Power distribution system fault prediction method based on fractal theory prediction and BP neural network classification
By combining fractal theory and BP neural networks, and using current data for power distribution system fault prediction, the problem of unreliable prediction in existing technologies has been solved, and the fault early warning and handling capabilities of smart grids have been realized.
Patent Information
- Application Number
- CN202610817771.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-08
- Publication Date
- 2026-07-24
AI Technical Summary
Existing technologies are insufficient for reliably predicting power distribution system faults, and accurate prediction results cannot be obtained by relying solely on current analysis.
Combining fractal theory and BP neural networks, the system determines whether the current is normal and predicts impending faults by comparing the predicted current with the actual current. It uses fractal prediction steps to establish the current history sequence and measurement scale, performs linear regression, and combines BP neural networks for diagnosis.
It enables reliable prediction of power distribution system faults, provides early warning and fault resolution strategies, and improves the fault handling capabilities of smart grids.
Smart Images
Figure CN122451753A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent technology for power grid fault diagnosis, specifically to a method for predicting power distribution system faults based on fractal theory prediction and BP neural network classification. Background Technology
[0002] Compared to traditional power grids, smart grids require the ability to acquire comprehensive information about the entire power grid, enabling real-time dynamic analysis, diagnosis, and optimization of power grid business flows to achieve precise, accurate, and timely power grid operation and management. Simultaneously, smart grids require bidirectional monitoring and intelligent control of information and power flow, a reliable, stable, and efficient mesh topology, and automatic fault isolation and self-healing capabilities. Predicting distribution system faults allows smart grid systems to effectively prevent fault occurrence and better respond to various faults, taking swift action. Therefore, predicting distribution system faults is of great significance to the operation of smart grids and is one of the urgent problems that smart grids need to solve.
[0003] Fractal geometry, a theory and methodology developed in the last decade or so, studies nonlinear phenomena. Founded by the French-American mathematician Mandelbrot, it studies disordered, irregular, and self-similar systems that are widespread in nature and social life. Self-similarity refers to the statistically significant similarity between the parts and the whole in terms of form, function, information, time, and space. Therefore, fractal theory can be applied to predicting current in power distribution system lines.
[0004] A backpropagation (BP) neural network is a multilayer feedforward neural network. Its main characteristic is that the signal propagates forward while the error propagates backward. In the forward propagation, the input signal is processed layer by layer from the input layer through the hidden layers until it reaches the output layer. The state of neurons in each layer only affects the state of neurons in the next layer. If the output layer does not produce the desired output, backpropagation begins. The network weights and thresholds are adjusted based on the prediction error, allowing the BP neural network to continuously approximate the desired output.
[0005] In summary, fractal theory can be used to predict the current in the power distribution system lines in the future. By comparing the predicted current with the actual measured current, it can be determined whether the current value is within the normal range. A backpropagation (BP) neural network can effectively perform this comparison, thereby determining whether the power distribution system lines are functioning correctly. Summary of the Invention
[0006] The technical problem this invention aims to solve is that predicting power distribution system faults is one of the most pressing issues in smart grids, and relying solely on current analysis is insufficient to obtain reliable prediction results. This invention provides a power distribution system fault prediction method based on fractal theory and BP neural network classification. By combining the excellent predictive capabilities of fractal theory with the data classification capabilities of BP neural networks, it compares the predicted current with the actual measured current to determine whether the current is normal and predicts impending faults, providing early warnings and solutions for further processing.
[0007] The technical solution adopted by this invention to solve its technical problem is: a method for predicting faults in a power distribution system based on fractal theory prediction and BP neural network classification. The method is characterized by collecting current data of the line to be predicted in the power distribution system and performing the following fractal prediction steps: S1. Establishing a historical current sequence to obtain a cumulative sum sequence and a measurement scale; S2. Performing linear regression on the cumulative sum sequence curve in a double logarithmic coordinate system to obtain the fractal dimension; S3. Predicting the current value at future times. After prediction, the actual current value is measured, and the difference between the actual value and the predicted value is calculated as input for BP neural network diagnosis to obtain the fault prediction result.
[0008] The fractal prediction step S1 includes: dividing the current data into time intervals. Sampling is used to obtain a discretized current history sequence. ,in To determine the number of discretized current data, a first-order cumulative sum sequence of the historical sequence is constructed. ,in It is a measurement scale.
[0009] The fractal prediction step S2 includes: in a double logarithmic coordinate system, with... The horizontal axis is... Using the vertical axis as the ordinate, historical data points are plotted, and a linear regression is performed on the curve using the least squares method to obtain the linear regression equation. The slope of the regression line is the fractal dimension. ,and axial intercept is , It is a constant.
[0010] The fractal prediction step S3 includes: based on the obtained... and Value, pair Find By deducing the serial number The predicted current value is And so on, to find ,in Given the number of data points, we obtain the predicted current values for future times.
[0011] After the fractal prediction is completed, the actual current value is measured. Calculate the difference between the actual value and the predicted value. , which serves as the input to the BP neural network.
[0012] The BP neural network diagnosis consists of three parts: BP neural network construction, BP neural network training, and BP neural network classification.
[0013] The BP neural network construction determines the structure of the BP neural network based on the characteristics of the input and output data, where the input is... There are a total of The data is used to output the fault type of the power distribution line. Therefore, the structure of the BP neural network is as follows: ,in The number of fault types, i.e., the number of fault types in the input layer. There are nodes, and the hidden layer has... There are nodes, and the output layer has [number] nodes. Each node.
[0014] The types of power distribution line faults include, but are not limited to, high-resistance faults, single-phase grounding faults, intermittent faults, etc.
[0015] The BP neural network is trained using training data and test data, both of which contain multiple sets of training and test data. And fault type.
[0016] The BP neural network classification uses a trained neural network to classify the data to be predicted and outputs the normal or fault type of the power distribution line.
[0017] The beneficial effect of this invention is that it can predict the current in the power distribution system lines in the future using fractal theory. By comparing the predicted current with the actual measured current, a BP neural network can determine whether the current value is normal. If it is abnormal, it indicates the corresponding fault. Combining fractal theory and BP neural network can yield reliable prediction results. The obtained prediction results provide early warning methods and fault resolution strategies for line monitoring in smart grids. Attached Figure Description
[0018] Figure 1 This is a flowchart of the fractal prediction process in the power distribution system fault prediction method of the present invention.
[0019] Figure 2 This is a diagram of the BP neural network topology in the power distribution system fault prediction method of the present invention.
[0020] Figure 3 This is the flowchart of the BP neural network classification algorithm in the power distribution system fault prediction method of the present invention.
[0021] Figure 4 The current data (24 hours) of the power distribution system to be predicted in this embodiment of the invention.
[0022] Figure 5 This is the current data (1 hour) of the line to be predicted in the power distribution system according to an embodiment of the present invention.
[0023] Figure 6 These are the current prediction results and measured values of the circuit to be predicted in the power distribution system according to an embodiment of the present invention. Detailed Implementation
[0024] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.
[0025] like Figure 1 As shown, the power distribution system fault prediction method based on fractal theory prediction and BP neural network classification collects current data of the line to be predicted in the power distribution system and performs the following fractal prediction steps: S1, establish the current history sequence to obtain the cumulative sum sequence and measurement scale; S2, perform linear regression on the cumulative sum sequence curve in a double logarithmic coordinate system to obtain the fractal dimension; S3, predict the current value in the future time.
[0026] like Figure 2 As shown, after the fractal prediction is completed, the difference between the actual value and the predicted value is calculated. , which serves as the input to the BP neural network. Figure 2 The structure of the BP neural network is shown, where there are a total of The data is used to output the fault type of the power distribution line. Therefore, the structure of the BP neural network is as follows: ,in The number of fault types, i.e., the number of fault types in the input layer. There are nodes, and the hidden layer has... There are nodes, and the output layer has [number] nodes. Each node.
[0027] like Figure 3 As shown, BP neural network classification consists of three parts: BP neural network construction, BP neural network training, and BP neural network classification. BP neural network construction determines the structure of the BP neural network based on the characteristics of the input and output data. BP neural network training uses training data and test data to train the BP neural network, and both training data and test data contain multiple sets of data. And fault type; BP neural network classification uses a trained neural network to classify the data to be predicted and output the normal or fault type of the power distribution line.
[0028] Example
[0029] This embodiment uses fractal theory prediction and BP neural network classification to predict faults in the main circuit of a library's power distribution system.
[0030] First, collect current data from the main circuits of the library's power distribution system, such as... Figure 4 and Figure 5 As shown. Among them, Figure 4 The display shows current data for a total of 24 hours in a day. Figure 5 The data displays current data for a specific hour. Based on the collected current data, different time periods are selected for fractal prediction and BP neural network training. In this example, a 60-second time period is used, meaning 6-second segments of current data are continuously extracted from the 24-hour current data to obtain sufficient training and testing data. A total of 14,400 data sets are obtained through continuous extraction. It's worth noting that repeating the extraction process, where each set contains some identical data, would yield even more data. However, in this example, 14,400 data sets were obtained through repeated extraction.
[0031] Secondly, for each set of data obtained, fractal prediction is performed as follows.
[0032] Execution step S1: Take the current data for the first 4 seconds of each data set, and then divide the data into time intervals. Sampling is used to obtain a discretized current history sequence. Then construct the first-order cumulative sum sequence of the historical sequence. ,in It is a measurement scale.
[0033] Execute step S2: In a double logarithmic coordinate system, with... The horizontal axis is... Using the vertical axis as the ordinate, historical data points are plotted, and a linear regression is performed on the curve using the least squares method to obtain the linear regression equation. In this example, the slope of the obtained regression line equation is... The intercept with the axis is .
[0034] Execute step S3: Based on the obtained and The value can be calculated. By deducing the predicted current value for serial number 401, And so on, to find .
[0035] This completes the fractal prediction and yields the current value prediction result. .
[0036] Next, the BP neural network is trained. Since the current data is processed at time intervals... Sampling resulted in a total of 600 sequences for a single set of current data (6 seconds in total). Of these, 400 sequences were used for the preceding fractal prediction, and the remaining 200 sequences represent the actual current values. Calculate the difference between the actual value and the predicted value. This serves as the input to the BP neural network. In this example, the structure of the BP neural network is as follows: The output includes two categories: "normal" and "abnormal". The 10,000 sets of collected data are divided into training data and test data, and then the neural network is trained.
[0037] Finally, a trained BP neural network is used to predict whether the power distribution system will experience a fault. For example... Figure 6 The image shows the predicted current and measured values for the circuit to be predicted in the power distribution system. Figure 6 The dataset contains current data over a total time span of 6 seconds. The first 4 seconds are historical current data, used for fractal theory prediction, resulting in predicted current data for the remaining 2 seconds. The last 2 seconds are measured current data. The difference between the predicted and measured current data is taken and input into a trained BP neural network to obtain the prediction result. In this example, the prediction result is "normal".
[0038] Based on the preferred embodiments of the present invention described above, those skilled in the art can make various changes and modifications without departing from the technical concept of the invention. The technical scope of this invention is not limited to the contents of the specification, but must be determined by the scope of the claims.
Claims
1. A method for predicting faults in a power distribution system based on fractal theory prediction and BP neural network classification, characterized in that, By collecting current data from the circuit to be predicted in the power distribution system, fractal prediction is performed using the following steps: S1, establish a historical current sequence to obtain the cumulative sum sequence and measurement scale; S2, perform linear regression on the cumulative sum sequence curve in a double logarithmic coordinate system to obtain the fractal dimension; S3, predict the current value for future times. After prediction, the actual current value is measured, and the difference between the actual and predicted values is calculated as input for BP neural network diagnosis to obtain the fault prediction result.
2. The power distribution system fault prediction method as described in claim 1, characterized in that, The fractal prediction step S1 includes: dividing the current data into time intervals. Sampling is used to obtain a discretized current history sequence. ,in To determine the number of discretized current data, a first-order cumulative sum sequence of the historical sequence is constructed. ,in It is a measurement scale.
3. The power distribution system fault prediction method as described in claim 1, characterized in that, The fractal prediction step S2 includes: in a double logarithmic coordinate system, with... The horizontal axis is... Using the vertical axis as the ordinate, historical data points are plotted, and a linear regression is performed on the curve using the least squares method to obtain the linear regression equation. The slope of the regression line is the fractal dimension. ,and axial intercept is , It is a constant.
4. The power distribution system fault prediction method as described in claim 1, characterized in that, The fractal prediction step S3 includes: based on the obtained and Value, pair Find By deducing the serial number The predicted current value is And so on, to find ,in Given the number of data points, we obtain the predicted current values for future times.
5. The power distribution system fault prediction method as described in claim 1, characterized in that, After the fractal prediction is completed, the actual current value is measured. Calculate the difference between the actual value and the predicted value. , which serves as the input to the BP neural network.
6. The power distribution system fault prediction method as described in claim 1, characterized in that, The BP neural network diagnosis consists of three parts: BP neural network construction, BP neural network training, and BP neural network classification.
7. The power distribution system fault prediction method as described in claims 1 and 6, characterized in that, The BP neural network construction determines the structure of the BP neural network based on the characteristics of the input and output data, where the input is... There are a total of The data is used to output the fault type of the power distribution line. Therefore, the structure of the BP neural network is as follows: ,in The number of fault types, i.e., the number of fault types in the input layer. There are nodes, and the hidden layer has... There are nodes, and the output layer has [number] nodes. Each node.
8. The power distribution system fault prediction method as described in claims 1, 6, and 7, characterized in that, The types of power distribution line faults include, but are not limited to, high-resistance faults, single-phase grounding faults, intermittent faults, etc.
9. The power distribution system fault prediction method as described in claims 1 and 6, characterized in that, The BP neural network is trained using training data and test data, both of which contain multiple sets of training and test data. And fault type.
10. The power distribution system fault prediction method as described in claims 1 and 6, characterized in that, The BP neural network classification uses a trained neural network to classify the data to be predicted and outputs the normal or fault type of the power distribution line.