of the present invention is implemented.

[0047] The beneficial effects of the present invention are as follows: by mapping the features extracted by the temporal convolutional layer to the complex domain, the expression ability of non-linear features is enhanced. The introduction of optical phase shift modulation further adjusts the phase and amplitude information, improving the non-linear modeling ability of temporal features. By combining multi-modal data fusion with multi-scale consistency verification of fractal dimension, the time-scale deviation of different data sources is eliminated, improving the effectiveness and consistency of multi-dimensional feature fusion. By combining the phase adjustment of optical phase shift modulation with the multi-scale self-similarity of fractal dimension, the non-linear features of current fluctuations at different frequencies and amplitudes can be comprehensively captured. The combination of the non-linear correction mechanism of the adjustment factor and the efficient calculation of photonic computing significantly improves the correction speed and accuracy of the predicted value. The current trend correction factor dynamically captures the current fluctuation trend and amplitude, and through the dimension anomaly index of the current fractal dimension, non-linear fluctuations and abnormal fluctuations are accurately detected, ensuring the safety of the power grid and the stable operation of equipment. Description of the Drawings

[0048] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0049] Figure 1 It is a schematic flowchart of the intelligent control method based on the pole-mounted circuit breaker device in Embodiment 1.

[0050] Figure 2 It is a schematic structural diagram of the intelligent control system based on the pole-mounted circuit breaker device in Embodiment 2. Detailed Embodiments

[0051] To make the above objects, features, and advantages of the present invention more obvious and understandable, the detailed embodiments of the present invention will be described in detail below with reference to the drawings in the specification.

[0052] In the following description, many specific details are set forth in order to provide a thorough understanding of the present invention. However, the present invention may be practiced in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the spirit of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0053] Secondly, as used herein, an "embodiment" or "embodiments" refers to specific features, structures, or characteristics that may be included in at least one implementation of the present invention. The appearances of "in one embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or alternative embodiments that exclude each other.

[0054] Embodiment 1, referring to Figure 1 and Figure 2 , is the first embodiment of the present invention. This embodiment provides an intelligent control method based on pole-mounted circuit breaker equipment, including the following steps:

[0055] S1, Collect the environmental data of the pole-mounted circuit breaker, use the Kalman filter fusion algorithm to output the fused state vector, construct time series data, and use the time series neural network accelerated by photon computing to map the extracted features in the complex domain and perform optical phase shift modulation to predict the future current value;

[0056] Preferably, collect the environmental data of the pole-mounted circuit breaker, use the Kalman filter fusion algorithm to output the fused state vector, construct time series data, and use the time series neural network accelerated by photon computing to map the extracted features in the complex domain and perform optical phase shift modulation to predict the future current value. The environmental data is collected through sensors based on the pole-mounted circuit breaker equipment, including temperature data, humidity data, current data, and leakage current data;

[0057] Perform data synchronization on the collected environmental data using linear interpolation and noise reduction using adaptive Wiener filtering;

[0058] Use the Kalman filter fusion algorithm to perform current state prediction and prediction covariance for the environmental vector composed of environmental data, expressed as:

[0059] ;

[0060] ;

[0061] Where represents the predicted current state vector at time t+1, represents the state transition matrix, which represents the transition relationship of the system from time t to time t+1 and is learned from historical data, Denote the predicted current state vector at time t, which includes the collected temperature data, humidity data, current data, leakage current data, and historical fault data. Denote the control input matrix, which includes the influence values of temperature change and humidity change on current and leakage current, and the influence value of current change on the thermal load of the circuit breaker. Denote the control input vector, which includes the temperature change amount, humidity change amount, current change amount, and leakage current change amount in the environmental data. Denote the predicted covariance matrix at time t + 1. Denote the covariance matrix at time t. Denote the process noise covariance matrix, which represents the uncertainty of state transition. Denote the transpose calculation of the state transition matrix.

[0062] Calculate the Kalman gain, use the measurement data to correct the state vector and update the covariance matrix, output the fused state vector, and construct the time series data.

[0063] Construct a time series model using a time series neural network accelerated by photonic computing, including an input layer, a temporal convolutional layer, a photonic computing layer, and an output layer.

[0064] The input layer inputs the fused state vector data after normalization processing. The temporal convolutional layer uses a temporal convolutional network (TCN) to extract temporal features, expressed as:

[0065] ;

[0066] where denotes the convolutional output of the i-th neuron, M denotes the time window size, denotes the fused state vector at the past time point t - j, denotes the convolutional weight;

[0067] Map the features extracted by the temporal convolutional layer to the complex domain and perform optical phase shift modulation, expressed as:

[0068] ;

[0069] ;

[0070] ;

[0071] where denotes the input data of the i-th neuron after optical phase shift modulation, denotes the i-th phase shift angle of the optical phase shift factor, denotes the phase shift adjustment coefficient, u denotes the imaginary unit, satisfying ;

[0072] Perform photon-computing layer calculations to make a final prediction of the future current value, expressed as:

[0073] ;

[0074] ;

[0075] where represents the output of the photon-computing layer of the i-th neuron, N represents the number of neurons in the hidden layer of the neural network, represents the current prediction at time t + Δt, represents the fully connected weight, b represents the bias term, represents the photon-computing weight;

[0076] Use the calibrated training set to train the model. Select the cross-entropy loss function to calculate the computational loss between the output predicted by the model and the actual label. Use the chain rule to calculate the gradients of the loss with respect to the parameters of each layer for backpropagation calculation. Use the Adam optimizer to update the parameters. Stop updating when the loss of the model no longer decreases significantly during consecutive updates, and output the model parameters;

[0077] Use the newly collected data to predict the current data through a time series model.

[0078] Through multi-modal environmental data acquisition, comprehensively cover the environmental factors affecting current fluctuations, making the predicted value closer to the actual scenario and environmental conditions. Through data synchronization and adaptive noise reduction, improve the effectiveness of multi-dimensional feature fusion;

[0079] Through the Kalman filter fusion algorithm, quantify the uncertainty in the state transition process, reduce the interference of noise on state prediction and covariance estimation. Through the time series neural network accelerated by photon computing, photon computing transmits data at the speed of light in convolution calculations, achieving ultra-high bandwidth and ultra-low latency convolution operations, parallel computing multiple convolution kernels and time windows, significantly accelerating time convolution operations and feature extraction. Through the long-term dependence modeling ability of the time convolution network, it can capture the periodicity and trend of current data in a long time series. The combination of convolution kernel sharing and photon-computing acceleration improves the accuracy and stability of time series feature extraction;

[0080] Map the features extracted by the temporal convolutional layer to the complex domain, convert the temporal features into complex forms, thereby enhancing the expression ability of non-linear features. The introduction of optical phase shift modulation further adjusts the phase and amplitude information, improving the non-linear modeling ability of temporal features. Utilize the phase modulation and interference principles of photonic computing to perform matrix operations and complex convolutions in the complex domain, while retaining both amplitude and phase information, achieving accurate modeling of non-linear and periodic fluctuation features; Through the high parallelism and low latency of photonic computing, enhance the ability to handle dynamic changes in power grid load fluctuations and environmental disturbances, and thereby improve the effectiveness of predicting future current values.

[0081] S2. Perform multi-scale partitioning on the predicted current time series, calculate the current fractal dimension, use the fractional Brownian motion algorithm to establish a non-linear perturbation model of current fluctuations to calculate the current fluctuation value, and verify the consistency of the fractal dimension through linear regression calculation, calculate the adjustment factor and perform prediction value correction;

[0082] Preferably, perform multi-scale partitioning on the predicted current time series, calculate the current fractal dimension, use the fractional Brownian motion algorithm to establish a non-linear perturbation model of current fluctuations to calculate the current fluctuation value, and verify the consistency of the fractal dimension through linear regression calculation, calculate the adjustment factor and perform prediction value correction, perform fractal dimension correction FDA on the predicted current data, form the predicted current data into sequence data, and set the scale parameter , set as , and m is determined according to historical data. Through the multi-scale Box-Counting algorithm at each scale , perform grid partitioning on the predicted current time series, count the number of intervals covering the entire current sequence, and calculate the current fractal dimension according to the box counting method, expressed as:

[0083] ;

[0084] Where represents the current fractal dimension, represents the scale the number of intervals at;

[0085] Statistically analyze the current fractal dimension data and calculate the sum of the mean and standard deviation as the fractal threshold, and determine the lowest threshold limit value based on historical data. Use the maximum value of the lowest threshold limit value and the fractal threshold as the dimension threshold. If the calculated current fractal dimension is greater than or equal to the dimension threshold, it is determined that correction is required;

[0086] Use the fractional Brownian motion algorithm fBM to establish a non-linear perturbation model of current fluctuations, expressed as:

[0087] ;

[0088] where represents the current fluctuation value under fractional Brownian motion at time t, represents the Hurst exponent of the current fluctuation at time t, which can be set to , represents the perturbation amplitude value, which is determined by the historical fluctuation standard deviation, represents the random perturbation value of standard Brownian motion at time t;

[0089] Based on for verification, through the multi-scale Box-Counting algorithm according to the scale parameter for perform multi-scale segmentation, and perform the calculation of the box technique, and calculate the corresponding verified fractal dimension through linear regression, expressed as:

[0090] ;

[0091] where represents the verified fractal dimension, C is the intercept term in linear regression, representing the cardinality of the shape object at the smallest scale, which can be obtained by the least squares method;

[0092] Verify the difference between the verified fractal dimension and , determine the minimum value as the verification threshold based on historical data. If the fractal difference is less than the verification threshold, it means is consistent at multiple scales and the verification passes;

[0093] Based on the verified set the adjustment factor and correct the predicted value according to the adjustment factor, expressed as:

[0094] ;

[0095] ;

[0096] where represents the adjustment factor, represents the minimum value to prevent the denominator from being zero, represents the corrected current prediction value at time t+Δt, represents the difference between the actual current value and the predicted value.

[0097] Multi-scale partitioning can comprehensively capture the self-similarity and multifractal structure of current fluctuations, reflect the complexity of current fluctuations at different time scales, and is particularly suitable for modeling non-linear and non-stationary current fluctuations, improving the prediction accuracy of mutations, spikes, and periodic fluctuations. Through fractal dimension calculation, it describes the complexity and multi-scale self-similarity of current fluctuations, can accurately quantify the chaotic degree of current fluctuations, accurately identify non-linear fluctuations and abnormal fluctuations, and trigger the correction mechanism of predicted values, avoiding misjudgment of stationary fluctuations, effectively reducing the redundant calculation of error correction, and improving the accuracy and efficiency of correction calculation. By establishing a non-linear perturbation model of current fluctuations through fractional Brownian motion (fBM), it can dynamically adjust the long-term trend and short-term random fluctuations of current fluctuations. Through fractal dimension consistency verification, it eliminates the influence of random fluctuations and external noise in the non-linear perturbation model, ensuring the reliability and stability of the fBM model. Through the calculation of adjustment factors and the correction of predicted values, it effectively eliminates short-term fluctuations and noise interference, and improves the robustness and accuracy of the corrected predicted values;

[0098] By organically combining multi-modal data fusion and multi-scale consistency verification of fractal dimensions, it eliminates the time-scale deviation of different data sources, improves the effectiveness and consistency of multi-dimensional feature fusion. The multi-scale time-series features extracted by TCN accelerated by photon computing provide high-precision and high-timeliness feature inputs for the multi-scale segmentation and self-similarity verification of the Box-Counting algorithm. By quantitatively describing the complexity and multi-scale self-similarity of current fluctuations through the current fractal dimension, it enhances the performance of multi-modal data in detecting non-linear fluctuations and abnormal fluctuations. By combining the phase adjustment of optical phase shift modulation with the multi-scale self-similarity of fractal dimensions, it can comprehensively capture the non-linear features of current fluctuations at different frequencies and amplitudes. The non-linear correction mechanism of the adjustment factor combined with the efficient calculation of photon computing significantly improves the correction speed and accuracy of predicted values. The dynamic adjustment factor based on fractal dimensions combined with the phase adjustment and complex domain mapping of the photon computing layer realizes the precise correction and fast calculation of non-linear fluctuations and complex fluctuations.

[0099] S3. Calculate the current trend correction factor, dimension anomaly index, and magnetic field perturbation index, determine the optimal strategy through the reinforcement learning algorithm, calculate the short-term perturbation value, and conduct decision delay verification;

[0100] Preferably, calculate the current trend correction factor, dimension anomaly index, and magnetic field perturbation index, determine the optimal strategy through the reinforcement learning algorithm, and based on the corrected current predicted value, define the current trend correction factor, expressed as:

[0101] ;

[0102] where represents the current trend correction factor, represents the corrected current prediction value at time ti;

[0103] The dimension anomaly index is calculated based on the current fractal dimension and is expressed as:

[0104] ;

[0105] in represents the dimension anomaly index, represents the mean value of the current fractal dimension;

[0106] Based on the finite element simulation FEM, the magnetic induction intensity is determined, and the electromagnetic torque of the circuit breaker at the current time t is calculated, and the magnetic field disturbance index is calculated, which is expressed as:

[0107] ;

[0108] ;

[0109] in represents the magnetic field perturbation index, and denote the electromagnetic torque of the circuit breaker at time t and time t-1, respectively, represents the current density, represents the magnetic induction intensity, and V represents the action area of the circuit breaker;

[0110] The curriculum learning mechanism is used to dynamically adjust the difficulty of the reinforcement learning task based on the current trend correction factor, dimension anomaly index, and magnetic field disturbance index, which can be expressed as:

[0111] ;

[0112] ;

[0113] in represents the complexity of the task, T represents the time window of the course learning phase, Indicates the initial time, represents a fixed reward scale parameter (determined based on historical data), Indicates the task complexity modification reward item;

[0114] based on The Q-learning update performed by the reinforcement learning algorithm as a reward correction term and the dynamic adjustment of the learning rate is expressed as:

[0115] ;

[0116] in represents the dynamic learning rate, represents the initial learning rate;

[0117] A state space is composed of a predicted current value, a current fractal dimension, a magnetic field perturbation index, and a task complexity. An action space is composed of the adjustment, closing, and holding states of a circuit breaker. A Q-value table is updated through Q-learning, including the corresponding Q-values for different state spaces and action spaces, and an optimal strategy is obtained to determine the state of the circuit breaker under different states.

[0118] The current trend correction factor is used to dynamically capture the current fluctuation trend and amplitude. The dimension anomaly index of the current fractal dimension is used to accurately detect non-linear fluctuations and abnormal fluctuations. The electromagnetic torque perturbation of the finite element simulation (FEM) is used to accurately quantify the intensity and volatility of the electromagnetic perturbation of the circuit breaker. The task complexity adjustment and reward correction of the curriculum learning mechanism are used to effectively improve the scenario adaptability and generalization of the reinforcement learning model. The Q-learning update mechanism of the reinforcement learning algorithm is used to dynamically find the optimal strategy in the state space and action space to achieve fine control of the closing, holding, adjustment, etc. operations of the circuit breaker.

[0119] Furthermore, a short-term perturbation value is calculated, and decision delay verification is performed. The short-term perturbation is considered based on the optimal strategy, where a perturbation time window is defined to calculate the short-term perturbation, expressed as:

[0120] ;

[0121] where represents the short-term perturbation index, represents the perturbation time window;

[0122] Based on the sum of the historical mean and twice the standard deviation of the short-term perturbation index as the delay threshold, if the short-term perturbation index is less than or equal to the delay threshold, it is judged as not delayed, and the optimal strategy is directly executed. If the short-term perturbation index is greater than the delay threshold, decision delay is performed, expressed as:

[0123] ;

[0124] where represents the decision delay value, represents the delay threshold, represents the maximum delay time;

[0125] Time delay is performed based on the decision delay value, and the Q-value table is updated again through Q-learning to determine the optimal strategy after delay.

[0126] Calculating short-time disturbance values by disturbing the dynamic changes of current fluctuations within a time window can accurately identify transient fluctuations, short-time pulses, and high-frequency noise, significantly improve the response speed and decision-making accuracy of the power grid under transient disturbances and sudden fluctuations, significantly reduce misjudgments and misoperations, and ensure the safety of the power grid and the stable operation of equipment.

[0127] S4. Monitor the breaker status data, calculate the thermal anomaly index and issue a warning, analyze the current stable value, and perform a closing operation.

[0128] Preferably, monitor the breaker status data, calculate the thermal anomaly index and issue a warning, implement actions based on the optimal strategy, monitor the breaker status data, collect the breaker temperature data and calculate the thermal anomaly index, expressed as:

[0129] ;

[0130] Where represents the thermal anomaly index, m represents the total number of acquisition times, represents the breaker temperature at time i, represents the ambient temperature at time i;

[0131] Based on the sum of the breaker rated temperature and the allowable temperature rise value of the breaker as the temperature threshold, if the thermal anomaly index is greater than or equal to the temperature threshold, it is determined that the breaker temperature is abnormal, and an alarm is issued through the warning device.

[0132] By cumulatively calculating the difference between the breaker temperature and the ambient temperature, it can accurately reflect the internal heat accumulation and temperature rise of the breaker, can capture the abnormal temperature change trend in real time, especially when the temperature continues to rise or fluctuates violently, can quickly identify the abnormality and issue a timely warning, reduce the dependence on a single temperature threshold, can accurately identify the abnormal change trend in temperature fluctuations, and effectively improve the accuracy and stability of temperature monitoring.

[0133] Furthermore, analyzing the current stable value and performing a closing operation refers to the implementation of actions based on the optimal strategy. If the breaker is in the open state, the mean value of the corrected current prediction value within the acquisition window is used as the current stable value, and the sum of the mean value and the standard deviation of the current stability is used as the current threshold. If the current stable value is less than the current threshold, it is determined that the closing condition is met and the closing operation is performed.

[0134] By calculating the mean value of the corrected current prediction value within the acquisition window, it can accurately reflect the stable state of the current over a period of time, especially when the load fluctuation is small, effectively identify the stable state and continuous state of the current.

[0135] This embodiment also provides a system for an intelligent control method based on pole-mounted breaker equipment, including

[0136] The data acquisition and processing module collects the environmental data of the pole-mounted circuit breaker and performs time synchronization and noise reduction processing on the multi-modal data;

[0137] The data fusion and time series construction module uses the Kalman filtering algorithm to fuse the multi-modal environmental data and constructs multi-dimensional time series data based on the fused state vector;

[0138] The time series neural network module uses the TCN network accelerated by photon computing to extract multi-scale time series features of the time series of the fused state vector, maps the time series features to the complex domain and performs optical phase shift modulation to predict the current;

[0139] The non-linear correction module performs multi-scale partitioning on the predicted current time series, calculates the current fractal dimension and performs consistency verification, and non-linearly corrects the current prediction value based on the adjustment factor calculated from the fractal dimension;

[0140] The reinforcement learning adaptive control module calculates the current trend correction factor, the dimension anomaly index and the magnetic field perturbation index, dynamically adjusts the complexity of the reinforcement learning task, updates the Q-value table through Q-learning, dynamically calculates the optimal strategy and performs decision delay verification;

[0141] The intelligent warning module collects the circuit breaker status data and performs status monitoring and fault warning;

[0142] The communication module is modularly integrated with edge computing communication to achieve multi-module linkage and real-time data transmission.

[0143] This embodiment also provides a computer device applicable to the situation of the intelligent control method based on the pole-mounted circuit breaker device, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the intelligent control method based on the pole-mounted circuit breaker device proposed in the above embodiment.

[0144] The computer device may be a terminal, which includes a processor, a memory, a communication interface, a display screen, and an input device connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the housing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0145] This embodiment also provides a storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the intelligent control method based on the pole-mounted circuit breaker device as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM for short), Electrically Erasable Programmable Read-Only Memory (EEPROM for short), Erasable Programmable Read Only Memory (EPROM for short), Programmable Red-Only Memory (PROM for short), Read-Only Memory (ROM for short), magnetic memory, flash memory, a magnetic disk, or an optical disc.

[0146] In summary, the present invention maps the features extracted by the temporal convolutional layer into the complex domain, converts the temporal features into complex forms, thereby enhancing the expression ability of non-linear features. The introduction of optical phase shift modulation further adjusts the phase and amplitude information, improving the non-linear modeling ability of temporal features. Through the organic combination of multi-modal data fusion and multi-scale consistency verification of the fractal dimension, the time scale deviation of different data sources is eliminated, and the effectiveness and consistency of multi-dimensional feature fusion are improved. By combining the phase adjustment of optical phase shift modulation with the multi-scale self-similarity of the fractal dimension, the non-linear features of current fluctuations at different frequencies and amplitudes can be comprehensively captured. The combination of the non-linear correction mechanism of the adjustment factor and the efficient calculation of photon computing significantly improves the correction speed and accuracy of the predicted value. The current trend correction factor dynamically captures the current fluctuation trend and amplitude, and through the dimension anomaly index of the current fractal dimension, non-linear fluctuations and abnormal fluctuations are accurately detected, ensuring the safety of the power grid and the stable operation of equipment.

[0147] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. An intelligent control method based on pole-mounted circuit breaker equipment, characterized in that, Including: Collect the environmental data of the pole-mounted circuit breaker, use the Kalman filter fusion algorithm to output the fusion state vector, construct time series data, and use the time series neural network accelerated by photon computing to map the extracted features in the complex domain and perform optical phase shift modulation to predict the future current value; Perform multi-scale division on the predicted current time series, calculate the current fractal dimension, use the fractional Brownian motion algorithm to establish a non-linear perturbation model of current fluctuations to calculate the current fluctuation value, and verify the consistency of the fractal dimension through linear regression calculation, calculate the adjustment factor and correct the predicted value; Calculate the current trend correction factor, dimension anomaly index, and magnetic field perturbation index, determine the optimal strategy through the reinforcement learning algorithm and calculate the short-term perturbation value, and perform decision delay verification; Monitor the circuit breaker status data, calculate the thermal anomaly index and give an early warning, analyze the current stable value, and perform the closing operation; The collection of the environmental data of the pole-mounted circuit breaker and the use of the Kalman filter fusion algorithm to output the fusion state vector refer to the collection of environmental data through sensors based on the pole-mounted circuit breaker equipment, including temperature data, humidity data, current data, and leakage current data; Use the linear interpolation method for data synchronization of the collected environmental data and use the adaptive Wiener filter for noise reduction; Use the Kalman filter fusion algorithm to predict the current state and predict the covariance for the environmental vector composed of environmental data, calculate the Kalman gain, correct the state vector using the measurement data and update the covariance matrix, output the fusion state vector, and construct time series data; Construct a time series model using the time series neural network accelerated by photon computing, including an input layer, a temporal convolutional layer, a photon computing layer, and an output layer; The input layer inputs the normalized fusion state vector data. The temporal convolutional layer uses the temporal convolutional network (TCN) to extract temporal features, map the features extracted by the temporal convolutional layer in the complex domain, perform optical phase shift modulation, and finally predict the future current value through the photon computing layer; Use the calibrated training set to train the model, select the cross-entropy loss function to calculate the calculation loss between the output of the model prediction and the actual label, use the chain rule to calculate the gradient of the loss with respect to the parameters of each layer for backpropagation calculation, use the Adam optimizer to update the parameters, and stop updating when the loss of the model no longer decreases significantly during continuous updates, and output the model parameters; Use the newly collected data to predict the current data through the time series model.

2. The intelligent control method based on the pole-mounted circuit breaker device according to claim 1, wherein: The multi-scale division of the predicted current time series and the calculation of the current fractal dimension refer to the fractal dimension correction FDA for the predicted current data. The predicted current data is composed into sequence data, and the scale parameter is set. ; At each scale through the multi-scale Box-Counting algorithm divide the predicted current time series into grids, count the number of intervals covering the entire current series, calculate the number of scale intervals according to the box counting method, and calculate the current fractal dimension by linear regression fitting; Statistically analyze the current fractal dimension data and calculate the sum of the mean and standard deviation as the fractal threshold, determine the lowest threshold limit value based on historical data, and use the maximum value of the lowest threshold limit value and the fractal threshold as the dimension threshold. If the calculated current fractal dimension is greater than or equal to the dimension threshold, it is determined that correction is required; Use the fractal Brownian motion algorithm fBM to establish a non-linear perturbation model of current fluctuations, and determine the current fluctuation value through the Hurst exponent and perturbation amplitude value of the current fluctuations ; Based on the current fluctuation value Verify by performing multi-scale Box-Counting algorithm according to the scale parameter for perform multi-scale segmentation, calculate the box technique, and calculate the corresponding verified fractal dimension through linear regression ; Verify the fractal dimension and Verify the difference value. Determine the minimum value as the verification threshold based on historical data. If the fractal difference is less than the verification threshold, it means consistency is maintained at multiple scales and the verification passes; Based on the verified Set an adjustment factor and correct the predicted value according to the adjustment factor.

3. The intelligent control method based on the pole-mounted circuit breaker device according to claim 2, wherein: The calculation of the current trend correction factor, dimension anomaly index, and magnetic field perturbation index, determine the optimal strategy through the reinforcement learning algorithm, and define the current trend correction factor based on the relative change rate of the corrected current prediction value and the over-predicted value; Determine the dimension anomaly index based on the difference between the current fractal dimension and the mean value of the fractal dimension ;​ Determine the magnetic induction intensity based on finite element simulation (FEM), calculate the electromagnetic torque of the circuit breaker at the current time t, and calculate the relative change rate of the electromagnetic torque at different times as the magnetic field disturbance index; Use the course learning mechanism to dynamically adjust the difficulty of the reinforcement learning task based on the current trend correction factor, dimension anomaly index, and magnetic field perturbation index, and determine the task complexity ; Based on Q-learning update is performed through the reinforcement learning algorithm as a reward correction term, and the learning rate is dynamically adjusted; Based on the predicted current value, current fractal dimension, magnetic field disturbance index, and task complexity, a state space is formed. Based on the adjustment, closing, and holding states of the circuit breaker, an action space is formed. The Q-value table is updated through Q-learning, including the corresponding Q-values for different state spaces and action spaces, and the optimal strategy is obtained to determine the state of the circuit breaker under different states.

4. The intelligent control method based on the pole-mounted circuit breaker device according to claim 3, wherein: Calculate the short-term disturbance value, conduct decision delay verification, and calculate the short-term disturbance based on the optimal strategy, where a disturbance time window is defined and the short-term disturbance is calculated based on the difference in the corrected predicted values. Based on the sum of the historical mean and twice the standard deviation of the short-term disturbance index as the delay threshold, if the short-term disturbance index is less than or equal to the delay threshold, it is judged as not delayed and the optimal strategy is directly executed. If the short-term disturbance index is greater than the delay threshold, decision delay is performed. Perform time delay based on the decision delay value and update the Q-value table again through Q-learning to determine the optimal strategy after delay.

5. The intelligent control method based on the pole-mounted circuit breaker device according to claim 4, characterized in that: Monitor the state data of the circuit breaker, calculate the thermal anomaly index and issue an early warning. Based on the implementation of the actions of the optimal strategy, monitor the state data of the circuit breaker, collect the temperature data of the circuit breaker, and calculate the thermal anomaly index based on the difference between the environmental temperature and the collected temperature. Based on the sum of the rated temperature of the circuit breaker and the allowable temperature rise value of the circuit breaker as the temperature threshold, if the thermal anomaly index is greater than or equal to the temperature threshold, it is judged that the temperature of the circuit breaker is abnormal, and an alarm is issued through the warning device.

6. The intelligent control method based on the pole-mounted circuit breaker device according to claim 5, characterized in that: Analyze the current stability value. The closing operation is based on the implementation of the actions of the optimal strategy. If the circuit breaker is in the open state, the mean value of the corrected current predicted value within the acquisition window is used as the current stability value, and the sum of the mean value and the standard deviation of the current stability is used as the current threshold. If the current stability value is less than the current threshold, it is judged that the closing condition is met and the closing operation is performed.

7. A system for an intelligent control method based on pole-mounted circuit breaker equipment, based on the intelligent control method based on pole-mounted circuit breaker equipment according to any one of claims 1 to 6, characterized in that: Including, A data acquisition and processing module that collects the environmental data of the pole-mounted circuit breaker and performs time synchronization and noise reduction processing on the multi-modal data; A time series construction module that uses the Kalman filter algorithm to fuse the multi-modal environmental data and constructs multi-dimensional time series data based on the fused state vector; A time series neural network module that uses the TCN network accelerated by photon computing to extract multi-scale time series features of the time series of the fused state vector; A non-linear correction module that calculates the current fractal dimension of the predicted current time series and conducts consistency verification, and non-linearly corrects the current predicted value based on the adjustment factor calculated from the fractal dimension; A reinforcement learning adaptive control module that calculates the current trend correction factor, dimension anomaly index, and magnetic field disturbance index, dynamically adjusts the complexity of the reinforcement learning task, updates the Q-value table through Q-learning, dynamically calculates the optimal strategy, and conducts decision delay verification; An intelligent early warning module that collects the state data of the circuit breaker and conducts state monitoring and fault early warning.

8. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that: When the processor executes the computer program, the steps of the intelligent control method based on the pole-mounted circuit breaker device according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, the steps of the intelligent control method based on the pole-mounted circuit breaker device according to any one of claims 1 to 6 are implemented.

Images