Intelligent power distribution network first-aid repair scheduling system

Abstract

The invention belongs to the field of power grid maintenance, and particularly relates to an intelligent power distribution network first-aid repair scheduling system. The system is composed of a data acquisition module, a data preprocessing module, a data analysis module, a data updating module and a display module, the data acquisition module monitors multi-parameter data of each node of the power distribution network in real time through a multi-parameter sensor, and the data acquired by the data acquisition module is transmitted to the data preprocessing module in real time in a wireless or wired mode. The data preprocessing module cleans and standardizes multi-parameter data, the preprocessed multi-parameter data is subjected to model training through the data analysis module, a model comprises an LSTM and an RNN, meanwhile, the data updating module adopts an Adam optimization algorithm to update parameters of the RNN of the data analysis module, and the data analysis module outputs the specific position of a fault. The display module is used for displaying, so that the repair process is completed.

Description

Technical Field

[0001] The invention belongs to the field of power grid maintenance, and in particular relates to an intelligent power distribution network emergency repair dispatching system. Background Art

[0002] The distribution network refers to the power grid that receives electric energy from the transmission network or regional power plants and distributes it locally or step by step according to voltage to various users through distribution facilities. The distribution network maintains the stable life of residents and the stable production of industry, so the stable operation of the distribution network is the basis of people's livelihood. Although the distribution network is relatively stable in daily life, it is expected that the distribution network will fail in extreme situations, so controlling the efficiency of the emergency repair and dispatching of the distribution network is a major issue to stabilize people's livelihood.

[0003] Although the existing distribution network emergency repair and dispatching system has improved the emergency efficiency through manual repair reporting, data analysis and other means, it faces severe challenges: first, the response mechanism is lengthy and relies on manual reporting step by step, which significantly delays the timeliness of fault handling; second, the fault location technology is lagging behind and lacks accuracy, often requiring manual field investigation, which is costly and inefficient; third, the level of intelligence is limited. In the face of complex fault scenarios, the system lacks independent analysis and decision-making capabilities, and it is difficult to quickly generate efficient emergency repair strategies; fourth, the barriers to data sharing are high, which hinders the flow of information and coordinated operations across departments and regions, and limits the optimization of emergency repair efficiency and resource allocation. Therefore, the present invention provides an intelligent distribution network emergency repair and dispatching system to improve the intelligence, precision and coordination level of distribution network emergency repair and dispatching. Summary of the invention

[0004] The present invention provides an intelligent power distribution network emergency repair dispatching system, which is used to solve the technical problems of a traditional power distribution network emergency repair dispatching system, such as a lengthy response mechanism, lagging fault location technology, insufficient accuracy and low intelligence level.

[0005] In order to achieve the above-mentioned purpose, the technical scheme adopted by the present invention is as follows: an intelligent distribution network emergency repair and dispatching system, the system comprising: a data acquisition module, a data preprocessing module, a data analysis module, a data update module and a display module, the data acquisition module monitors the multi-parameter data of each node of the distribution network in real time through a multi-parameter sensor, the multi-parameter data collected by the data acquisition module is transmitted to the data preprocessing module in real time by wireless or wired means, the data preprocessing module preprocesses the multi-parameter data, the preprocessing includes cleaning and standardization, the multi-parameter data after the preprocessing is trained by the data analysis module for model, the model includes LSTM and RNN, at the same time, the data update module adopts the Adam optimization algorithm to update the parameters of the RNN, the data analysis module analyzes the multi-parameter data in real time, and the model outputs the prediction result of the fault location, thereby determining the specific location of the fault, and displaying it through the display module, thereby completing the emergency repair process.

[0006] Furthermore, the data acquisition module uses a standard API interface to connect to the multi-parameter sensor monitoring network, and uses RFID tags to identify and track power equipment at each node in the monitoring distribution network to ensure rapid positioning during emergency repairs.

[0007] Furthermore, the multi-parameter data of each node in the distribution network includes voltage, current, temperature, humidity, power factor, harmonic content, electromagnetic intensity, and frequency.

[0008] Furthermore, the cleaning is to identify and process outliers using a standard deviation method.

[0009] Furthermore, the standardization is to standardize the multi-parameter data by improving the Z-score standardization so that they can be compared and analyzed on the same scale.

[0010] Furthermore, the standardized multi-parameter data is normalized to convert the multi-parameter data into a dimensionless value between 0 and 1, and a minimum-maximum normalization formula is obtained.

[0011] Furthermore, the LSTM model in the data analysis module includes a cell state and a hidden state, the cell state is responsible for storing and transmitting information, and the hidden state is responsible for extracting and outputting information, and the RNN model includes an input layer, a hidden layer, and an output layer.

[0012] Furthermore, the data analysis module performs model training, and the model includes LSTM and RNN. The training uses the training set data to train the model, calculates the loss through forward propagation, and then updates the RNN model parameters through back propagation and gradient descent, and evaluates the model performance on the validation set to avoid overfitting, and adjusts the model structure and hyperparameters as needed.

[0013] Furthermore, the specific process of the Adam optimization algorithm is as follows:

[0014] (1) Initialization: Set data x i The initial learning rate α, the exponential decay rate β1 of the first-order moment estimate, the exponential decay rate β2 of the second-order moment estimate, and the fraction ε, which is used to prevent the denominator from being zero in the implementation;

[0015] (2) Calculate gradients: In each iteration, use the mini-batch gradient descent method to calculate the gradient of the loss function with respect to each parameter;

[0016] Calculate the estimates of the first-order moment and the second-order moment: Calculate the estimate of the mean of the first-order moment and the variance of the second-order moment of the gradient by the exponential decay averaging method;

[0017] (3) Correction bias: Since the momentum value is zero at the beginning, the estimation of the first and second moments is adjusted by bias correction;

[0018] (4) Update parameters: Update the parameters based on the corrected first-order moment and second-order moment estimates and the learning rate.

[0019] Furthermore, the specific location of the fault is output on the display module, and the display module includes a cloud platform, which can collect feedback data according to the actual results of the fault repair, analyze the difference between the predicted results and the actual results, and evaluate the model performance.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] (1) The present invention adopts a fault location model that integrates a deep learning algorithm. Through comprehensive analysis of historical fault data and real-time monitoring data, the fault location can be quickly and accurately located, thereby reducing on-site investigation time.

[0022] (2) The dispatching system of the present invention has the ability of self-learning and optimization. By continuously accumulating emergency repair cases and data, it continuously optimizes the fault prediction model and dispatching strategy, thereby improving the overall intelligence level.

[0023] (3) The present invention realizes seamless connection and sharing of data between different systems through open data interfaces and cloud platforms, promotes cross-departmental and cross-regional collaborative repair operations, and improves overall emergency response capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is a structural diagram of the energy distribution network emergency repair dispatching system of the present invention. DETAILED DESCRIPTION

[0025] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0026] An intelligent power distribution network emergency repair dispatching system is composed of a data acquisition module, a data preprocessing module, a data analysis module, a data update module, and a display module. The data acquisition module monitors the multi-parameter data of each node of the power distribution network in real time through a multi-parameter sensor. The data collected by the data acquisition module is transmitted to the data preprocessing module in real time through a wireless or wired manner. The data preprocessing module cleans and standardizes the multi-parameter data. The preprocessed multi-parameter data is used for model training through the data analysis module. The model includes LSTM+RNN. At the same time, the data update module uses the Adam optimization algorithm to update the parameters of the data analysis model RNN. The data analysis module analyzes the multi-parameter data in real time, and the model outputs the prediction result of the fault position, thereby determining the specific position of the fault, and displaying it through the display module, thereby completing the emergency repair process.

[0027] Data acquisition module: Multi-parameter sensors can collect the voltage u of each node in the distribution network in real time i 、Current I i , Temperature T i 、Humidity i , Power factor P i , Harmonic content F i , electromagnetic intensity Q i , frequency f i Key parameters, these data are instantly transmitted to the data pre-processing module through wireless communication networks (such as 4G / 5G, LoRa, NB-IoT, etc.). RFID tags are used to identify and track power equipment at each node of the monitoring distribution network to ensure that the required information can be quickly located and obtained during the emergency repair process.

[0028] Data preprocessing module: (1) Data cleaning: By setting reasonable thresholds and filtering algorithms, noise and outliers are removed to ensure the accuracy and reliability of the data.

[0029] Use the standard deviation method to identify and handle outliers. For each data point x i (Voltage u i 、Current I i , Temperature T i 、Humidity i , Power factor P i , Harmonic content F i , electromagnetic intensity Q i , frequency f i , x i =ω u ·u i +ω I I i +ω T ·T i +ω W ·W i +ωP ·P i +ω F ·F i +ω Q Q i +ω f ·f i , i∈{1,2,3...,N}), where ω xx is the normalized weight value. Calculate its average value The difference χ i =x i -u, and with standard deviation Make a comparison.

[0030] If the difference χ i If the value of -s exceeds the threshold m=3s (3 times the standard deviation), the data point is considered an outlier and is removed or replaced.

[0031] (2) Data standardization: Data from different sources and units are standardized so that they can be compared and analyzed on the same scale.

[0032] Data normalization method: Improved Z-score normalization.

[0033]

[0034] Among them, x is the multi-parameter data after data cleaning, μ is the mean of the data, and σ is the standard deviation of the data. After standardization, the mean of the data is 0 and the standard deviation is 1.

[0035] (3) Data normalization: Through the normalization algorithm, the data is converted into dimensionless values ​​between 0 and 1, which facilitates the processing and analysis of subsequent algorithms.

[0036] Minimum-maximum normalization formula:

[0037] Among them, x is the multi-parameter data after data cleaning and standardization, xmin and xmax are the minimum and maximum values ​​in the data respectively. After normalization, the value range of the data is between 0 and 1.

[0038] (4) Missing value filling: Use linear interpolation to fill in the missing value x m , whose two non-missing values ​​are x a and x b , and a<m<b, then

[0039] Data Analysis Module:

[0040] (1) Selecting an RNN model

[0041] A suitable RNN variant, LSTM (Long Short-Term Memory Networks), was chosen as they are able to handle long-term dependencies and are suitable for time series data.

[0042] Here we choose LSTM to process the above data. LSTM God General Network:

[0043] LSTM (Long Short-Term Memory) is a commonly used variant of recurrent neural network (RNN) for processing sequence data. Compared with traditional RNN, LSTM introduces a gating mechanism, which can better solve the gradient vanishing and gradient exploding problems in traditional RNN and has stronger memory capacity.

[0044] LSTM controls the flow of information and the update of memory through forget gate, input gate and output gate. Each LSTM unit contains a cell state and a hidden state. The cell state is responsible for storing and transmitting information, and the hidden state is responsible for extracting and outputting information.

[0045] Specifically, the operation process of the LSTM unit is as follows:

[0046] Forget gate: determines whether to forget previous information from the cell state. It inputs the input of the current time step and the hidden state of the previous time step, passes through a sigmoid activation function, and outputs a value between 0 and 1, indicating the degree of information retention at the corresponding position.

[0047] Input gate: determines the impact of the input of the current time step on the cell state. It inputs the input of the current time step and the hidden state of the previous time step, passes through a sigmoid activation function, and outputs a value between 0 and 1, indicating the importance of the input at the corresponding position.

[0048] Update cell state: Update the cell state according to the forget gate, input gate and the input of the current time step. First, the forget gate is used to control the degree of forgetting of the previous cell state, and then the input gate is used to control the degree of updating of the cell state by the input of the current time step.

[0049] Output gate: determines the output of the hidden state of the current time step. It inputs the input of the current time step and the hidden state of the previous time step, passes through a sigmoid activation function, and outputs a value between 0 and 1, indicating the output degree of the hidden state at the corresponding position. At the same time, the cell state passes through a tanh activation function to obtain a value between -1 and 1, indicating the candidate value of the hidden state of the current time step. Finally, the output gate and the candidate hidden state are multiplied to obtain the final hidden state.

[0050] LSTM is used to process sequence data. It controls the flow of information and the update of memory through forget gate, input gate and output gate. Multi-layer LSTM network can increase the depth of the network and improve the modeling ability.

[0051] (2) Model building

[0052] Define the architecture of the RNN model, including the input layer, hidden layer (multi-layer RNN unit), and output layer. Determine the number of hidden layers, the number of units in each hidden layer, and hyperparameters such as activation functions. If necessary, you can add additional layers, such as fully connected layers (DenseLayer), for classification or regression.

[0053] An RNN neural network mainly consists of the following 5 layers:

[0054] Input units: {x0,...x t-1 ,x t ,x t+1 ……}

[0055] Hidden units: {s0,...s t-1 ,s t ,s t+1 ……}

[0056] Output units: {o0,...o t-1 ,o t ,o t+1 ……}

[0057] Input layer: xt represents the input at time t;

[0058] The state of the hidden layer in the middle of the tth time is:

[0059] s t =f(U*x t +W*s t-1 ), f is a nonlinear activation function.

[0060] Output layer: o t =softmax(Vs t ), the softmax function is a normalized exponential function that makes each element range between 0 and 1 and the sum of all elements is 1.

[0061] Fully connected layer: y = σ(W x + b)

[0062] Here, W is the weight matrix, x is the input vector, b is the bias term, and σ is the activation function (such as softmax for classification).

[0063] RNN network architecture: RNN stores information from previous time steps through hidden states and has a cyclic structure. A typical RNN network architecture contains an input, an output, and a neural network unit.

[0064] The difference from ordinary feedforward neural networks is that the neural network units of RNNs are not only connected to the input and output, but also have a cycle / loop / circuit / loop themselves. This loop allows information to be passed from one step in the network to the next step.

[0065] Data update module: Select the Adam optimization algorithm to update the model parameters.

[0066] Specific process:

[0067] Initialization: Set data x i The initial learning rate (α), the exponential decay rate of the first moment estimate (β1), the exponential decay rate of the second moment estimate (β2), and a very small number (ε) to prevent the denominator from being zero in the implementation.

[0068] Calculate gradients: At each iteration, the gradient of the loss function with respect to each parameter is calculated using the mini-batch gradient descent method.

[0069] Compute estimates of the first and second moments: Estimates of the first (mean) and second (variance) moments of the gradient are computed using the exponentially decaying averaging method.

[0070] First-order moment estimation (Momentum): m t =β1·m t-1 +(1-β1)·g t , where g t is the current gradient, m t is the current first moment estimate.

[0071] Second moment estimation (RMSProp): where ν t is the current second moment estimate.

[0072] Correcting bias: Since the momentum value is zero initially, the estimation of the first and second moments is adjusted through bias correction to make them more accurate.

[0073] Corrected first moment:

[0074] Corrected second moment:

[0075] Update parameters: Update the parameters based on the corrected first-order moment and second-order moment estimates and the learning rate.

[0076] Parameter update formula:

[0077] The model is trained using the training set data, the loss is calculated through forward propagation, and then the model parameters are updated through backpropagation and gradient descent.

[0078] Evaluate model performance on the validation set to avoid overfitting and adjust model structure and hyperparameters as needed.

[0079] Display module: Input the preprocessed real-time monitoring data into the trained RNN model. The model outputs the prediction result of the fault location, including a specific coordinate, equipment number or line segment identification. Analyze the output of the model to determine the specific location of the fault.

[0080] The data acquisition module, data preprocessing module, data analysis module, data updating module and display module of the present application are all controlled by a microprocessor.

[0081] The cloud platform collects feedback data based on the actual results of fault repair. The difference between the predicted results and the actual results is analyzed to evaluate the model performance. The cloud platform transmits the data to the microprocessor to retrain or fine-tune the model using the new fault data and feedback data to improve the accuracy and robustness of the model.

[0082] Based on the prediction results of the deep learning model, the system can quickly locate the fault location and provide specific coordinate information. This fast and accurate positioning capability greatly reduces the time and labor costs of on-site investigations and improves the efficiency of emergency repairs.

[0083] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. An intelligent distribution network emergency repair dispatching system, characterized in that: The system includes: a data acquisition module, a data preprocessing module, a data analysis module, a data update module and a display module. The data acquisition module monitors the multi-parameter data of each node of the distribution network in real time through a multi-parameter sensor. The multi-parameter data collected by the data acquisition module is transmitted to the data preprocessing module in real time through a wireless or wired manner. The data preprocessing module preprocesses the multi-parameter data, and the preprocessing includes cleaning and standardization. The multi-parameter data after the preprocessing is trained by a data analysis module for a model, and the model includes LSTM and RNN. At the same time, the data update module uses the Adam optimization algorithm to update the parameters of the RNN. The data analysis module analyzes the multi-parameter data in real time, and the model outputs the prediction result of the fault location, thereby determining the specific location of the fault, and displaying it through the display module, thereby completing the emergency repair process.

2. The intelligent distribution network emergency repair dispatching system according to claim 1, characterized in that: The data acquisition module uses a standard API interface to connect to the multi-parameter sensor monitoring network, and uses RFID tags to identify and track power equipment at each node in the monitoring distribution network to ensure rapid positioning during emergency repairs.

3. The intelligent distribution network emergency repair dispatching system according to claim 1, characterized in that: The multi-parameter data of each node in the distribution network include voltage, current, temperature, humidity, power factor, harmonic content, electromagnetic intensity, and frequency.

4. The intelligent distribution network emergency repair dispatching system according to claim 3 is characterized in that: The cleaning is to identify and process outliers using the standard deviation method.

5. The intelligent distribution network emergency repair dispatching system according to claim 4, characterized in that: The standardization is to standardize the multi-parameter data by improving the Z-score standardization, so that the data can be compared and analyzed on the same scale.

6. The intelligent distribution network emergency repair dispatching system according to claim 5, characterized in that: The standardized multi-parameter data are normalized to convert the multi-parameter data into dimensionless values ​​between 0 and 1, and the minimum-maximum normalization formula is obtained.

7. The intelligent distribution network emergency repair dispatching system according to claim 1, characterized in that: The LSTM model in the data analysis module contains a cell state and a hidden state. The cell state is responsible for storing and transmitting information, and the hidden state is responsible for extracting and outputting information. The RNN includes an input layer, a hidden layer, and an output layer.

8. The intelligent distribution network emergency repair dispatching system according to claim 7, characterized in that: The data analysis module performs model training, and the model includes LSTM and RNN. The training uses the training set data to train the model, calculates the loss through forward propagation, and then updates the RNN model parameters through back propagation and gradient descent, and evaluates the model performance on the validation set to avoid overfitting, and adjusts the model structure and hyperparameters as needed.

9. The intelligent distribution network emergency repair dispatching system according to claim 1, characterized in that: The specific process of the Adam optimization algorithm is as follows: (1) Initialization: Set data x i The initial learning rate α, the exponential decay rate β1 of the first-order moment estimate, the exponential decay rate β2 of the second-order moment estimate, and the fraction ε, which is used to prevent the denominator from being zero in the implementation; (2) Calculate gradients: In each iteration, use the mini-batch gradient descent method to calculate the gradient of the loss function with respect to each parameter; Calculate the estimates of the first-order moment and the second-order moment: Calculate the estimate of the mean of the first-order moment and the variance of the second-order moment of the gradient by the exponential decay averaging method; (3) Correction bias: Since the momentum value is zero at the beginning, the estimation of the first and second moments is adjusted by bias correction; (4) Update parameters: Update the parameters based on the corrected first-order moment and second-order moment estimates and the learning rate.

10. The intelligent distribution network emergency repair dispatching system according to claim 8, characterized in that: The specific location of the fault is output on the display module, and the display module includes a cloud platform, which can collect feedback data according to the actual results of the fault repair, analyze the difference between the predicted results and the actual results, and evaluate the model performance.

Images