A distribution automation voltage control system

Abstract

The present invention discloses a distribution automation voltage control system, which relates to the field of distribution automation and includes a data acquisition and processing module, a parameter estimation and correction module, a central control execution module, a dynamic response control module, a monitoring fault diagnosis module and a user interface module; the data acquisition and processing module obtains real-time data values ​​in the power grid, and preprocesses the collected real-time data values ​​to obtain preprocessed data and transmit them to the parameter estimation and correction module and the central control execution module; the present invention solves the problem that parameters may change due to changes in the power grid structure, equipment aging, etc. by setting up the parameter estimation and correction module, realizes timely parameter updating, avoids control strategy failure, and improves system operation efficiency and stability; by setting up the dynamic response control module, it solves the situation that transient impacts in the power grid may cause voltage drops or surges; by adjusting the control parameters in advance, the system's impact resistance and response speed are improved.

Description

Technical Field

[0001] The present invention relates to the field of power distribution automation, and in particular to a power distribution automation voltage control system. Background Art

[0002] The development of modern smart grids places higher demands on the stability and reliability of distribution systems. Voltage fluctuations not only affect the normal operation of user equipment but can also lead to equipment failures and energy waste. Therefore, the Distributed Automatic Voltage Control System (DAVCS) plays a crucial role in smart grids. Through automated monitoring and regulation, DAVCS maintains stable and precise voltage levels in the distribution network, improving power supply quality and reliability.

[0003] The existing technology has the following deficiencies: the existing distribution automation voltage control process usually requires grid parameters such as line impedance, load characteristics, and transformer parameters for algorithm design. In actual applications, due to the limited accuracy of sensors and measuring equipment, or insufficient reliability in complex environments, or the influence of changes in environmental factors such as temperature and humidity, there may be a large difference between the collected parameter estimates and the actual values, causing the voltage control to deviate from the target value, affecting the accuracy and control performance of the voltage regulation, and thus affecting the power supply quality; in addition, a large number of users operating high-power electrical appliances at the same time will cause large fluctuations in current and power in a short period of time, generating transient shocks, causing the voltage to drop or rise instantaneously. When there is a short-term high-power shock caused by the uncertainty of user electricity consumption behavior, the system's predictive ability and adaptability cannot respond in time, resulting in difficulty in regulation and control delays, thereby affecting the stability and reliability of the entire distribution network.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0005] The purpose of the present invention is to provide a distribution automation voltage control system, which solves the problems in the above-mentioned background technology.

[0006] In order to achieve the above-mentioned object, the present invention provides the following technical solutions: a distribution automation voltage control system, comprising a data acquisition and processing module, a parameter estimation and correction module, a central control execution module, a dynamic response control module, a monitoring fault diagnosis module and a user interface module;

[0007] The data acquisition and processing module obtains real-time data values ​​in the power grid, and pre-processes the collected real-time data values ​​to obtain pre-processed data and transmits it to the parameter estimation and correction module and the central control execution module;

[0008] The parameter estimation and correction module uses a deep learning algorithm to estimate, analyze and correct the grid parameters in real time on the pre-processed data obtained by the data acquisition and processing module, generates high-precision parameter results and feeds them back to the central control execution module;

[0009] The central control execution module controls the data communication between the modules, and generates the control instructions for the optimal control action based on the comparison of the pre-processed data and the high-precision parameter results with the preset target value of the distribution voltage, and sends the instructions to the dynamic response control module;

[0010] The dynamic response control module receives the control instructions from the central control execution module, combines the user's electricity consumption behavior data, uses the time series prediction algorithm to predict the changes in the user's electricity consumption behavior in advance, and adjusts the control parameters to respond to transient impacts;

[0011] The monitoring fault diagnosis module continues to monitor the grid operation status and system operation data in real time after the control parameters are adjusted, analyzes, identifies and diagnoses fault signals, and feeds back the fault signals to the central control execution module and the user interface module for alarm operator intervention and control;

[0012] The user interface module utilizes a human-computer interaction interface to alert and display the central control execution module and the monitoring fault diagnosis module to control the system operation status data, receive the operator's control instructions, and record the log information of the system detection fault.

[0013] In a preferred embodiment, the data acquisition and processing module uses a smart meter with a built-in low-pass filter to obtain real-time data values ​​in the power grid in real time and calibrate them into , the acquired real-time data values ​​are corrected by using a cyclic redundancy check (CRC) algorithm, and the data is detected for outliers using an isolation forest algorithm to obtain preprocessed data. The cyclic redundancy check (CRC) algorithm is used to arrange the real-time data values ​​in the power grid into a byte sequence according to the timestamp sequence and then convert it into binary form. Polynomial division is then used to generate a CRC check code to verify the comparison result of the isolation forest algorithm on the real-time data values ​​after filtering and transmission, ensuring that accurate and reliable real-time data values ​​in the power grid are obtained. The real-time data values ​​after CRC verification are synchronized using the network time protocol (NTP) to detect outliers. The isolation forest algorithm is used to identify and process outliers to generate preprocessed data.

[0014] In a preferred embodiment, the steps of obtaining the pre-processed data are as follows:

[0015] Using Isolation Forest Algorithm to Calculate the Value from Real-time Data Constructed from the dataset Isolation tree, and calculate each real-time data value The path length of the data point in the isolation tree, where the calculation formula of the path length is as follows: , where Represented as real-time data values exist The path length in the isolated tree, Expressed as the base of natural logarithms;

[0016] Calculate the anomaly score of each data point of the real-time data value, where the calculation formula of the anomaly score is as follows: and , where Represented as real-time data values exist Anomaly scores in isolated trees, Expressed as The average path length of isolated trees;

[0017] Set the outlier score threshold to identify and calibrate abnormal data points as preprocessing data, where the outlier score threshold is set as ,when When the real-time data value is an abnormal data point, that is, the preprocessed data is calibrated as , and the preprocessed data includes .

[0018] In a preferred embodiment, the parameter estimation and correction module uses a deep neural network (DNN) algorithm to perform real-time estimation on the preprocessed data obtained in the data acquisition and processing module, combines historical data and real-time data training models to achieve real-time estimation and correction of power grid parameters, and generates high-precision parameter results.

[0019] In a preferred embodiment, the logic for obtaining the high-precision parameter results is as follows:

[0020] By acquiring the system's long-term operating data, including real-time data and historical data, and dividing it into training sets and test sets, the deep neural network (DNN) model is trained.

[0021] Calculate the input layer, 1 hidden layer, and output layer structure of the DNN model, and adjust the weights through the back-propagation algorithm so that the model can accurately estimate the power grid parameters;

[0022] Will get preprocessed data Input the trained DNN model and preprocess the input data through the DNN model Perform forward propagation and output high-precision parameter results, where the calculation formula of the input layer is: , where preprocessed data Represented as the input data of the input layer, Expressed as the output result of the input layer, the input and output calculation formula of the lth hidden layer is 、 , where Represented as the output of the hidden layer, is represented as the weight matrix of the lth hidden layer, Represents the activation function input data for the lth hidden layer, is represented as the bias vector of the lth hidden layer, Expressed as the activation function of the hidden layer, the input and output calculation formula of the output layer is, 、 , where Represented as the input data of the output layer, is represented as the weight matrix of the output layer, Represented as the output data of the output layer, Represented as the bias vector of the output layer, Represented as the activation function of the output layer, and the output data of the output layer It is the high-precision parameter result;

[0023] According to the output high-precision parameter results , combined with the actual operation of the power grid, parameter correction is performed to obtain high-precision parameter results after correction to ensure the accuracy of the estimated value, and the high-precision parameter results are transmitted to the central control execution module for subsequent voltage control. The steps of parameter correction are as follows:

[0024] By calculating the pre-processed data of the actual operation of the power grid High-precision parameter results output by the DNN model The error value between and the correction factor is calculated, where the calculation formula of the error value is: , where Represented as preprocessed data and high-precision parameter results The error value between the two, the calculation formula of the correction factor is , where Expressed as a correction factor;

[0025] Apply the calculated correction factor to the high-precision parameter result output by the DNN model to perform parameter correction calculation to obtain the corrected high-precision parameter result. The calculation formula for the corrected high-precision parameter result is: , where Expressed as the high-precision parameter result after correction, Expressed as an adjustment coefficient to control the impact of the correction factor.

[0026] In a preferred embodiment, the central control execution module uses a model predictive control (MPC) algorithm to dynamically adjust control parameters based on real-time data values ​​and parameter correction values ​​of high-precision parameter results to ensure that the system always approaches a preset target value, wherein the preset target value is a threshold for voltage stability control in the power grid;

[0027] The logical steps for dynamically adjusting control parameters of the model predictive control MPC algorithm are as follows:

[0028] Construct a dynamic model of the distribution voltage in the power grid, where the construction formula of the dynamic model is: , where and They are respectively represented as real-time data values ​​and high-precision parameter results after CNN model correction. Respectively expressed as The state vector of the power grid operation and the output vector of the system operation in the MPC model at time t, Expressed as The value of the moment, Expressed as the state transfer matrix in the MPC model, Expressed as the control input matrix in the MPC model, Represented as the output matrix in the MPC model, Represented as a direct transposed matrix in the MPC model;

[0029] The high-precision parameter results after the current CNN model correction are used to predict the power grid operation status in the future;

[0030] Define the objective function of the optimized MPC model to solve the optimal control action and dynamically adjust the voltage to obtain the system operating state. The optimization objective function formula of the MPC model is: , where Expressed as the optimization objective function to evaluate the performance of the control action, Represented as the ability to predict future time periods Time value and Time period, Represented as the next The time corresponds to the state vector of the power grid operation and the output vector of the system operation in the MPC model. Indicates the preset target value in the future The vector of time, Expressed as The time interval between time steps, Represented as the next The weight matrix of the output error at time t and the weight matrix of the control input;

[0031] By executing the optimal control actions of the MPC model, the grid status data is obtained and the control effect is evaluated.

[0032] In a preferred embodiment, the dynamic response control module uses a time series prediction algorithm, combined with the control instructions of the optimal control action of the central control execution module and the user's electricity consumption behavior data, to predict changes in the user's electricity consumption behavior in the short term in the future. Adjusting the control parameters in advance can quickly respond to transient shocks. The time series prediction algorithm uses a long short-term memory network LSTM model for time series prediction;

[0033] The prediction logic of the LSTM model is as follows:

[0034] Obtain the control instructions of the optimal control action in the model predictive control MPC algorithm and the user's electricity consumption behavior data, among which the output result of the optimization objective function in the MPC model For control instructions, set the user's electricity behavior data as ;

[0035] The LSTM model predicts the changes in user electricity consumption behavior in the short term in the future, obtains the predicted data of user electricity consumption behavior data, and generates a comprehensive control strategy. The calculation formula of the comprehensive control strategy is: , where Expressed as a comprehensive control strategy, It is expressed as a correction coefficient to adjust the impact of changes in user electricity consumption behavior. It represents the predicted data of the user's electricity consumption behavior data after the LSTM model is trained and predicted on the user's electricity consumption behavior data;

[0036] Apply the generated comprehensive control strategy to the control parameters for real-time adjustment and optimization;

[0037] According to the parameter regulation of the comprehensive control strategy, it can quickly respond to transient impacts.

[0038] In a preferred embodiment, the predicted data of the user's electricity consumption behavior data is the output prediction value result calculated by forward propagation of the LSTM model through the input gate, forget gate, output gate, cell state and hidden state. The specific calculation steps of the predicted data of the user's electricity consumption behavior data are as follows: The input gate calculation formula of the LSTM model is , where Represented as an LSTM model in Input the activation value of the gate at any moment, Represented as the activation function of the LSTM model input gate, The user's electricity consumption behavior data is represented by In the weight matrix of the input gate, Expressed as The weight matrix of the hidden state of the LSTM model at the input gate at this moment, Expressed as The user electricity consumption behavior data input into the LSTM model at any moment is , Expressed as The hidden state of the LSTM model at this moment, Represented as the bias vector of the LSTM model input gate;

[0039] The calculation formula of the forget gate is: , where Represented as an LSTM model in The activation value of the forget gate at any moment, Represented as the activation function of the forget gate of the LSTM model, The user's electricity consumption behavior data is represented by In the weight matrix of the forget gate, Expressed as The weight matrix of the hidden state of the LSTM model in the forget gate at this moment, Expressed as The user electricity consumption behavior data input into the LSTM model at any moment is , Expressed as The hidden state of the LSTM model at this moment, Represented as the bias vector of the forget gate of the LSTM model;

[0040] The cell state calculation formula is: , where Expressed as element-wise multiplication, Represented as in the LSTM model Update cell status at all times, Represented as an LSTM model in The cell state at any moment, Expressed as the hyperbolic tangent activation function, The user's electricity consumption behavior data is represented by The weight matrix in the cell state, Expressed as The weight matrix of the hidden state of the LSTM model in the cell state at this moment, Represented as the bias vector of the LSTM model cell state;

[0041] The output gate calculation formula is , where Represented as an LSTM model in The activation value of the output gate at the moment, Represented as the activation function of the output gate of the LSTM model, The user's electricity consumption behavior data is represented by In the weight matrix of the output gate, Expressed as The weight matrix of the hidden state of the LSTM model at the output gate at this moment, Represented as the bias vector of the output gate of the LSTM model;

[0042] The hidden state calculation formula is: , where Expressed as element-wise multiplication, Expressed as The hidden state of the LSTM model at this moment, Expressed as a hyperbolic tangent activation function;

[0043] The output prediction value calculation formula is: , where Expressed as At this moment, the LSTM model outputs the predicted value. Expressed as The weight matrix of the hidden state of the LSTM model at this moment, Represents the bias vector for the LSTM model output prediction value.

[0044] In a preferred embodiment, the monitoring fault diagnosis module uses a wavelet transform feature extraction algorithm to identify and diagnose existing fault signals by analyzing system operating status data and power grid operating status information. The logical steps of the wavelet transform feature extraction algorithm for diagnosing fault signals are as follows:

[0045] De-noising and normalizing the collected system operation status data and power grid operation status information to obtain pre-processed signals;

[0046] Decompose the preprocessed signal into different frequency bands using continuous wavelet transform;

[0047] And extract fault features from the pre-processed signal after continuous wavelet transform decomposition, and use statistical methods to analyze, identify and classify fault signals;

[0048] The fault diagnosis results are fed back to the central control execution module so that appropriate fault handling measures can be taken.

[0049] In a preferred embodiment, the logical steps of the continuous wavelet transform algorithm for feature extraction are as follows:

[0050] Collect voltage signals from the power grid;

[0051] De-noising and normalizing the voltage signal to obtain a normalized voltage signal;

[0052] Perform continuous wavelet transform decomposition on the normalized voltage signal;

[0053] Calculate the energy, variance and logical entropy of the wavelet coefficients, and input the energy, variance and logical entropy of the wavelet coefficients as feature vectors into the support vector machine (SVM) to classify the extracted features and identify the fault type;

[0054] The fault diagnosis report generated by SVM is transmitted to the central control execution module and user interface module to issue timely alarms and take corresponding fault handling measures.

[0055] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0056] The present invention solves the problem that parameters may change due to changes in the power grid structure, equipment aging, etc. by setting up a parameter estimation and correction module, realizes timely updating of parameters, avoids failure of control strategy, and improves system operation efficiency and stability. By setting up a dynamic response control module, the situation that transient impacts in the power grid may cause voltage drops or surges is solved. By adjusting the control parameters in advance, the system's impact resistance and response speed are improved, the voltage is kept stable while the time of voltage fluctuation is reduced. The setting of a monitoring fault diagnosis module can accurately diagnose the fault type and identify the fault location in real time, thereby improving the system's fault diagnosis capability, reducing the power outage time caused by the fault, further reducing the workload of maintenance personnel, and improving maintenance efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction to the drawings required for use in the embodiments will be given below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0058] Figure 1 This is a module diagram of the power distribution automation voltage control system of the present invention.

[0059] Figure 2 This is the flow chart of the distribution automation voltage control of the present invention. DETAILED DESCRIPTION

[0060] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0061] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the described features. In the description of this application, "plurality" means two or more, unless otherwise specifically specified.

[0062] In the description of this application, the term "for example" is used to mean "used as an example, illustration or explanation". Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any person skilled in the art to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art will recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in this application.

[0063] The present invention provides Figure 1-2 A distribution automation voltage control system shown includes a data acquisition and processing module, a parameter estimation and correction module, a central control execution module, a dynamic response control module, a monitoring fault diagnosis module and a user interface module;

[0064] The data acquisition and processing module obtains the real-time data values ​​of voltage, current, active power, reactive power, temperature and humidity in the power grid, and pre-processes the collected real-time data values ​​to obtain pre-processed data and transmit it to the parameter estimation and correction module and the central control execution module;

[0065] The parameter estimation and correction module uses a deep learning algorithm to estimate, analyze and correct the grid parameters in real time on the pre-processed data obtained by the data acquisition and processing module, generates high-precision parameter results and feeds them back to the central control execution module;

[0066] The central control execution module controls the data communication between the modules, and generates the control instructions for the optimal control action based on the comparison of the pre-processed data and the high-precision parameter results with the preset target value of the distribution voltage, and sends the instructions to the dynamic response control module;

[0067] The dynamic response control module receives the control instructions from the central control execution module, combines the user's electricity consumption behavior data, uses the time series prediction algorithm to predict the changes in the user's electricity consumption behavior in advance, and adjusts the control parameters to respond to transient impacts;

[0068] The monitoring fault diagnosis module continues to monitor the grid operation status and system operation data in real time after the control parameters are adjusted, analyzes, identifies and diagnoses fault signals, and feeds back the fault signals to the central control execution module and the user interface module for alarm operator intervention and control;

[0069] The user interface module utilizes a human-computer interaction interface to alert and display the central control execution module and the monitoring fault diagnosis module to control the system operation status data, receive the operator's control instructions, and record the log information of the system detection fault.

[0070] It should be specifically explained that the data acquisition and processing module uses a smart meter with a built-in low-pass filter to obtain the data values ​​of voltage, current, active power, reactive power, temperature and humidity in the power grid in real time, which are calibrated as , the real-time data value obtained is corrected by using the cyclic redundancy check CRC algorithm, where the real-time data value is calibrated as , and real-time data values ​​include voltage, current, active power, reactive power, temperature and humidity data values , that is, real-time data value , and use the isolation forest algorithm to detect outliers on the data to obtain preprocessed data, wherein the cyclic redundancy check CRC algorithm is used to arrange the real-time data values ​​in the power grid into a byte sequence according to the timestamp sequence and then convert it into binary form, and then use polynomial division to generate a CRC check code to verify the comparison result of the isolation forest algorithm on the real-time data value after filtering and transmission, ensuring that accurate and reliable real-time data values ​​in the power grid are obtained, and the real-time data value after CRC verification is synchronized using the network time protocol NTP to perform outlier detection, and the isolation forest algorithm is used to identify and process outliers to generate preprocessed data. The steps for obtaining the preprocessed data are as follows:

[0071] Using Isolation Forest Algorithm to Calculate the Value from Real-time Data Constructed from the dataset Isolation tree, and calculate each real-time data value The path length of the data point in the isolation tree, where the calculation formula of the path length is as follows: , where Represented as real-time data values exist The path length in the isolated tree, Expressed as the base of natural logarithms;

[0072] Calculate the anomaly score of each data point of the real-time data value, where the calculation formula of the anomaly score is as follows: and , where Represented as real-time data values exist Anomaly scores in isolated trees, Expressed as The average path length of isolated trees;

[0073] Set the outlier score threshold to identify and calibrate abnormal data points as preprocessing data, where the outlier score threshold is set as ,when When the real-time data value is an abnormal data point, that is, the preprocessed data is calibrated as , and the preprocessed data includes .

[0074] It should be noted that the parameter estimation and correction module uses the deep neural network (DNN) algorithm to perform real-time estimation on the pre-processed data obtained in the data acquisition and processing module. It combines historical data and real-time data training models to achieve real-time estimation and correction of power grid parameters and generate high-precision parameter results. The logic for obtaining high-precision parameter results is as follows:

[0075] By acquiring the system's long-term operating data, including real-time data and historical data, and dividing it into training sets and test sets, the deep neural network (DNN) model is trained.

[0076] Calculate the input layer, 1 hidden layer, and output layer structure of the DNN model, and adjust the weights through the back-propagation algorithm so that the model can accurately estimate the power grid parameters;

[0077] Will get preprocessed data Input the trained DNN model and preprocess the input data through the DNN model Perform forward propagation and output high-precision parameter results, where the calculation formula of the input layer is: , where preprocessed data Represented as the input data of the input layer, Expressed as the output result of the input layer, the input and output calculation formula of the lth hidden layer is 、 , where Represented as the output of the hidden layer, is represented as the weight matrix of the lth hidden layer, Represents the activation function input data for the lth hidden layer, is represented as the bias vector of the lth hidden layer, Expressed as the activation function of the hidden layer, the input and output calculation formula of the output layer is, 、 , where Represented as the input data of the output layer, is represented as the weight matrix of the output layer, Represented as the output data of the output layer, Represented as the bias vector of the output layer, Represented as the activation function of the output layer, and the output data of the output layer It is the high-precision parameter result;

[0078] According to the output high-precision parameter results , combined with the actual operation of the power grid, parameter correction is performed to obtain high-precision parameter results after correction to ensure the accuracy of the estimated value, and the high-precision parameter results are transmitted to the central control execution module for subsequent voltage control. The steps of parameter correction are as follows:

[0079] By calculating the pre-processed data of the actual operation of the power grid High-precision parameter results output by the DNN model The error value between and the correction factor is calculated, where the calculation formula of the error value is: , where Represented as preprocessed data and high-precision parameter results The error value between the two, the calculation formula of the correction factor is , where Expressed as a correction factor;

[0080] Apply the calculated correction factor to the high-precision parameter result output by the DNN model to perform parameter correction calculation to obtain the corrected high-precision parameter result. The calculation formula for the corrected high-precision parameter result is: , where Expressed as the high-precision parameter result after correction, Expressed as an adjustment coefficient to control the impact of the correction factor.

[0081] It should be noted that the real-time estimation steps of the deep neural network DNN algorithm are as follows:

[0082] Extracting real-time data values ​​of voltage, current, active power, reactive power, temperature and humidity from the long-term operation data of the system as feature data;

[0083] The feature data is divided into training sets and test sets, and applied to the DNN model for training. The weights and biases are adjusted through the back-propagation algorithm to minimize the loss function.

[0084] Then the preprocessed data Input into the trained DNN model for real-time estimation calculation to obtain the parameter values ​​predicted by the DNN model;

[0085] The parameter values ​​predicted by the DNN model are then calibrated to generate high-precision parameter results.

[0086] It should be noted that the central control execution module uses the model predictive control (MPC) algorithm to dynamically adjust control parameters based on real-time data values ​​and the parameter correction values ​​of high-precision parameter results to ensure that the system always approaches the preset target value. The preset target value is the threshold for voltage stability regulation in the power grid.

[0087] The logical steps for dynamically adjusting control parameters of the model predictive control MPC algorithm are as follows:

[0088] Construct a dynamic model of the distribution voltage in the power grid, where the construction formula of the dynamic model is: , where and They are respectively represented as real-time data values ​​and high-precision parameter results after CNN model correction. Respectively expressed as The state vector of the power grid operation and the output vector of the system operation in the MPC model at time t, Expressed as The value of the moment, Expressed as the state transfer matrix in the MPC model, Expressed as the control input matrix in the MPC model, Represented as the output matrix in the MPC model, Represented as a direct transposed matrix in the MPC model;

[0089] The high-precision parameter results after the current CNN model correction are used to predict the power grid operation status in the future;

[0090] Define the objective function of the optimized MPC model to solve the optimal control action and dynamically adjust the voltage to obtain the system operating state. The optimization objective function formula of the MPC model is: , where Expressed as the optimization objective function to evaluate the performance of the control action, Represented as the ability to predict future time periods Time value and Time period, Represented as the next The time corresponds to the state vector of the power grid operation and the output vector of the system operation in the MPC model. Indicates the preset target value in the future The vector of time, Expressed as The time interval between time steps, Represented as the next The weight matrix of the output error at time t and the weight matrix of the control input;

[0091] By executing the optimal control actions of the MPC model, the grid status data is obtained and the control effect is evaluated.

[0092] It should be noted that the dynamic response control module uses a time series prediction algorithm, combined with the control instructions of the optimal control action of the central control execution module and the user's electricity consumption behavior data, to predict changes in user electricity consumption behavior in the short term in the future. Adjusting the control parameters in advance can quickly respond to transient shocks. The time series prediction algorithm uses the long short-term memory network (LSTM) model for time series prediction.

[0093] It should be noted that the prediction logic of the LSTM model is as follows:

[0094] Obtain the control instructions of the optimal control action in the model predictive control MPC algorithm and the user's electricity consumption behavior data, among which the output result of the optimization objective function in the MPC model For control instructions, set the user's electricity behavior data as ;

[0095] The LSTM model predicts the changes in user electricity consumption behavior in the short term in the future, obtains the predicted data of user electricity consumption behavior data, and generates a comprehensive control strategy. The calculation formula of the comprehensive control strategy is: , where Expressed as a comprehensive control strategy, It is expressed as a correction coefficient to adjust the impact of changes in user electricity consumption behavior. It is represented as the predicted data of user electricity consumption behavior data after LSTM model is trained and predicted on user electricity consumption behavior data, and the predicted data of user electricity consumption behavior data is the output prediction value result calculated by LSTM model through input gate, forget gate, output gate, cell state and hidden state. The specific calculation steps of the predicted data of the user's electricity consumption behavior data are as follows: The input gate calculation formula of the LSTM model is , where Represented as an LSTM model in Input the activation value of the gate at any moment, Represented as the activation function of the LSTM model input gate, The user's electricity consumption behavior data is represented by In the weight matrix of the input gate, Expressed as The weight matrix of the hidden state of the LSTM model at the input gate at this moment, Expressed as The user electricity consumption behavior data input into the LSTM model at any moment is , Expressed as The hidden state of the LSTM model at this moment, Represented as the bias vector of the LSTM model input gate;

[0096] The calculation formula of the forget gate is: , where Represented as an LSTM model in The activation value of the forget gate at any moment, Represented as the activation function of the forget gate of the LSTM model, The user's electricity consumption behavior data is represented by In the weight matrix of the forget gate, Expressed as The weight matrix of the hidden state of the LSTM model in the forget gate at this moment, Expressed as The user electricity consumption behavior data input into the LSTM model at any moment is , Expressed as The hidden state of the LSTM model at this moment, Represented as the bias vector of the forget gate of the LSTM model;

[0097] The cell state calculation formula is: , where Expressed as element-wise multiplication, Represented as in the LSTM model Update cell status at all times, Represented as an LSTM model in The cell state at any moment, Expressed as the hyperbolic tangent activation function, The user's electricity consumption behavior data is represented by The weight matrix in the cell state, Expressed as The weight matrix of the hidden state of the LSTM model in the cell state at this moment, Represented as the bias vector of the LSTM model cell state;

[0098] The output gate calculation formula is , where Represented as an LSTM model in The activation value of the output gate at the moment, Represented as the activation function of the output gate of the LSTM model, The user's electricity consumption behavior data is represented by In the weight matrix of the output gate, Expressed as The weight matrix of the hidden state of the LSTM model at the output gate at this moment, Represented as the bias vector of the output gate of the LSTM model;

[0099] The hidden state calculation formula is: , where Expressed as element-wise multiplication, Expressed as The hidden state of the LSTM model at this moment, Expressed as a hyperbolic tangent activation function;

[0100] The output prediction value calculation formula is: , where Expressed as At this moment, the LSTM model outputs the predicted value. Expressed as The weight matrix of the hidden state of the LSTM model at this moment, Represents the bias vector for the LSTM model output prediction value;

[0101] Apply the generated comprehensive control strategy to the control parameters for real-time adjustment and optimization;

[0102] According to the parameter regulation of the comprehensive control strategy, it can quickly respond to transient impacts and ensure the stability and accuracy of the power grid operation status.

[0103] It should be specifically noted that the monitoring fault diagnosis module uses a wavelet transform feature extraction algorithm to identify and diagnose existing fault signals by analyzing system operating status data and power grid operating status information. The logical steps of the wavelet transform feature extraction algorithm for diagnosing fault signals are as follows:

[0104] De-noising and normalizing the collected system operation status data and power grid operation status information to obtain pre-processed signals;

[0105] Decompose the preprocessed signal into different frequency bands using continuous wavelet transform;

[0106] And extract fault features from the pre-processed signal after continuous wavelet transform decomposition, and use statistical methods to analyze, identify and classify fault signals;

[0107] The fault diagnosis results are fed back to the central control execution module so that appropriate fault handling measures can be taken.

[0108] It should be specifically noted that the logical steps of the continuous wavelet transform algorithm for feature extraction are as follows:

[0109] Collect voltage signals from the power grid;

[0110] De-noising and normalizing the voltage signal to obtain a normalized voltage signal;

[0111] Perform continuous wavelet transform decomposition on the normalized voltage signal;

[0112] Calculate the energy, variance and logical entropy of the wavelet coefficients, and input the energy, variance and logical entropy of the wavelet coefficients as feature vectors into the support vector machine (SVM) to classify the extracted features and identify the fault type;

[0113] The fault diagnosis report generated by SVM is transmitted to the central control execution module and user interface module to issue timely alarms and take corresponding fault handling measures.

[0114] The specific method and process of a distribution automation voltage control system provided by an embodiment of the present invention are detailed in the above-mentioned embodiment of a distribution automation voltage control system, which will not be repeated here.

[0115] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0116] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0117] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0118] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0119] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A distribution automation voltage control system, characterized in that: It includes data acquisition and processing module, parameter estimation and correction module, central control execution module, dynamic response control module, monitoring fault diagnosis module and user interface module; The data acquisition and processing module obtains real-time data values ​​in the power grid, and pre-processes the collected real-time data values ​​to obtain pre-processed data and transmits it to the parameter estimation and correction module and the central control execution module; The parameter estimation and correction module uses a deep learning algorithm to estimate, analyze and correct the grid parameters in real time on the pre-processed data obtained by the data acquisition and processing module, generates high-precision parameter results and feeds them back to the central control execution module; The central control execution module controls the data communication between the modules, and generates the control instructions for the optimal control action based on the comparison of the pre-processed data and the high-precision parameter results with the preset target value of the distribution voltage, and sends the instructions to the dynamic response control module; The central control execution module uses the model predictive control (MPC) algorithm to dynamically adjust control parameters based on real-time data values ​​and parameter correction values ​​of high-precision parameter results to ensure that the system always approaches a preset target value, wherein the preset target value is a threshold for voltage stability regulation in the power grid. The steps of dynamically adjusting control parameters of the model predictive control (MPC) algorithm are as follows: Construct a dynamic model of distribution voltage in the power grid; The high-precision parameter results after the current CNN model correction are used to predict the power grid operation status in the future; Define the objective function of the optimized MPC model to solve the optimal control action and dynamically adjust the voltage to obtain the system operating state; By executing the optimal control actions of the MPC model, the grid status data is obtained and the control effect is evaluated; The dynamic response control module receives the control instructions from the central control execution module, combines the user's electricity consumption behavior data, uses the time series prediction algorithm to predict the changes in the user's electricity consumption behavior in advance, and adjusts the control parameters to respond to transient impacts; The monitoring fault diagnosis module continues to monitor the grid operation status and system operation data in real time after the control parameters are adjusted, analyzes, identifies and diagnoses fault signals, and feeds back the fault signals to the central control execution module and the user interface module for alarm operator intervention and control; The user interface module utilizes a human-computer interaction interface to alert and display the central control execution module and the monitoring fault diagnosis module to control the system operation status data, receive the operator's control instructions, and record the log information of the system detection fault.

2. A distribution automation voltage control system according to claim 1, characterized in that: The data acquisition and processing module uses a smart meter with a low-pass filter to obtain real-time data values ​​in the power grid in real time and calibrates them as D I , the acquired real-time data values ​​are corrected by using a cyclic redundancy check (CRC) algorithm, and the data is detected for outliers using an isolation forest algorithm to obtain preprocessed data. The cyclic redundancy check (CRC) algorithm is used to arrange the real-time data values ​​in the power grid into a byte sequence according to the timestamp sequence and then convert it into binary form. Polynomial division is then used to generate a CRC check code to verify the comparison result of the isolation forest algorithm on the real-time data values ​​after filtering and transmission, ensuring that accurate and reliable real-time data values ​​in the power grid are obtained. The real-time data values ​​after CRC verification are synchronized using the network time protocol (NTP) to detect outliers. The isolation forest algorithm is used to identify and process outliers to generate preprocessed data.

3. A distribution automation voltage control system according to claim 2, characterized in that: The steps for obtaining the preprocessed data are as follows: Using the isolation forest algorithm to calculate the value D from real-time data I Construct N isolated trees in the data set and calculate each real-time data value D I The path length of the data point in the isolation tree, where the calculation formula of the path length is as follows: Where, Expressed as real-time data value D I The path length in N isolated trees, e, is expressed as the base of the natural logarithm; Calculate the anomaly score of each data point of the real-time data value, where the calculation formula of the anomaly score is as follows: and Where, S(D I , N) is represented by the real-time data value D I The anomaly score among N isolated trees, d(N) is expressed as the average path length of N isolated trees; Set the abnormal value score threshold, identify and calibrate abnormal data points as preprocessing data, where the abnormal value score threshold is set as T H , when S(D I , N)>T H When the real-time data value D I is an abnormal data point, that is, the preprocessed data is marked as Dc, and the preprocessed data includes Dc={S(D I , N)>T H }.

4. A distribution automation voltage control system according to claim 3, characterized in that: The parameter estimation and correction module uses a deep neural network (DNN) algorithm to perform real-time estimation on the pre-processed data obtained in the data acquisition and processing module, combines historical data with real-time data training models, realizes real-time estimation and correction of power grid parameters, and generates high-precision parameter results.

5. A distribution automation voltage control system according to claim 4, characterized in that: The logic for obtaining the high-precision parameter results is as follows: By acquiring the system's long-term operating data, including real-time data and historical data, and dividing it into training sets and test sets, the deep neural network (DNN) model is trained. Calculate the input layer, 1 hidden layer, and output layer structure of the DNN model, and adjust the weights through the back-propagation algorithm so that the model can accurately estimate the power grid parameters; The preprocessed data Dc is input into the trained DNN model, and the input preprocessed data Dc is forward propagated through the DNN model to output high-precision parameter results. The calculation formula of the input layer is a0=Dc, where the preprocessed data Dc is represented as the input data of the input layer, a0 is represented as the output result of the input layer, and the input and output calculation formula of the lth hidden layer is z l =W l ×a l-1 +b l 、a l =f(z l ), where z l Represented as the output of the hidden layer, W l Represented as the weight matrix of the lth hidden layer, a l-1 Represents the activation function input data for the lth hidden layer, b l It is represented as the bias vector of the lth hidden layer, f is represented as the activation function of the hidden layer, and the input and output calculation formula of the output layer is z′=W′×a ′-1 +b′, y′=f′(z′), where z′ represents the input data of the output layer, W′ represents the weight matrix of the output layer, y′ represents the output data of the output layer, and b l Represented as the bias vector of the output layer, f′ represents the activation function of the output layer, and the output data y′ of the output layer is the high-precision parameter result; Based on the output high-precision parameter result y′, combined with the actual operation of the power grid, parameter correction is performed to obtain the corrected high-precision parameter result to ensure the accuracy of the estimated value, and the high-precision parameter result is transmitted to the central control execution module for subsequent voltage control. The steps of parameter correction are as follows: The error between the pre-processed data Dc of the actual operation of the power grid and the high-precision parameter result y′ output by the DNN model is calculated, and the correction factor is obtained. The calculation formula of the error value is ei = Dc - y′, where ei represents the error between the pre-processed data Dc and the high-precision parameter result y′. The calculation formula of the correction factor is ci = ei / y′, where ci represents the correction factor. Apply the calculated correction factor to the high-precision parameter result output by the DNN model to perform parameter correction calculation to obtain the corrected high-precision parameter result. The calculation formula for the corrected high-precision parameter result is y″=y′+α×ci, where y″ represents the corrected high-precision parameter result and α represents the adjustment coefficient for controlling the influence of the correction factor.

6. A distribution automation voltage control system according to claim 5, characterized in that: The logical steps of dynamically adjusting control parameters of the model predictive control MPC algorithm are as follows: Construct a dynamic model of the distribution voltage in the power grid, where the construction formula of the dynamic model is: Where Dc and y″ represent the real-time data value and the high-precision parameter result after CNN model correction, respectively. t+1 、Y t+1 They are respectively represented as the state vector of the power grid operation and the output vector of the system operation in the MPC model at time t, t represents the value at time t, A represents the state transfer matrix in the MPC model, B represents the control input matrix in the MPC model, C represents the output matrix in the MPC model, and D represents the direct transposed matrix in the MPC model; The high-precision parameter results after the current CNN model correction are used to predict the power grid operation status in the future; Define the objective function of the optimized MPC model to solve the optimal control action and dynamically adjust the voltage to obtain the system operating state. The optimization objective function formula of the MPC model is: Where Q is the optimization objective function for evaluating the performance of the control action, m and M are the values ​​at time m and time period M that can be predicted in the future, respectively, and X t+m 、Y t+m They are respectively represented as the state vector of the power grid operation and the output vector of the system operation in the MPC model at the future time t+m, R t+m Represented as a vector of preset target values ​​at the future t+m time, |t| It is represented as the time interval of t time steps, w′ and w″ are the weight matrix of output error and control input at the future time t+m respectively; By executing the optimal control actions of the MPC model, the grid status data is obtained and the control effect is evaluated.

7. A distribution automation voltage control system according to claim 6, characterized in that: The dynamic response control module uses a time series prediction algorithm, combined with the control instructions of the optimal control action of the central control execution module and the user's electricity consumption behavior data, to predict changes in user electricity consumption behavior in the short term in the future. Adjusting the control parameters in advance can quickly respond to transient shocks. The time series prediction algorithm uses a long short-term memory network (LSTM) model for time series prediction; The prediction logic of the LSTM model is as follows: Obtain the control instruction of the optimal control action in the model predictive control (MPC) algorithm and the user's electricity consumption behavior data, wherein the output result Q of the optimization objective function in the MPC model is the control instruction, and the user's electricity consumption behavior data is set as U; The LSTM model predicts changes in user electricity usage behavior in the short term, obtains predicted data for user electricity usage behavior data, and generates a comprehensive control strategy. The calculation formula for the comprehensive control strategy is G = Q + μ × (U′ - U), where G represents the comprehensive control strategy, μ represents the correction coefficient for adjusting the impact of changes in user electricity usage behavior, and U′ represents the predicted data of user electricity usage behavior data after the LSTM model is trained and predicted on the user electricity usage behavior data. Apply the generated comprehensive control strategy to the control parameters for real-time adjustment and optimization; According to the parameter regulation of the comprehensive control strategy, it can quickly respond to transient impacts.

8. A distribution automation voltage control system according to claim 7, characterized in that: The predicted data of the user's electricity consumption behavior data is the output prediction value result calculated by the LSTM model forward propagating through the input gate, forget gate, output gate, cell state and hidden state. i The calculation steps of the prediction data of the user's electricity consumption behavior data are as follows: The input gate calculation formula of the LSTM model is i k =ζ i (ω i k U k +υ i k h k-1 +b i k ), where i k It is represented by the activation value of the input gate of the LSTM model at time k, ζ i Represented as the activation function of the LSTM model input gate, ω i k The user's electricity consumption behavior data is represented as the weight matrix U at the input gate, υ i k It is represented as the weight matrix of the hidden state of the LSTM model at the input gate at k-1 time, U k The user electricity consumption behavior data input into the LSTM model at time k is represented as U, h k-1 Represented as the hidden state of the LSTM model at k-1 time, b i k Represented as the bias vector of the LSTM model input gate; The calculation formula of the forget gate is g k =ζ g (ω g k U k +υ g k h k-1 +b g k ), where g k It is expressed as the activation value of the forget gate of the LSTM model at time k, ζ g Represented as the activation function of the forget gate of the LSTM model, ω g k The user's electricity consumption behavior data is represented as the weight matrix U in the forget gate, υ g k It is represented as the weight matrix of the hidden state of the LSTM model in the forget gate at the k-1 moment, U k The user electricity consumption behavior data input into the LSTM model at time k is represented as U, h k-1 Represented as the hidden state of the LSTM model at k-1 time, b g k Represented as the bias vector of the forget gate of the LSTM model; The cell state calculation formula is: Where, Expressed as element-wise multiplication, Cc k Represented as the updated cell state of the LSTM model at time k, Cc k-1 It represents the cell state of the LSTM model at the k-1 moment, and tanh represents the hyperbolic tangent activation function. It is represented as the weight matrix of the user's electricity consumption behavior data U in the cell state, It is represented as the weight matrix of the hidden state of the LSTM model in the cell state at k-1 time. Represented as the bias vector of the LSTM model cell state; The output gate calculation formula is o k =ζ o (ω o k U k +υ o k h k-1 +b o k ), where o k It is represented by the activation value of the output gate of the LSTM model at time k, ζ o Represented as the activation function of the output gate of the LSTM model, ω o k The user's electricity consumption behavior data is represented as the weight matrix U at the output gate, υ o k It is represented as the weight matrix of the hidden state of the LSTM model at the output gate at the k-1 moment, b o k Represented as the bias vector of the output gate of the LSTM model; The hidden state calculation formula is: Where, Expressed as element-wise multiplication, h k It represents the hidden state of the LSTM model at time k, and tanh represents the hyperbolic tangent activation function; The output prediction value calculation formula is p k =ω p k h k +b p k , where p k Represented as the output prediction value of the LSTM model at time k, ω p k Represented as the weight matrix of the hidden state of the LSTM model at time k, b p k Represents the bias vector for the LSTM model output prediction value.

9. A distribution automation voltage control system according to claim 8, characterized in that: The monitoring fault diagnosis module uses a wavelet transform feature extraction algorithm to identify and diagnose existing fault signals by analyzing system operation status data and power grid operation status information. The logical steps of the wavelet transform feature extraction algorithm for diagnosing fault signals are as follows: De-noising and normalizing the collected system operation status data and power grid operation status information to obtain pre-processed signals; Decompose the preprocessed signal into different frequency bands using continuous wavelet transform; And extract fault features from the pre-processed signal after continuous wavelet transform decomposition, and use statistical methods to analyze, identify and classify fault signals; The fault diagnosis results are fed back to the central control execution module so that appropriate fault handling measures can be taken.

10. A distribution automation voltage control system according to claim 9, characterized in that: The logical steps of the continuous wavelet transform algorithm for feature extraction are as follows: Collect voltage signals from the power grid; De-noising and normalizing the voltage signal to obtain a normalized voltage signal; Perform continuous wavelet transform decomposition on the normalized voltage signal; Calculate the energy, variance and logical entropy of the wavelet coefficients, and input the energy, variance and logical entropy of the wavelet coefficients as feature vectors into the support vector machine (SVM) to classify the extracted features and identify the fault type; The fault diagnosis report generated by SVM is transmitted to the central control execution module and user interface module to issue timely alarms and take corresponding fault handling measures.

Images