A method and system for monitoring power cable faults based on big data

Abstract

The present invention relates to the field of cables, and more specifically, to a method and system for monitoring power cable faults based on big data. The method comprises: using a time series decomposition algorithm to decompose the voltage value obtained at each sampling moment into a trend term, seasonality, and residual value; comparing the residual consistency with a preset threshold to obtain a first prediction value; calculating a second prediction value through a time series model; and monitoring power cable faults through the prediction value. Through the technical solution of the present invention, a more certain voltage value prediction value can be obtained, the real-time and accuracy of power cable fault monitoring can be improved, and the probability of accidents caused by faults can be reduced.

Description

Technical Field

[0001] The present invention relates to the field of cables, and more specifically, to a method and system for monitoring power cable faults based on big data. Background Art

[0002] As an important part of the power transmission system, power cables are responsible for transmitting electricity from power stations to thousands of households and various industrial and commercial facilities. With the development of social economy and the acceleration of urbanization, the scale and complexity of power cable networks are increasing, and their safe and stable operation is of vital importance to ensuring social production and people's lives. Power cable faults, such as broken wires, short circuits, and grounding faults, will lead to power supply interruptions and affect normal production and living order. In addition, faults may also cause safety accidents such as equipment damage and fire, causing huge economic losses and social impacts. Therefore, timely detection and accurate positioning of power cable faults are of great significance to ensuring the stable operation of the power system and improving power supply reliability.

[0003] The existing Chinese patent application document with publication number CN114563661A discloses a method and device for monitoring the status of power cables. The method obtains power cable status monitoring information, configures and sorts the health status levels of the power cables, and displays the monitoring information of high-risk power cables with low health status on the front side, which is convenient for staff to view and monitor, thereby realizing power cable fault detection.

[0004] However, the above method only involves obtaining power cable status monitoring information, configuring and sorting the health status level of the power cables, and does not process the collected voltage data, which results in low credibility of the collected voltage data and reduced accuracy of power cable fault monitoring. Summary of the invention

[0005] In order to solve the problem of low credibility of collected voltage data, the present invention proposes a power cable fault monitoring method and system based on big data.

[0006] In a first aspect, the present invention discloses a power cable fault monitoring method based on big data, comprising: decomposing a voltage value sequence obtained in a sampling period into a trend item sequence, a seasonal sequence and a residual value sequence by using a time series decomposition algorithm; constructing the residual values ​​of multiple historical sampling periods at the current sampling moment as a reference sequence, and taking the normalized variance of the reference sequence as the residual consistency at the current sampling moment; in response to the residual consistency being no greater than a preset threshold, taking the voltage value of the previous sampling period at the current sampling moment as the first predicted value of the current sampling period at the current sampling moment; in response to the residual consistency being greater than the preset threshold, inputting the reference sequence into a preset neural network, setting the number of floating training times, outputting the residual predicted value at the current sampling moment, and adding the trend item and seasonality of the previous sampling period at the current sampling moment to the residual predicted value as the first predicted value of the current sampling period at the current sampling moment, wherein the number of floating training times satisfies the relationship: , Indicates the number of floating training times, represents the correction factor, represents residual consistency, Indicates the initial training times of the preset neural network; and completes the monitoring of the power cable fault according to the first prediction value.

[0007] The constructed reference sequence and the calculation of residual consistency provide a quantitative method for evaluating the stability of voltage values. If the residual consistency is not greater than the preset threshold, it means that the voltage data at the current sampling moment is similar to the historical data, and the data can be considered stable. By adjusting the number of neural network training times through residual consistency, the predicted value that conforms to the residual value changes of multiple sampling cycles can be obtained, and a more accurate voltage value prediction value can be calculated.

[0008] Preferably, the multiple historical sampling periods are multiple sampling periods before the current sampling period and continuous with the current sampling period.

[0009] Multiple consecutive sampling cycles can ensure the real-time nature of the residual data and avoid the reduction in the accuracy of the predicted residual value due to the large time span between the historical sampling cycle and the current sampling cycle.

[0010] Preferably, the preset neural network is a BP neural network, which includes an input layer, a hidden layer and an output layer. The input layer is used to receive the input of a reference sequence. After the hidden layer extracts features from the input information, the extracted features are input to the output layer to output the residual prediction value at the current sampling moment. The loss function of the preset neural network uses cross entropy loss, and the training method uses gradient descent.

[0011] Preferably, the monitoring of power cable faults based on the first prediction value includes: comparing the first prediction value with a preset voltage threshold, and in response to the first prediction value being greater than the preset voltage threshold, recording it as suspected abnormal data; when the suspected abnormal data appears a preset number of times consecutively, generating and sending an alarm signal.

[0012] By setting the suspected abnormal data, it is possible to prevent false alarms caused by the influence of harmonics in the power on the voltage value.

[0013] Preferably, one day is set as a sampling period.

[0014] A one-day data cycle can provide more real-time analysis and feedback, can provide a more detailed observation window, and can more easily adapt to changes in voltage values ​​using a one-day data cycle.

[0015] Preferably, it also includes: constructing the voltage values ​​at historical sampling moments before the current sampling moment in this sampling cycle into a historical voltage sequence, inputting the historical voltage sequence into a time series model, and outputting the second predicted value of this sampling cycle at the current sampling moment; taking the average of the first predicted value and the second predicted value as the final predicted value of this sampling cycle at the current sampling moment; and completing the monitoring of power cable faults according to the final predicted value.

[0016] The data in this sampling period provides the latest information and can be used to build real-time or near real-time prediction models. It can also reflect immediate changes in the system or environment more quickly and help the model capture short-term trends.

[0017] Preferably, the monitoring of power cable faults based on the final predicted value includes: comparing the final predicted value with a preset voltage threshold, and in response to the final predicted value being greater than the preset voltage threshold, recording it as suspected abnormal data; when the suspected abnormal data appears for a preset number of consecutive times, generating and sending an alarm signal.

[0018] In a second aspect, the present invention discloses a power cable fault monitoring system based on big data, comprising: a processor; and a memory, wherein the memory stores computer instructions, and when the computer instructions are executed by the processor, the device executes the above-mentioned power cable fault monitoring method based on big data.

[0019] Beneficial effects of the present invention:

[0020] The first prediction value obtained at the current moment from multiple historical sampling cycles can capture long-term trends and seasonal patterns; the second prediction value obtained from multiple historical sampling moments in this sampling cycle can capture short-term trends and tendencies of voltage changes, and obtain more comprehensive prediction results from different dimensions, thereby increasing the credibility of the predicted value of the voltage value at the current sampling moment in this sampling cycle, and thus improving the accuracy of power cable fault monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0022] Figure 1 It is a flow chart of a power cable fault monitoring method based on big data according to an embodiment of the present invention.

[0023] Figure 2 It is a schematic diagram of a power cable fault monitoring system based on big data according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0025] It should be understood that when the terms "first", "second", etc. are used in the claims, descriptions, and drawings of the present invention, they are only used to distinguish different objects, rather than to describe a specific order. The terms "include" and "comprise" used in the description and claims of the present invention indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their collections.

[0026] The present invention provides a method for monitoring power cable faults based on big data. Figure 1 As shown, a power cable fault monitoring method based on big data includes steps S1 to S4, which are described in detail below.

[0027] S1, using the time series decomposition algorithm, the voltage value sequence obtained in a sampling period is decomposed into a trend term sequence, a seasonal sequence and a residual value sequence.

[0028] In one embodiment, the voltage value of the power cable is collected according to a preset sampling time. For example, the present invention is configured to collect the voltage value once per minute. The specific time can be set by those skilled in the art.

[0029] The time series decomposition algorithm is a method for analyzing and predicting time series data, and exemplary methods include STL decomposition method (Seasonal and Trend decomposition using Loess), ARIMA model (Autoregressive Integrated Moving Average model), state space model and Kalman filter, etc. The present invention uses STL decomposition method to decompose the voltage value obtained at each sampling moment into trend term, seasonality and residual value.

[0030] It should be noted that the trend term represents the overall change trend of voltage data in the long term; seasonality represents the periodic change of voltage values ​​that recurs in a specific time period; the residual value refers to the random fluctuation or noise remaining in addition to the trend term and seasonality. The residual value includes the part that is not explained by the trend term and seasonality, which is usually considered to be the random change or unknown factor in the data. Since the voltage value in continuous time may fluctuate, the voltage value is decomposed into trend term, seasonality and residual value. The trend term and seasonality in continuous time are generally stable, and the voltage value is usually affected by the fluctuation or noise, that is, the residual value.

[0031] S2, comparing the residual consistency with a preset threshold to obtain a first prediction value.

[0032] In one embodiment, one week is set as a sampling period, which can include a large number of voltage values ​​and improve the accuracy of prediction.

[0033] In another embodiment, setting one day as a sampling period can reduce the amount of data during prediction, speed up the prediction, reduce the time span, shorten the time difference between different sampling days at the current sampling time, and improve the credibility of the voltage value of the historical sampling day at the current sampling time.

[0034] The residual values ​​of multiple historical sampling periods at the current sampling moment are constructed as a reference sequence, and the normalized variance of the reference sequence is used as the residual consistency at the current sampling moment.

[0035] Exemplarily, March 20 is taken as the sampling period, 12 noon is taken as the current sampling time, and the residual value at 12 noon in each historical sampling period from March 10 to March 19 is taken as the reference sequence to predict the residual value at 12 noon on March 20, where the specific time can be set by technicians in this field.

[0036] The residual consistency is compared with a preset threshold, and in response to the residual consistency being not greater than the preset threshold, the voltage value of the previous sampling cycle at the current sampling moment is used as the first predicted value of the current sampling cycle at the current sampling moment.

[0037] Exemplarily, March 20 is used as the sampling period, the preset threshold is set to 0.6, and when the residual consistency is not greater than the preset threshold, the voltage value at 12:00 noon on March 19 is used as the first prediction value at 12:00 noon on March 20.

[0038] In response to the residual consistency being greater than a preset threshold, the reference sequence is input into a preset neural network. Exemplarily, the preset neural network can be an FCN (Fully Convolutional Network) network, a U-net network, a GCN (Graph Convolutional Network) network, a GAN (Generative Adversarial Network) network, etc. The present invention adopts a BP neural network.

[0039] The BP neural network includes an input layer, a hidden layer, and an output layer. The input layer is used to receive the input of the reference sequence. After the hidden layer extracts the features of the input information, the extracted features are input to the output layer to output the residual prediction value at the current sampling moment. The loss function of the preset neural network uses the cross entropy loss, and the training method uses the gradient descent method.

[0040] The residual values ​​of any consecutive historical sampling periods at the same sampling time and the residual values ​​of a historical sampling period after any consecutive historical sampling periods are used as label samples, and the label samples are input into the BP neural network for training. During the training process, the cross entropy loss function between the predicted value and the true value output by the label sample is calculated, and the BP neural network parameters are adjusted using the gradient descent method to minimize the error between the predicted value and the true value. When the number of BP neural network training times reaches the set floating training times, the trained BP neural network is obtained.

[0041] Input the reference sequence into the preset neural network, output the residual prediction value at the current sampling time, and add the trend item and seasonality of the previous sampling period at the current sampling time to the residual prediction value as the first prediction value of this sampling period at the current sampling time. Among them, the floating training times satisfy the relationship: , Indicates the number of floating training times, represents the correction factor, represents residual consistency, Indicates the initial training times of the preset neural network. The correction factor is used to correct the residual consistency to prevent the data difference between the residual consistency and the initial training times from affecting the accuracy of the calculated floating training times.

[0042] At this point, the first prediction value can be obtained.

[0043] S3, calculating a second prediction value through a time series model.

[0044] In one embodiment, the voltage values ​​at the historical sampling moments before the current sampling moment in the current sampling cycle are constructed as a historical voltage sequence, the historical voltage sequence is input into the time series model, and the second predicted value of the current sampling cycle at the current sampling moment is output.

[0045] For example, March 20 is taken as the sampling period and 12 noon is taken as the current sampling time. Then, the voltage values ​​at 8:00, 9:00, 10:00 and 11:00 on March 20 need to be taken as the historical voltage sequence.

[0046] It should be noted that the time series model is a statistical model used to analyze time series data. It can predict future development trends and provide decision support by understanding the past behavior of the data. Time series models include autoregressive models (AR), moving average models (MA), seasonal autoregressive integrated moving average models (SARIMA), etc. The present invention adopts an autoregressive model.

[0047] Specifically, the order of the autoregressive model is determined to be 4, and the specific order can be adjusted and set according to experiments or experience; the autoregressive model is trained, the training set is input into the autoregressive model for training, and the least squares method is used to estimate the parameters of the autoregressive model; the test set is used to evaluate the performance of the model, the mean square error is calculated to evaluate the performance of the model, and the parameters in the model are adjusted according to the evaluation results; the voltage value at the historical moment is input into the trained autoregressive model, and the second predicted value of this sampling period at the current sampling moment is obtained.

[0048] At this point, the second prediction value is obtained.

[0049] S4, monitoring power cable faults through prediction values.

[0050] In one embodiment, the monitoring of power cable faults can be completed based on the first prediction value, and the first prediction value and the preset voltage threshold are compared. In response to the first prediction value being greater than the preset voltage threshold, it is recorded as suspected abnormal data; when the suspected abnormal data appears a preset number of times in succession, an alarm signal is generated and sent, which can avoid the influence of harmonics in the power cable on the voltage value, causing the voltage value at the current sampling moment to increase or decrease, and then generating a false alarm.

[0051] In another embodiment, the average of the first prediction value and the second prediction value is used as the final prediction value of this sampling cycle at the current sampling moment; the final prediction value is compared with a preset voltage threshold, and in response to the final prediction value being greater than the preset voltage threshold, it is recorded as suspected abnormal data; when the suspected abnormal data appears for a preset number of times in succession, an alarm signal is generated and sent.

[0052] The embodiment of the present invention also discloses a power cable fault monitoring system based on big data, referring to Figure 2 , including a processor and a memory, the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a power cable fault monitoring method based on big data according to the present invention is implemented.

[0053] The above system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, and their configuration and functions are known in the art, so they will not be described in detail here.

[0054] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, apparatus or device. For example, a computer-readable storage medium may be any appropriate magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM (Resistive Random Access Memory), a dynamic random access memory DRAM (Dynamic Random Access Memory), a static random access memory SRAM (Static Random-Access Memory), an enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), a high-bandwidth memory HBM (High-Bandwidth Memory), a hybrid memory cube HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module or both. Any such computer storage medium may be part of a device or accessible or connectable to a device.

[0055] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

[0056] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A power cable fault monitoring method based on big data, characterized in that: include: The voltage value sequence obtained in one sampling period is decomposed into trend term sequence, seasonal sequence and residual value sequence by using time series decomposition algorithm; The residual values ​​of multiple historical sampling periods at the current sampling moment are constructed as a reference sequence, and the normalized variance of the reference sequence is used as the residual consistency at the current sampling moment; In response to the residual consistency being not greater than a preset threshold, taking the voltage value of the previous sampling cycle at the current sampling time as the first predicted value of the current sampling cycle at the current sampling time; In response to the residual consistency being greater than a preset threshold, the reference sequence is input into a preset neural network, the number of floating training times is set, the residual prediction value at the current sampling moment is output, and the trend item and seasonality of the previous sampling period at the current sampling moment are added to the residual prediction value as the first prediction value of this sampling period at the current sampling moment, wherein the number of floating training times satisfies the relationship: , Indicates the number of floating training times, represents the correction factor, represents residual consistency, Indicates the initial training times of the preset neural network; The voltage values ​​at the historical sampling moments before the current sampling moment in this sampling cycle are constructed as a historical voltage sequence, the historical voltage sequence is input into the time series model, and the second predicted value of this sampling cycle at the current sampling moment is output; the average of the first predicted value and the second predicted value is taken as the final predicted value of this sampling cycle at the current sampling moment; Complete monitoring of the power cable fault according to the first prediction value or the final prediction value; The monitoring of the power cable fault according to the first prediction value includes: comparing the first predicted value with a preset voltage threshold, and in response to the first predicted value being greater than the preset voltage threshold, recording it as suspected abnormal data; When suspected abnormal data appears for a preset number of consecutive times, an alarm signal is generated and sent; The monitoring of power cable faults based on the final prediction value includes: Comparing the final prediction value with the preset voltage threshold, and in response to the final prediction value being greater than the preset voltage threshold, recording it as suspected abnormal data; When suspected abnormal data appears for a preset number of consecutive times, an alarm signal is generated and sent.

2. A power cable fault monitoring method based on big data according to claim 1, characterized in that: The multiple historical sampling periods are multiple sampling periods before the current sampling period and continuous with the current sampling period.

3. The method for monitoring power cable faults based on big data according to claim 1, characterized in that: The preset neural network is a BP neural network, which includes an input layer, a hidden layer and an output layer. The input layer is used to receive the input of the reference sequence. After the hidden layer extracts the features of the input information, the extracted features are input to the output layer to output the residual prediction value at the current sampling moment. The loss function of the preset neural network uses cross entropy loss, and the training method uses gradient descent.

4. The method for monitoring power cable faults based on big data according to claim 1, characterized in that: Set one day as a sampling period.

5. A power cable fault monitoring system based on big data, characterized in that: include: Processor; and A memory storing computer instructions, wherein when the computer instructions are executed by a processor, the device executes a power cable fault monitoring method based on big data according to any one of claims 1 to 4.

Images