Method and device for predicting service life of distribution transformer and electronic equipment

Abstract

The invention provides a service life prediction method and device of a distribution transformer and electronic equipment. The method comprises the steps that related data of the distribution transformer are acquired, and the related data comprise load data, operation data, model parameter data and environment data of the distribution transformer; characteristic data of the distribution transformer are determined according to the related data, and the characteristic data comprise load characteristics in the load data extracted by adopting a short-time Fourier transform technology, related characteristics of load fluctuation and environmental factors extracted based on the load data and the environmental data, and operation state characteristics determined based on the model parameter data and the operation data; and constructing a distribution transformer life prediction model, obtaining real-time monitoring data of the distribution transformer, and predicting the life of the distribution transformer by adopting the distribution transformer life prediction model at least according to the real-time monitoring data and the characteristic data. The problem of low accuracy of predicting the service life of the distribution transformer under complex working conditions in the prior art is solved.

Description

Technical Field

[0001] The present application relates to the technical field of life prediction of distribution transformers, and in particular to a method, device, computer-readable storage medium and electronic device for life prediction of distribution transformers. Background Art

[0002] As a key equipment in the power system, the safe and stable operation of distribution transformers is crucial to ensuring the quality of power supply. With the continuous growth of power demand and the development of smart grids, higher requirements are placed on the operating status monitoring and prediction capabilities of distribution transformers.

[0003] In terms of distribution transformer load forecasting, traditional methods such as time series analysis (such as ARIMA model) can capture the linear change law of load to a certain extent, but it is difficult to deal with nonlinear characteristics and complex influencing factors (such as meteorological factors, environmental factors, etc.) in load data. Although neural network models (such as BiLSTM-AM) have advantages in dealing with nonlinear problems, single models have limitations and are not adaptable enough when facing a changing operating environment. At the same time, previous data collection is often limited to the electrical parameters of the distribution transformer itself, lacking comprehensive integration of external environmental factors (meteorological, geographical, etc.), resulting in incomplete input information for the prediction model, affecting the prediction accuracy.

[0004] In terms of distribution transformer life prediction, traditional methods based on empirical formulas or simple statistical analysis cannot accurately reflect the aging process of distribution transformers under complex working conditions. Some existing prediction models fail to fully consider the impact of load fluctuations, environmental factors and their synergistic effects on distribution transformer life, and there are deficiencies in model updating and prediction accuracy assessment, making it difficult to achieve dynamic and accurate prediction of distribution transformer life, which is not conducive to the timely formulation of reasonable operation and maintenance strategies. Summary of the invention

[0005] The main purpose of the present application is to provide a method, device, computer-readable storage medium and electronic device for predicting the life of a distribution transformer, so as to at least solve the problem of low accuracy in predicting the life of a distribution transformer under complex working conditions in the prior art.

[0006] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a life prediction method for a distribution transformer is provided, comprising: obtaining relevant data of the distribution transformer, wherein the relevant data includes load data, operation data, model parameter data and environmental data of the distribution transformer; determining characteristic data of the distribution transformer according to the relevant data, wherein the characteristic data includes load characteristics extracted from the load data using short-time Fourier transform technology, relevant characteristics of load fluctuation and environmental factors extracted based on the load data and the environmental data, and operating state characteristics determined based on the model parameter data and the operation data; constructing a distribution transformer life prediction model, wherein the distribution transformer life prediction model is obtained by training using multiple groups of training data, each group of training data in the multiple groups of training data includes loads in historical data, environmental factors and corresponding actual remaining life data of the distribution transformer; obtaining real-time monitoring data of the distribution transformer, and predicting the life of the distribution transformer using the distribution transformer life prediction model at least based on the real-time monitoring data and the characteristic data.

[0007] Optionally, after obtaining the relevant data of the distribution transformer, the method further includes: using a data compression technology based on wavelet transform to compress the relevant data to obtain compressed data, and using a dynamic threshold calculation technology to determine abnormal values ​​in the compressed data; using a weighted average correction technology to correct the abnormal values ​​in the compressed data to obtain pre-processed relevant data.

[0008] Optionally, the life of the distribution transformer is predicted using the distribution transformer life prediction model at least based on the real-time monitoring data and the characteristic data, including: constructing a distribution transformer load prediction model based on a multi-scale convolutional neural network, wherein the distribution transformer life prediction model is trained using multiple sets of training data, and each set of training data in the multiple sets of training data includes operating data of the distribution transformer in historical data and load data at the next moment corresponding to the operating data; inputting the real-time monitoring data into the distribution transformer load prediction model to generate predicted load data, and adjusting the predicted load data based on the relationship between the environmental data and the load correction amount to obtain predicted corrected load data; and predicting the life of the distribution transformer using the distribution transformer life prediction model based on the predicted corrected load data, the real-time monitoring data and the characteristic data.

[0009] Optionally, a distribution transformer load forecasting model based on a multi-scale convolutional neural network is constructed, including: constructing multiple load forecasting models with different structures or algorithms, wherein the load forecasting model includes a long short-term memory network model and an ARIMA model; in the process of training each of the load forecasting models, calculating the performance indicators of each of the load forecasting models, and determining the model weights of each of the load forecasting models according to the performance indicators; and fusing the load forecasting models according to the model weights to obtain the distribution transformer load forecasting model.

[0010] Optionally, calculating the performance index of each load forecasting model includes: using a mean absolute percentage error formula: Determine the performance index of each load forecasting model, where MAPE m For the performance index, is the load value predicted by the model, L(t) is the true load value, and N is the number of samples.

[0011] Optionally, short-time Fourier transform technology is used to extract load characteristics from the load data, including: using the short-time Fourier transform technology to determine the time-frequency diagram of the load data, and extracting energy distribution characteristics from the time-frequency diagram; determining the mean and variance of the energy distribution based on the energy distribution characteristics, and determining the load characteristics of the load data based on the mean and the variance.

[0012] Optionally, after determining the characteristic data of the distribution transformer according to the relevant data, the method further includes: determining the information entropy of each characteristic data; and screening the characteristic data according to the information entropy to obtain the screened characteristic data.

[0013] According to another aspect of the present application, a life prediction device for a distribution transformer is provided, comprising: an acquisition unit, used to acquire relevant data of the distribution transformer, wherein the relevant data includes load data, operation data, model parameter data and environmental data of the distribution transformer; a first determination unit, used to determine characteristic data of the distribution transformer according to the relevant data, wherein the characteristic data includes load characteristics extracted from the load data using short-time Fourier transform technology, relevant characteristics of load fluctuation and environmental factors extracted based on the load data and the environmental data, and operating state characteristics determined based on the model parameter data and the operation data; a construction unit, used to construct a distribution transformer life prediction model, wherein the distribution transformer life prediction model is obtained by training using multiple groups of training data, each group of training data in the multiple groups of training data includes loads in historical data, environmental factors and corresponding actual remaining life data of the distribution transformer; a prediction unit, used to acquire real-time monitoring data of the distribution transformer, and predict the life of the distribution transformer using the distribution transformer life prediction model at least based on the real-time monitoring data and the characteristic data.

[0014] According to another aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute any one of the life prediction methods for distribution transformers.

[0015] According to another aspect of the present application, an electronic device is provided, comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include a method for executing any one of the life prediction methods for a distribution transformer.

[0016] The technical scheme of the present application is applied to obtain relevant data of the distribution transformer, wherein the relevant data includes the load data, operation data, model parameter data and environmental data of the distribution transformer; the characteristic data of the distribution transformer is determined according to the relevant data, wherein the characteristic data includes the load characteristics extracted from the load data by using the short-time Fourier transform technology, the relevant characteristics of the load fluctuation and the environmental factors extracted based on the load data and the environmental data, and the operating state characteristics determined based on the model parameter data and the operating data; a distribution transformer life prediction model is constructed, wherein the distribution transformer life prediction model is obtained by training using multiple sets of training data, and each set of training data in the multiple sets of training data includes the load, environmental factors and the corresponding distribution transformer actual remaining life data in the historical data; the real-time monitoring data of the distribution transformer is obtained, and the distribution transformer life prediction model is used to predict the life of the distribution transformer at least according to the real-time monitoring data and the characteristic data. By combining the characteristics of the environmental data of the distribution transformer, the accuracy of the distribution transformer life prediction is improved, and the trend of the operating state change of the distribution transformer can be grasped more accurately, the reliable operation of the distribution transformer is guaranteed and the overall efficiency of the power system is improved, and the problem of low accuracy in predicting the life of the distribution transformer under complex working conditions in the prior art is solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings constituting part of the present application are used to provide a further understanding of the present application. The exemplary embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0018] Figure 1 A hardware structure block diagram of a mobile terminal for executing a life prediction method for a distribution transformer provided in an embodiment of the present application is shown;

[0019] Figure 2 A schematic flow chart of a life prediction method for a distribution transformer provided according to an embodiment of the present application is shown;

[0020] Figure 3 A schematic flow chart of a specific life prediction method for a distribution transformer provided according to an embodiment of the present application is shown;

[0021] Figure 4 A structural block diagram of a life prediction device for a distribution transformer provided according to an embodiment of the present application is shown.

[0022] The above drawings include the following reference numerals:

[0023] 102, processor; 104, memory; 106, transmission device; 108, input and output devices. DETAILED DESCRIPTION

[0024] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0025] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0026] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0027] As introduced in the background technology, the prior art has low accuracy in predicting the life of distribution transformers under complex working conditions. In order to solve the problem of low accuracy in predicting the life of distribution transformers under complex working conditions in the prior art, the embodiments of the present application provide a method, device, computer-readable storage medium and electronic device for predicting the life of a distribution transformer.

[0028] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0029] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking running on a mobile terminal as an example, Figure 1 FIG. 1 is a hardware structure block diagram of a mobile terminal of a method for predicting the life of a distribution transformer according to an embodiment of the present invention. Figure 1 As shown, the mobile terminal may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data, wherein the mobile terminal may also include a transmission device 106 and an input / output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the mobile terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0030] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the life prediction method of the distribution transformer in the embodiment of the present invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, the above method is implemented. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories can be connected to the mobile terminal via a network. Examples of the above network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof. The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the mobile terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0031] In this embodiment, a life prediction method for a distribution transformer running on a mobile terminal, a computer terminal or a similar computing device is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0032] Figure 2 FIG. 1 is a flow chart of a method for predicting the life of a distribution transformer according to an embodiment of the present application. Figure 2 As shown, the method comprises the following steps:

[0033] Step S201, obtaining relevant data of the distribution transformer, wherein the relevant data includes load data, operation data, model parameter data and environmental data of the distribution transformer;

[0034] Specifically, load data includes power, voltage, and current data of distribution transformers; model parameter data includes rated capacity and model of distribution transformers; environmental data includes real-time meteorological data (including temperature, humidity, air pressure, etc.) and distribution transformer geographical location-related data in the geographic information system (GIS) (such as altitude, surrounding environment type, etc.). Through the fusion collection of multi-source data, factors affecting the operating status and life of distribution transformers can be considered more comprehensively.

[0035] Step S202, determining characteristic data of the distribution transformer according to the relevant data, wherein the characteristic data includes load characteristics extracted from the load data using short-time Fourier transform technology, characteristics related to load fluctuations and environmental factors extracted based on the load data and the environmental data, and operating state characteristics determined based on the model parameter data and the operating data;

[0036] Specifically, the cross-correlation coefficient between the load and the environmental factors is calculated, and the relevant characteristics of the load fluctuation and the environmental factors are extracted by analyzing the peak value and lag time of the cross-correlation coefficient. The load characteristics in the above load data are extracted by using short-time Fourier transform technology, which can reflect the distribution of the load on different frequency components and help capture the frequency characteristics of load changes. These characteristics can reflect the driving effect of environmental factors on load changes and provide more basis for load forecasting.

[0037] Step S203, constructing a distribution transformer life prediction model, wherein the distribution transformer life prediction model is obtained by training using multiple sets of training data, and each set of training data in the multiple sets of training data includes loads and environmental factors in historical data and corresponding actual remaining life data of the distribution transformer;

[0038] Specifically, the distribution transformer life prediction model is a distribution transformer life prediction model based on a hybrid kernel function support vector machine (HK-SVM). For the distribution transformer life prediction, a HK-SVM model is constructed, which combines the advantages of the polynomial kernel function and the Gaussian kernel function.

[0039] Step S204, obtaining real-time monitoring data of the distribution transformer, and predicting the life of the distribution transformer using the distribution transformer life prediction model at least based on the real-time monitoring data and the characteristic data.

[0040] Specifically, in the dynamic prediction of the life of the distribution transformer based on real-time monitoring data, as the distribution transformer is running, new real-time monitoring data (such as voltage, current, temperature, etc.) are continuously obtained, and these real-time data are used to update the model input to dynamically predict the life of the distribution transformer.

[0041] Through this embodiment, the relevant data of the distribution transformer is obtained, wherein the relevant data includes the load data, operation data, model parameter data and environmental data of the distribution transformer; the characteristic data of the distribution transformer is determined according to the relevant data, wherein the characteristic data includes the load characteristics extracted from the load data by using the short-time Fourier transform technology, the relevant characteristics of the load fluctuation and the environmental factors extracted based on the load data and the environmental data, and the operating state characteristics determined based on the model parameter data and the operating data; a distribution transformer life prediction model is constructed, wherein the distribution transformer life prediction model is obtained by training using multiple sets of training data, and each set of training data in the multiple sets of training data includes the load, environmental factors and the corresponding actual remaining life data of the distribution transformer in the historical data; the real-time monitoring data of the distribution transformer is obtained, and the life of the distribution transformer is predicted by using the distribution transformer life prediction model at least according to the real-time monitoring data and the characteristic data. By combining the characteristics of the environmental data of the distribution transformer, the accuracy of the life prediction of the distribution transformer is improved, and the trend of the operating state change of the distribution transformer can be grasped more accurately, the reliable operation of the distribution transformer is guaranteed and the overall efficiency of the power system is improved, and the problem of low accuracy in predicting the life of the distribution transformer under complex working conditions in the prior art is solved.

[0042] In the specific implementation process, after obtaining the relevant data of the distribution transformer, the above method also includes: using a data compression technology based on wavelet transform to compress the above relevant data to obtain compressed data, and using a dynamic threshold calculation technology to determine the abnormal values ​​in the above compressed data; using a weighted average correction technology to correct the above abnormal values ​​in the above compressed data to obtain pre-processed relevant data.

[0043] This method uses data compression technology to compress the collected data in real time, reducing the data transmission bandwidth requirements and improving data transmission efficiency. Using a data compression algorithm based on wavelet transform, the data volume is compressed to a certain proportion of the original without losing key information. The weighted average correction method can effectively correct abnormal values ​​while retaining the overall trend of the data, improve the accuracy and reliability of the data, and provide a better data basis for subsequent distribution transformer load and life prediction.

[0044] Specifically, the life of the distribution transformer is predicted at least based on the real-time monitoring data and the characteristic data using the distribution transformer life prediction model, including: constructing a distribution transformer load prediction model based on a multi-scale convolutional neural network, wherein the distribution transformer life prediction model is trained using multiple sets of training data, and each set of training data in the multiple sets of training data includes the operation data of the distribution transformer in the historical data and the load data at the next moment corresponding to the operation data; inputting the real-time monitoring data into the distribution transformer load prediction model to generate predicted load data, and adjusting the predicted load data based on the relationship between the environmental data and the load correction amount to obtain predicted corrected load data; predicting the life of the distribution transformer using the distribution transformer life prediction model based on the predicted corrected load data, the real-time monitoring data and the characteristic data.

[0045] This method establishes a distribution transformer load forecasting model based on a multi-scale convolutional neural network (MS-CNN). The MS-CNN model contains multiple convolutional layers of different scales to adapt to the characteristics of load data at different time scales.

[0046] Assume that the input load data sequence is X = (x1, x2, ... x T ), where x t It represents the load feature vector at time t (composed of multiple features obtained in the feature extraction stage), and T is the sequence length.

[0047] The first convolution layer uses a larger convolution kernel size, k1×1 (k can be set according to the long-term characteristics of the load data, such as k1=12, corresponding to 12 time steps) to capture the long-term trend characteristics of the load data. The convolution operation formula is: where w 1i is the convolution kernel weight, b1 is the bias, σ is the activation function (such as ReLu function σ(x) = max(0,x)), and C1(t) is the output feature of the first convolution layer at time t.

[0048] Subsequent convolutional layers use gradually decreasing convolution kernel sizes, such as k2×1 (k2=6) and k3×1 (k3=6), to extract local features of load data at different time scales. The second convolution operation formula is: The third layer convolution operation formula is:

[0049] Finally, the output of the convolutional layer is mapped to the predicted load value through the fully connected layer The calculation formula of the fully connected layer is Where n is the number of output features of the last convolutional layer, w 4i and b4 are the weights and biases of the fully connected layer.

[0050] More specifically, a distribution transformer load forecasting model based on a multi-scale convolutional neural network is constructed, including: constructing multiple load forecasting models with different structures or algorithms, wherein the load forecasting models include long short-term memory network models and ARIMA models; in the process of training each of the above load forecasting models, calculating the performance indicators of each of the above load forecasting models, and determining the model weights of each of the above load forecasting models according to the above performance indicators; fusing the above load forecasting models according to the above model weights to obtain the above distribution transformer load forecasting model.

[0051] This method uses multiple load forecasting models with different structures or algorithms (such as the bidirectional long short-term memory network (BiLSTM-AM) with fusion attention mechanism built in the model construction and training stage, and other classic forecasting models such as ARIMA model, etc.) to make forecasts, and then dynamically weights and fuses the forecast results according to the performance of the model. This method of multi-model fusion and dynamic weighting can comprehensively utilize the advantages of different models to improve the accuracy and stability of load forecasting.

[0052] Further, the performance index of each of the above load forecasting models is calculated, including: using the mean absolute percentage error formula: Determine the above performance indicators of each of the above load forecasting models, among which MAPE m For the above performance indicators, is the load value predicted by the model, L(t) is the true load value, and N is the number of samples.

[0053] This method calculates the performance index of each model, uses the mean absolute percentage error (MAPE) to measure the accuracy of the model on the training set or validation set, and calculates the dynamic weight w according to the model performance index. m , a weighted method based on the ranking of performance indicators is adopted, such as In this way, the model with better performance (smaller MAPE) has a larger weight. This method of multi-model fusion and dynamic weighting can comprehensively utilize the advantages of different models to improve the accuracy and stability of load forecasting.

[0054] Furthermore, short-time Fourier transform technology is used to extract load characteristics from the above-mentioned load data, including: using the above-mentioned short-time Fourier transform technology to determine the time-frequency diagram of the above-mentioned load data, and extracting energy distribution characteristics from the above-mentioned time-frequency diagram; determining the mean and variance of the energy distribution according to the above-mentioned energy distribution characteristics, and determining the above-mentioned load characteristics of the above-mentioned load data according to the above-mentioned mean and the above-mentioned variance.

[0055] This method uses STFT to obtain the time-frequency diagram of load data and extracts energy distribution features from the time-frequency diagram. These energy features can reflect the distribution of load on different frequency components and help capture the frequency characteristics of load changes.

[0056] Specifically, after determining the characteristic data of the distribution transformer according to the relevant data, the method further includes: determining the information entropy of each characteristic data; screening the characteristic data according to the information entropy to obtain the screened characteristic data.

[0057] This method uses the basic parameters and operating data of the distribution transformer to calculate the operating status characteristics of the distribution transformer in the feature importance evaluation and selection based on information entropy. The information entropy of each extracted feature is calculated to evaluate its uncertainty and importance, and the features are sorted according to the size of the information entropy, and the features with smaller information entropy are selected. This method can automatically screen out features containing more effective information and reduce redundant features.

[0058] In order to enable those skilled in the art to more clearly understand the technical solution of the present application, the implementation process of the life prediction method of the distribution transformer of the present application will be described in detail below in combination with specific embodiments.

[0059] This embodiment relates to a specific life prediction method for a distribution transformer, such as Figure 3 As shown, the following steps are included:

[0060] Step S100: data collection and preprocessing;

[0061] Step S200, feature extraction and selection;

[0062] Step S300: model building and training;

[0063] Step S400, load and life prediction;

[0064] Wherein, step S100 includes the following steps:

[0065] S110, multi-source data fusion collection;

[0066] In the multi-source data fusion collection, not only load data (such as power, current, voltage, etc.) and basic parameters (such as rated capacity, model, etc.) are obtained from the traditional distribution transformer monitoring system, but also real-time meteorological data (including temperature, humidity, air pressure, etc.) from the meteorological department and distribution transformer geographical location related data (such as altitude, surrounding environment type, etc.) in the geographic information system (GIS) are integrated. Through the fusion collection of multi-source data, the factors affecting the operating status and life of the distribution transformer can be considered more comprehensively.

[0067] For load data, high-precision sensors are used for real-time collection to ensure the accuracy and timeliness of the data. The sampling frequency of the sensor can be adaptively adjusted according to the dynamic characteristics of the distribution transformer operation. When the load fluctuates greatly, the sampling frequency can be increased to capture more detailed load change information.

[0068] S120, distributed data collection architecture;

[0069] In the distributed data acquisition architecture, a distributed data acquisition network is constructed, and data acquisition terminals are set up at multiple key nodes of the distribution transformer (such as the high-voltage side, low-voltage side, windings, etc.) to achieve all-round monitoring of the distribution transformer operation status. Each acquisition terminal transmits data to the data center in real time through wireless communication technology (such as 5G, LoRa, etc.) to ensure the timeliness and integrity of the data.

[0070] Data compression technology is used to compress the collected data in real time, reducing the data transmission bandwidth requirements and improving data transmission efficiency. Using a data compression algorithm based on wavelet transform, the data volume can be compressed to a certain proportion of the original without losing key information.

[0071] S130, dynamic threshold calculation;

[0072] In order to more accurately detect abnormal values ​​in load data, we introduce a dynamic threshold calculation method. Let L(t) be the load value at time t. First, calculate the average value of the n data points before time t. and standard deviation σ(t). Calculate the mean The formula is The formula for calculating the standard deviation σ(t) is

[0073]

[0074] Then determine the dynamic upper threshold and dynamic lower threshold Where k is a threshold coefficient determined based on data, distribution characteristics, and experience. The value of k can be determined by analyzing historical data and simulation experiments, and is generally between 2 and 3. When the data fluctuates greatly, k takes a larger value to ensure that outliers can be effectively detected.

[0075] S140, outlier detection;

[0076] In outlier detection, for each load value L(t) at time t, it is compared with the dynamic upper threshold U(t) and the dynamic lower threshold L(t). If L(t)>U(t) or L(t)<U(t), the data point is judged as an outlier. This dynamic threshold detection method can adapt to the changes of load data in different operating states and time stages, avoiding the misjudgment or missed judgment that may be caused by the fixed threshold method.

[0077] S150, outlier correction;

[0078] In the abnormal value correction, when an abnormal value is detected, it is corrected by using the weighted average correction method. Suppose the corrected load value is Lcorr (t), then Where α is the weight coefficient determined according to the degree of deviation of the outlier.

[0079] Calculate the degree of deviation of outliers The value of α is determined according to the size of d. When d is large, it means that the abnormal value deviates far from the normal range. At this time, α takes a smaller value, 0.1-0.3, to reduce the impact of the abnormal value on the correction result: when d is small, α takes a larger value, 0.7-0.9, to make the correction result closer to the trend of the original data. Such a weighted average correction method can effectively correct the abnormal value while retaining the overall trend of the data, improve the accuracy and reliability of the data, and provide a better data basis for the subsequent distribution transformer load and life prediction.

[0080] Wherein, step S200 includes the following steps:

[0081] S210, load feature extraction based on time-frequency analysis;

[0082] In the load feature extraction based on time-frequency analysis, the load data L(t) after preprocessing (such as outlier detection and correction based on dynamic threshold, denoising and data reconstruction based on wavelet transform, normalization based on time series decomposition and adaptive weighting, etc.) is subjected to time-frequency analysis using short-time Fourier transform (STFT). The formula of STFT is: Where w(t) is the window function (such as Hanning window).

[0083] The time-frequency diagram of the load data is obtained by STFT, and the energy distribution characteristics are extracted from the time-frequency diagram. Assume that the energy at frequency f in the time-frequency diagram is Calculate the mean of the energy distribution and variance (F is the frequency range). These energy characteristics can reflect the distribution of load on different frequency components and help capture the frequency characteristics of load changes.

[0084] S220, load fluctuation feature extraction integrating environmental factors;

[0085] In the extraction of load fluctuation characteristics integrating environmental factors, the influence of environmental data (such as temperature T(t), humidity H(t), etc., obtained through multi-source data fusion collection) on load fluctuation is taken into account, and the cross-correlation coefficient between load and environmental factors is calculated. Assume that the cross-correlation coefficient r between load data L(t) and temperature data T(t) is LT (τ) is in and are the means of load and temperature data, respectively.

[0086] By analyzing the peak value and lag time of the cross-correlation coefficient, the correlation characteristics between load fluctuation and environmental factors are extracted. LT (τ) reaches its maximum value at τ = 3 (time interval), which means that the impact of temperature change on load fluctuation has a lag of 3 time intervals. These characteristics can reflect the driving effect of environmental factors on load changes and provide more basis for load forecasting.

[0087] S230, Feature importance evaluation and selection based on information entropy;

[0088] The basic parameters of distribution transformers (such as rated capacity S rated , short circuit impedance Z sc The load factor of the distribution transformer is calculated by using the operating data (such as voltage U(t), current I(t), etc.) and the operating status characteristics of the distribution transformer. (in is the apparent power, P(t) is the active power, Q(t) is the reactive power) and the impedance change rate (in is the running impedance).

[0089] These operating status characteristics can reflect the working intensity and internal electrical characteristics changes of the distribution transformer, which is of great significance for the life prediction of the distribution transformer because the aging and loss degree of the distribution transformer will be different under different load rates and impedance changes.

[0090] In feature importance evaluation and selection based on information entropy, the information entropy of each extracted feature is calculated to evaluate its uncertainty and importance. Let feature f i The value of x ij (j=1,2,…,m is the sample number), feature f i The probability distribution p(ij) is (where n(ij) is the value x ij The number of samples), then the feature f i Information entropy

[0091]

[0092] Sort the features according to their information entropy and select features with smaller information entropy (i.e., smaller uncertainty and higher importance). Set a threshold T H , when H(f i )<T H When , retain the feature f i , otherwise it will be discarded. This method can automatically filter out features containing more effective information and reduce redundant features.

[0093] S240, feature correlation screening based on grey correlation analysis;

[0094] In the feature correlation screening based on grey correlation analysis, the grey correlation between features and between features and target variables (such as the remaining service life RUL of distribution transformers) is calculated. i (feature sequence), then the grey relational degree where Δ ik =|x0(k)-x i (k) is the difference between x0 and x at the kth sample point i The absolute difference of ρ is the resolution coefficient.

[0095] Wherein, step S300 includes the following steps:

[0096] S310, establishing a distribution transformer load prediction model based on a multi-scale convolutional neural network (MS-CNN);

[0097] In establishing a distribution transformer load forecasting model based on a multi-scale convolutional neural network (MS-CNN), an MS-CNN model is constructed, which contains multiple convolutional layers of different scales to adapt to the characteristics of load data at different time scales. Assume that the input load data sequence is X = (x1, x2, ... x T ), where x t It represents the load feature vector at time t (composed of multiple features obtained in the feature extraction stage), and T is the sequence length.

[0098] The first convolution layer uses a larger convolution kernel size, k1×1 (k can be set according to the long-term characteristics of the load data, such as k1=12, corresponding to 12 time steps) to capture the long-term trend characteristics of the load data. The convolution operation formula is: where w 1i is the convolution kernel weight, b1 is the bias, σ is the activation function (such as ReLu function σ(x) = max(0,x)), and C1(t) is the output feature of the first convolution layer at time t.

[0099] Subsequent convolutional layers use gradually decreasing convolution kernel sizes, such as k2×1 (k2=6) and k3×1 (k3=6), to extract local features of load data at different time scales. The second convolution operation formula is: The third layer convolution operation formula is:

[0100] Finally, the output of the convolutional layer is mapped to the predicted load value through the fully connected layer The calculation formula of the fully connected layer is Where n is the number of output features of the last convolutional layer, w4i and b4 are the weights and biases of the fully connected layer.

[0101] S320, establishing a distribution transformer life prediction model based on a hybrid kernel function support vector machine (HK-SVM);

[0102] In the establishment of the distribution transformer life prediction model based on the hybrid kernel function support vector machine (HK-SVM), the HK-SVM model is constructed for the distribution transformer life prediction, which combines the advantages of the polynomial kernel function and the Gaussian kernel function. Assume that the training sample is (x i ,y i )(i=1,2,…,N,x i is the feature vector, y i is the remaining life of the corresponding distribution transformer).

[0103] Polynomial kernel function K p (x i ,y i )=(x i T x j +c) d (where c is a constant and d is the degree of the polynomial) can better handle nonlinear relationships in low-dimensional feature spaces. The Gaussian kernel function (where σ is the kernel width parameter) has good mapping ability for high-dimensional data.

[0104] The decision function of HK-SVM is where α i is the Lagrange multiplier, b is the bias, and the kernel function K(x,x i )=μK p (x,x i )+(1-μ)K g (x,x i )(μ is the mixing ratio coefficient, 0<μ<1). By adjusting the value of μ, the effects of the two kernel functions in the model can be balanced, so that the model can better adapt to the distribution transformer life prediction under different feature combinations.

[0105] Wherein, step S400 includes the following steps:

[0106] S410, distribution transformer load forecasting based on multi-model fusion and dynamic weighting;

[0107] In the distribution transformer load forecasting based on multi-model fusion and dynamic weighting, multiple load forecasting models with different structures or algorithms (such as the bidirectional long short-term memory network (BiLSTM-AM) with fusion attention mechanism built in the model construction and training stage, and other classic forecasting models such as ARIMA model, etc.) are used for forecasting, and then the forecast results are dynamically weighted and fused according to the performance of the model. Assume that there are M forecasting models in total, and the predicted load value of the mth model at time t+1 is

[0108] Calculate the performance indicators of each model, using the mean absolute percentage error (MAPE) to measure the accuracy of the model on the training set or validation set. The MAPE calculation formula for model m is: Where L(t) is the true load value and N is the number of samples.

[0109] Calculate dynamic weight w based on model performance indicators m , a weighted method based on the ranking of performance indicators is adopted, such as In this way, the model with better performance (smaller MAPE) has a larger weight.

[0110] The final fusion predicted load value This method of multi-model fusion and dynamic weighting can comprehensively utilize the advantages of different models to improve the accuracy and stability of load forecasting.

[0111] S420, load forecast correction considering the time-varying influence of meteorological factors;

[0112] In the load forecast correction considering the time-varying influence of meteorological factors, after obtaining the preliminary load forecast results, the time-varying influence of meteorological factors is further considered for correction. Assume that the meteorological factor vector is W(t) = (T(t), H(t), ...) (T(t) is temperature, H(t) is humidity, etc.), and adjust the forecast results by establishing a relationship model between meteorological factors and load correction.

[0113] A method based on nonlinear function fitting is adopted. Assuming that the relationship between the load correction value ΔL(t+1) and the meteorological factor is ΔL(t+1)=f(W(t),…), a neural network (such as a simple multilayer perceptron) can be used to fit the function. The input of the neural network is the meteorological factor vector W(t), and the output is the load correction value ΔL(t+1).

[0114] When training the neural network, the meteorological factors in the historical data and the difference between the corresponding actual load value and the initial forecast value are used as training samples. The loss function can be the mean square error (MSE), that is, Where ΔL(t) is the actual load correction, is the load correction value predicted by the neural network.

[0115] The final revised load forecast value is By considering the time-varying impact of meteorological factors, the accuracy of load forecasting under different meteorological conditions can be improved.

[0116] S430, dynamic prediction of distribution transformer life based on real-time monitoring data;

[0117] In the dynamic prediction of distribution transformer life based on real-time monitoring data, as the distribution transformer is running, new real-time monitoring data (such as voltage, current, temperature, etc.) are continuously obtained, and these real-time data are used to update the model input to dynamically predict the distribution transformer life. Suppose the real-time monitoring data sequence is X new =(x new,1 ,x new,2 ,…,x new,k )(k is the number of newly acquired data points), and combines it with historical data (after feature extraction and selection) to construct a dynamically updated feature vector.

[0118] For the life prediction model based on kernel principal component analysis (KPCA) and support vector regression (SVR) (built in the model building and training phase), the new data is first merged with the historical data, and KPCA dimension reduction is performed again (following the previous KPCA dimension reduction method, but using the updated data set) to obtain a new low-dimensional feature vector z new .

[0119] Then the new low-dimensional feature vector z new Input into the trained support vector regression (SVR) model to obtain the updated remaining life prediction value of the distribution transformer The SVR model recalculates the prediction results based on the new input data. The formula is: (where α i , z i and b are the trained SVR model parameters, K(z new ,z i ) is the kernel function that calculates the similarity between new data and historical data in low-dimensional space).

[0120] In order to evaluate the reliability of model prediction, confidence interval estimation is introduced. A confidence interval estimation based on the Bootstrap method is adopted to generate multiple new data sets (B, B is generally 100-1000) by performing random sampling with replacement on the original data set (including historical data and newly acquired real-time data).

[0121] For each sampled data set, KPCA dimension reduction and SVR prediction are performed according to the above method to obtain B predicted remaining life values Then calculate the mean of the predicted values and standard deviation

[0122] The final confidence interval is in For standard normal distribution Quantile (when the confidence level is 95%, α = 0.05, ). Through confidence interval estimation, we can have a more comprehensive understanding of the uncertainty of distribution transformer life prediction and provide richer information for operation and maintenance decisions.

[0123] S440, life prediction correction considering the synergy between load and environment;

[0124] In the correction of life prediction considering the synergistic effect of load and environment, the influence of the synergistic effect of load and environmental factors (such as temperature, humidity, etc.) on the life of distribution transformer is further considered to correct the prediction results. Assume that there is a complex nonlinear relationship RUL=g(L,E) between the life of distribution transformer RUL and load L and environmental factors E (such as E=(T,H,…)), and a deep learning-based method is used to fit this relationship. The input layer of the neural network contains load characteristics (such as load rate, fluctuation characteristics, etc., obtained in the feature extraction stage) and environmental factor characteristics, and the output is the predicted value of the remaining life of the distribution transformer. When training the neural network, the load, environmental factors in the historical data and the corresponding actual remaining life data of the distribution transformer are used as training samples. The loss function adopts the mean square error (MSE), that is Where RUL(t) is the actual remaining life, The remaining lifespan predicted by the neural network.

[0125] In the prediction stage, the real-time monitored load data and environmental data are input into the trained neural network to obtain the revised prediction value of the remaining life of the distribution transformer after considering the synergy. By considering the synergistic effect of load and environment, the remaining life of the distribution transformer under actual operating conditions can be predicted more accurately, the reliability and practicality of the prediction results can be improved, and a more scientific basis can be provided for the formulation of operation and maintenance strategies for the distribution transformer.

[0126] This embodiment effectively improves the accuracy of load and life prediction through multi-source data fusion collection, multi-model fusion and dynamic weighting, and can more accurately grasp the trend of distribution transformer operation status changes; on the other hand, the distributed data collection architecture and dynamic threshold outlier processing enhance the adaptability and stability of the method, enabling it to better cope with the complex and changeable actual operating environment. In addition, confidence interval estimation and correction considering synergy provide a scientific basis for operation and maintenance decisions, which helps to optimize resource scheduling, reduce operation and maintenance costs, ensure the reliable operation of distribution transformers and improve the overall efficiency of the power system.

[0127] The embodiment of the present application also provides a life prediction device for a distribution transformer. It should be noted that the life prediction device for a distribution transformer of the embodiment of the present application can be used to execute the life prediction method for a distribution transformer provided in the embodiment of the present application. The device is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware for a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceived.

[0128] The following is an introduction to the life prediction device for the distribution transformer provided in an embodiment of the present application.

[0129] Figure 4 Schematic diagram of a life prediction device for a distribution transformer according to an embodiment of the present application. Figure 4 As shown, the device comprises:

[0130] An acquisition unit 41 is used to acquire relevant data of the distribution transformer, wherein the relevant data includes load data, operation data, model parameter data and environmental data of the distribution transformer;

[0131] A first determining unit 42 is used to determine characteristic data of the distribution transformer according to the relevant data, wherein the characteristic data includes load characteristics extracted from the load data using short-time Fourier transform technology, characteristics related to load fluctuations and environmental factors extracted based on the load data and the environmental data, and operating state characteristics determined based on the model parameter data and the operating data;

[0132] A construction unit 43 is used to construct a distribution transformer life prediction model, wherein the distribution transformer life prediction model is obtained by training using multiple sets of training data, and each set of training data in the multiple sets of training data includes loads and environmental factors in historical data and corresponding actual remaining life data of the distribution transformer;

[0133] The prediction unit 44 is used to obtain the real-time monitoring data of the distribution transformer, and predict the life of the distribution transformer using the distribution transformer life prediction model at least based on the real-time monitoring data and the characteristic data.

[0134] In this embodiment, an acquisition unit is used to acquire relevant data of a distribution transformer, wherein the relevant data includes load data, operation data, model parameter data and environmental data of the distribution transformer; a first determination unit is used to determine characteristic data of the distribution transformer according to the relevant data, wherein the characteristic data includes load characteristics extracted from load data using short-time Fourier transform technology, relevant characteristics of load fluctuation and environmental factors extracted based on load data and environmental data, and operating state characteristics determined based on model parameter data and operation data; a construction unit is used to construct a distribution transformer life prediction model, wherein the distribution transformer life prediction model is obtained by training using multiple sets of training data, and each set of training data in the multiple sets of training data includes loads in historical data, environmental factors and corresponding actual remaining life data of the distribution transformer; a prediction unit is used to acquire real-time monitoring data of the distribution transformer, and predict the life of the distribution transformer using the distribution transformer life prediction model at least based on the real-time monitoring data and the characteristic data. By combining the characteristics of the environmental data of the distribution transformer, the accuracy of the distribution transformer life prediction is improved, and the changing trend of the operating status of the distribution transformer can be grasped more accurately, ensuring the reliable operation of the distribution transformer and improving the overall efficiency of the power system, solving the problem of low accuracy of the existing technology in predicting the life of the distribution transformer under complex working conditions.

[0135] As an optional solution, the device also includes a compression processing unit and a correction processing unit; the compression processing unit is used to compress the above-mentioned relevant data after obtaining the relevant data of the distribution transformer by adopting a data compression technology based on wavelet transform to obtain compressed data, and to determine the abnormal values ​​in the above-mentioned compressed data by adopting a dynamic threshold calculation technology; the correction processing unit is used to correct the above-mentioned abnormal values ​​in the above-mentioned compressed data by adopting a weighted average correction technology to obtain the pre-processed relevant data.

[0136] An optional solution, the prediction unit includes a construction module, an adjustment module and a prediction module; the construction module is used to construct a distribution transformer load prediction model based on a multi-scale convolutional neural network, wherein the above-mentioned distribution transformer life prediction model is obtained by training using multiple groups of training data, and each group of training data in the above-mentioned multiple groups of training data includes the operation data of the above-mentioned distribution transformer in the historical data and the load data at the next moment corresponding to the above-mentioned operation data; the adjustment module is used to input the above-mentioned real-time monitoring data into the above-mentioned distribution transformer load prediction model to generate predicted load data, and adjust the above-mentioned predicted load data based on the relationship between the above-mentioned environmental data and the load correction amount to obtain predicted corrected load data; the prediction module is used to predict the life of the above-mentioned distribution transformer using the above-mentioned distribution transformer life prediction model according to the above-mentioned predicted corrected load data, the above-mentioned real-time monitoring data and the above-mentioned characteristic data.

[0137] An optional solution, the construction module includes a construction submodule, a calculation submodule and a fusion processing submodule; the construction submodule is used to construct multiple load forecasting models with different structures or algorithms, wherein the load forecasting models include long short-term memory network models and ARIMA models; the calculation submodule is used to calculate the performance indicators of each of the above load forecasting models during the training process of each of the above load forecasting models, and determine the model weights of each of the above load forecasting models based on the above performance indicators; the fusion processing submodule is used to fuse the above load forecasting models according to the above model weights to obtain the above distribution transformer load forecasting model.

[0138] In an optional solution, the calculation submodule includes a determination submodule for using a mean absolute percentage error formula: Determine the above performance indicators of each of the above load forecasting models, among which MAPE m For the above performance indicators, is the load value predicted by the model, L(t) is the true load value, and N is the number of samples.

[0139] An optional scheme, the first determination unit includes a first determination module and a second determination module, the first determination module is used to determine the time-frequency diagram of the above-mentioned load data by adopting the above-mentioned short-time Fourier transform technology, and extract the energy distribution characteristics from the above-mentioned time-frequency diagram; the second determination module is used to determine the mean and variance of the energy distribution according to the above-mentioned energy distribution characteristics, and determine the above-mentioned load characteristics of the above-mentioned load data according to the above-mentioned mean and the above-mentioned variance.

[0140] An optional solution, the device also includes a second determination unit and a screening processing unit, the second determination unit is used to determine the information entropy of each of the above-mentioned characteristic data after determining the characteristic data of the above-mentioned distribution transformer according to the above-mentioned related data; the screening processing unit is used to screen the above-mentioned characteristic data according to the above-mentioned information entropy to obtain the characteristic data after screening.

[0141] The life prediction device for the distribution transformer includes a processor and a memory. The acquisition unit, the first determination unit, the construction unit and the prediction unit are stored in the memory as program units, and the processor executes the program units stored in the memory to implement corresponding functions. The modules are all located in the same processor; or, the modules are located in different processors in any combination.

[0142] The processor includes a kernel, which retrieves the corresponding program unit from the memory. One or more kernels can be set, and the problem of low accuracy in predicting the life of distribution transformers under complex working conditions in the prior art can be solved by adjusting kernel parameters.

[0143] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0144] An embodiment of the present invention provides a computer-readable storage medium, which includes a stored program, wherein when the program is running, the device where the computer-readable storage medium is located is controlled to execute the life prediction method for the distribution transformer.

[0145] Specifically, the life prediction method of the distribution transformer includes:

[0146] Step S201, obtaining relevant data of the distribution transformer, wherein the relevant data includes load data, operation data, model parameter data and environmental data of the distribution transformer;

[0147] Step S202, determining characteristic data of the distribution transformer according to the relevant data, wherein the characteristic data includes load characteristics extracted from the load data using short-time Fourier transform technology, characteristics related to load fluctuations and environmental factors extracted based on the load data and the environmental data, and operating state characteristics determined based on the model parameter data and the operating data;

[0148] Step S203, constructing a distribution transformer life prediction model, wherein the distribution transformer life prediction model is obtained by training using multiple sets of training data, and each set of training data in the multiple sets of training data includes loads and environmental factors in historical data and corresponding actual remaining life data of the distribution transformer;

[0149] Step S204, obtaining real-time monitoring data of the distribution transformer, and predicting the life of the distribution transformer using the distribution transformer life prediction model at least based on the real-time monitoring data and the characteristic data.

[0150] An embodiment of the present invention provides a processor, and the processor is used to run a program, wherein the life prediction method of the distribution transformer is executed when the program is run.

[0151] Specifically, the life prediction method of the distribution transformer includes:

[0152] Step S201, obtaining relevant data of the distribution transformer, wherein the relevant data includes load data, operation data, model parameter data and environmental data of the distribution transformer;

[0153] Step S202, determining characteristic data of the distribution transformer according to the relevant data, wherein the characteristic data includes load characteristics extracted from the load data using short-time Fourier transform technology, characteristics related to load fluctuations and environmental factors extracted based on the load data and the environmental data, and operating state characteristics determined based on the model parameter data and the operating data;

[0154] Step S203, constructing a distribution transformer life prediction model, wherein the distribution transformer life prediction model is obtained by training using multiple sets of training data, and each set of training data in the multiple sets of training data includes loads and environmental factors in historical data and corresponding actual remaining life data of the distribution transformer;

[0155] Step S204, obtaining real-time monitoring data of the distribution transformer, and predicting the life of the distribution transformer using the distribution transformer life prediction model at least based on the real-time monitoring data and the characteristic data.

[0156] An embodiment of the present invention provides an electronic device, the device comprising a processor, a memory, and a program stored in the memory and executable on the processor, and when the processor executes the program, at least the following steps are implemented:

[0157] Step S201, obtaining relevant data of the distribution transformer, wherein the relevant data includes load data, operation data, model parameter data and environmental data of the distribution transformer;

[0158] Step S202, determining characteristic data of the distribution transformer according to the relevant data, wherein the characteristic data includes load characteristics extracted from the load data using short-time Fourier transform technology, characteristics related to load fluctuations and environmental factors extracted based on the load data and the environmental data, and operating state characteristics determined based on the model parameter data and the operating data;

[0159] Step S203, constructing a distribution transformer life prediction model, wherein the distribution transformer life prediction model is obtained by training using multiple sets of training data, and each set of training data in the multiple sets of training data includes loads and environmental factors in historical data and corresponding actual remaining life data of the distribution transformer;

[0160] Step S204, obtaining real-time monitoring data of the distribution transformer, and predicting the life of the distribution transformer using the distribution transformer life prediction model at least based on the real-time monitoring data and the characteristic data.

[0161] The devices in this article can be servers, PCs, PADs, mobile phones, etc.

[0162] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program for initializing at least the following method steps:

[0163] Step S201, obtaining relevant data of the distribution transformer, wherein the relevant data includes load data, operation data, model parameter data and environmental data of the distribution transformer;

[0164] Step S202, determining characteristic data of the distribution transformer according to the relevant data, wherein the characteristic data includes load characteristics extracted from the load data using short-time Fourier transform technology, characteristics related to load fluctuations and environmental factors extracted based on the load data and the environmental data, and operating state characteristics determined based on the model parameter data and the operating data;

[0165] Step S203, constructing a distribution transformer life prediction model, wherein the distribution transformer life prediction model is obtained by training using multiple sets of training data, and each set of training data in the multiple sets of training data includes loads and environmental factors in historical data and corresponding actual remaining life data of the distribution transformer;

[0166] Step S204, obtaining real-time monitoring data of the distribution transformer, and predicting the life of the distribution transformer using the distribution transformer life prediction model at least based on the real-time monitoring data and the characteristic data.

[0167] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0168] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0169] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0170] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0171] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0172] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0173] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0174] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0175] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0176] From the above description, it can be seen that the above embodiments of the present application achieve the following technical effects:

[0177] 1) A life prediction method for a distribution transformer of the present application comprises: obtaining relevant data of the distribution transformer, wherein the relevant data comprises load data, operation data, model parameter data and environmental data of the distribution transformer; determining characteristic data of the distribution transformer according to the relevant data, wherein the characteristic data comprises load characteristics extracted from load data by using short-time Fourier transform technology, relevant characteristics of load fluctuation and environmental factors extracted based on load data and environmental data, and operating state characteristics determined based on model parameter data and operation data; constructing a distribution transformer life prediction model, wherein the distribution transformer life prediction model is obtained by training using multiple sets of training data, and each set of training data in the multiple sets of training data comprises loads in historical data, environmental factors and corresponding actual remaining life data of the distribution transformer; obtaining real-time monitoring data of the distribution transformer, and predicting the life of the distribution transformer using the distribution transformer life prediction model at least based on the real-time monitoring data and the characteristic data. By combining the characteristics of the environmental data of the distribution transformer, the accuracy of the distribution transformer life prediction is improved, and the changing trend of the operating status of the distribution transformer can be grasped more accurately, ensuring the reliable operation of the distribution transformer and improving the overall efficiency of the power system, solving the problem of low accuracy of the existing technology in predicting the life of the distribution transformer under complex working conditions.

[0178] 2) A life prediction device for a distribution transformer of the present application comprises: an acquisition unit, used to acquire relevant data of the distribution transformer, wherein the relevant data includes load data, operation data, model parameter data and environmental data of the distribution transformer; a first determination unit, used to determine characteristic data of the distribution transformer according to the relevant data, wherein the characteristic data includes load characteristics extracted from load data using short-time Fourier transform technology, relevant characteristics of load fluctuation and environmental factors extracted based on load data and environmental data, and operating state characteristics determined based on model parameter data and operation data; a construction unit, used to construct a distribution transformer life prediction model, wherein the distribution transformer life prediction model is obtained by training using multiple sets of training data, and each set of training data in the multiple sets of training data includes loads in historical data, environmental factors and corresponding actual remaining life data of the distribution transformer; a prediction unit, used to acquire real-time monitoring data of the distribution transformer, and predict the life of the distribution transformer using the distribution transformer life prediction model at least based on the real-time monitoring data and the characteristic data. By combining the characteristics of the environmental data of the distribution transformer, the accuracy of the distribution transformer life prediction is improved, and the changing trend of the operating status of the distribution transformer can be grasped more accurately, ensuring the reliable operation of the distribution transformer and improving the overall efficiency of the power system, solving the problem of low accuracy of the existing technology in predicting the life of the distribution transformer under complex working conditions.

[0179] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for predicting the life of a distribution transformer, characterized in that: include: Acquire relevant data of the distribution transformer, wherein the relevant data includes load data, operation data, model parameter data and environmental data of the distribution transformer; Determine characteristic data of the distribution transformer according to the relevant data, wherein the characteristic data includes load characteristics extracted from the load data using short-time Fourier transform technology, relevant characteristics of load fluctuation and environmental factors extracted based on the load data and the environmental data, and operating state characteristics determined based on the model parameter data and the operating data; Constructing a distribution transformer life prediction model, wherein the distribution transformer life prediction model is trained using multiple sets of training data, each set of training data in the multiple sets of training data includes loads and environmental factors in historical data and corresponding actual remaining life data of the distribution transformer; Real-time monitoring data of the distribution transformer is acquired, and the life of the distribution transformer is predicted using the distribution transformer life prediction model at least based on the real-time monitoring data and the characteristic data.

2. The method according to claim 1, characterized in that After obtaining the relevant data of the distribution transformer, the method further includes: Using a data compression technique based on wavelet transform to compress the relevant data to obtain compressed data, and using a dynamic threshold calculation technique to determine abnormal values ​​in the compressed data; The abnormal values ​​in the compressed data are corrected by using a weighted average correction technique to obtain pre-processed related data.

3. The method according to claim 1, characterized in that Predicting the life of the distribution transformer using the distribution transformer life prediction model at least according to the real-time monitoring data and the characteristic data, comprises: Constructing a distribution transformer load prediction model based on a multi-scale convolutional neural network, wherein the distribution transformer life prediction model is obtained by training using multiple sets of training data, and each set of training data in the multiple sets of training data includes operating data of the distribution transformer in historical data and load data at the next moment corresponding to the operating data; Inputting the real-time monitoring data into the distribution transformer load prediction model to generate predicted load data, and adjusting the predicted load data based on the relationship between the environmental data and the load correction amount to obtain predicted corrected load data; The life of the distribution transformer is predicted using the distribution transformer life prediction model according to the predicted corrected load data, the real-time monitoring data and the characteristic data.

4. The method according to claim 3, characterized in that Construct a distribution transformer load prediction model based on a multi-scale convolutional neural network, including: Construct multiple load forecasting models with different structures or algorithms, including long short-term memory network model and ARIMA model; During the training of each of the load forecasting models, the performance index of each of the load forecasting models is calculated, and the model weight of each of the load forecasting models is determined according to the performance index; The load forecasting model is fused according to the model weight to obtain the distribution transformer load forecasting model.

5. The method according to claim 4, characterized in that Calculating the performance indicators of each load forecasting model includes: Using the mean absolute percentage error formula: Determine the performance index of each load forecasting model, where MAPE m For the performance index, is the load value predicted by the model, L(t) is the true load value, and N is the number of samples.

6. The method according to claim 1, characterized in that The load characteristics in the load data are extracted using short-time Fourier transform technology, including: Determine the time-frequency diagram of the load data by using the short-time Fourier transform technique, and extract energy distribution characteristics from the time-frequency diagram; A mean and a variance of energy distribution are determined according to the energy distribution characteristics, and the load characteristics of the load data are determined according to the mean and the variance.

7. The method according to claim 1, characterized in that After determining the characteristic data of the distribution transformer according to the relevant data, the method further includes: Determining the information entropy of each of the feature data; The feature data is filtered according to the information entropy to obtain filtered feature data.

8. A life prediction device for a distribution transformer, characterized in that: include: An acquisition unit, used for acquiring relevant data of the distribution transformer, wherein the relevant data includes load data, operation data, model parameter data and environmental data of the distribution transformer; A first determining unit is used to determine characteristic data of the distribution transformer according to the relevant data, wherein the characteristic data includes load characteristics extracted from the load data using short-time Fourier transform technology, relevant characteristics of load fluctuation and environmental factors extracted based on the load data and the environmental data, and operating state characteristics determined based on the model parameter data and the operating data; A construction unit is used to construct a distribution transformer life prediction model, wherein the distribution transformer life prediction model is obtained by training using multiple sets of training data, each set of training data in the multiple sets of training data includes loads and environmental factors in historical data and corresponding actual remaining life data of the distribution transformer; A prediction unit is used to obtain real-time monitoring data of the distribution transformer, and predict the life of the distribution transformer using the distribution transformer life prediction model at least based on the real-time monitoring data and the characteristic data.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the life prediction method for the distribution transformer according to any one of claims 1 to 7.

10. An electronic device, characterized in that: include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include a method for executing the life prediction method of the distribution transformer described in any one of claims 1 to 7.

Images