A method and system for predicting faults in power equipment

Abstract

The application provides a power transformation equipment fault prediction method and system, and relates to the technical field of fault prediction.The method comprises the following steps: obtaining historical operation data of power transformation equipment; inputting the historical operation data into a fault prediction model to obtain a fault prediction result; wherein the fault prediction model performs state value prediction and fault category prediction in each time step of fault prediction.The application can reduce cumulative error in the prediction process and improve the accuracy of fault prediction.

Description

Technical Field

[0001] This invention relates to the field of fault prediction technology, and specifically to a method and system for predicting faults in power equipment. Background Technology

[0002] Substation equipment is a crucial component of the power system, and its failures can lead to power outages, impacting the normal operations of businesses and residents. Therefore, early warning systems for substation equipment failures are essential for timely identification and handling before problems occur, ensuring a stable power supply. Currently, the mainstream method for predicting substation equipment failures is to first predict and then diagnose the fault. Specifically, existing technologies typically predict future equipment states based on historical observed values, and then diagnose the fault based on these predicted values. However, in the prediction phase, each predicted value is calculated based on the previous prediction in a multi-step prediction process. Therefore, if an error or deviation occurs in a previous prediction, it will be propagated to subsequent predictions, leading to error accumulation. When predicting a long-term series, this accumulated error can cause a sharp decline in performance. Furthermore, in the diagnosis phase, the fault diagnosis result is also calculated based on the predicted value, introducing additional errors and reducing the accuracy of fault prediction. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the purpose of this invention is to provide a method and system for predicting faults in power equipment, thereby improving the accuracy of fault prediction.

[0004] To achieve the above objectives, according to some embodiments, a first aspect of the present invention provides a method for predicting faults in power equipment, comprising:

[0005] Obtain historical operating data of power equipment;

[0006] Historical operational data is input into the fault prediction model to obtain fault prediction results;

[0007] In this fault prediction model, at each time step of the fault prediction process, both state value prediction and fault category prediction are performed.

[0008] A second aspect of the present invention provides a fault prediction system for power equipment, comprising:

[0009] The data acquisition module is configured to acquire historical operating data of the power equipment;

[0010] The prediction module is configured to input historical operating data into the fault prediction model to obtain fault prediction results; wherein, the fault prediction model performs state value prediction and fault category prediction at each time step of fault prediction.

[0011] A third aspect of the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to perform the steps of the above-described method for predicting faults in power equipment.

[0012] In a fourth aspect, the present invention provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, complete the steps of the above-described method for predicting faults in power equipment.

[0013] In a fifth aspect, the present invention provides a computer program product comprising a computer program / instructions which, when executed by a processor, implement the steps of the above-described method for predicting faults in power equipment.

[0014] Compared with the prior art, the beneficial effects of the present invention are:

[0015] This invention provides a method and system for predicting faults in power equipment. At each time step of the prediction process, state value prediction and fault category prediction are performed separately. Furthermore, during the model optimization process, the state value prediction task and the fault category prediction task are jointly optimized by multiple tasks. This breaks the traditional fault warning paradigm of "predict first and then diagnose", effectively avoids the error accumulation problem caused by the sequential execution of the two tasks, and improves the accuracy of fault prediction.

[0016] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0017] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0018] Figure 1 This is a flowchart of the method of the present invention;

[0019] Figure 2 Optimize the flowchart for the model;

[0020] Figure 3 This is a model architecture diagram;

[0021] Figure 4 Flowchart for extracting historical time-series features. Detailed Implementation

[0022] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0023] Example 1

[0024] like Figure 1 As shown, Embodiment 1 of the present invention provides a method for predicting faults in power equipment, comprising:

[0025] Obtain historical operating data of power equipment;

[0026] Historical operational data is input into the fault prediction model to obtain fault prediction results; in which the fault prediction model performs state value prediction and fault category prediction at each time step of fault prediction.

[0027] In existing fault prediction technologies, numerical prediction is typically performed first, followed by fault prediction based on the predicted values ​​of future time series. However, this "prediction first, diagnosis later" approach suffers from error accumulation; the error in the predicted values ​​accumulates progressively at each time step, resulting in significant errors when predicting long time series. Therefore, this embodiment breaks away from the traditional "prediction first, diagnosis later" paradigm by adding two branches—state numerical prediction and fault category prediction—at each time step. This couples the numerical prediction and fault prediction tasks during model training, mitigating the error accumulation problem and achieving more accurate fault prediction.

[0028] Taking transformer fault prediction as an example, the model optimization process is as follows: Figure 2 As shown, the model architecture is as follows Figure 3 As shown. First, historical time-series features are extracted. The purpose of this stage is to extract features of the transformer's historical observation time-series sensor parameters, such as oil chromatography, electrical parameters, and meteorological data. First, for the input transformer raw multi-sensor parameter data X={x1,x2,…,x…},… ... h}, where x i (i=1…h)∈R n h is the historical time series length, n is the number of sensor parameters, and R n Let represent an n-dimensional real vector space. Data preprocessing is performed using standardization to ensure that all values ​​conform to a standard normal distribution with a mean of 0 and a standard deviation of 1. The specific operation is shown in the following formula:

[0029]

[0030] Where, x mean It is the mean vector, x std It is the standard deviation vector.

[0031] Secondly, regarding the preprocessed parameter data Linear embedding layers are used to map numerical values ​​to a high-dimensional feature space. The specific operation is shown in the following equation:

[0032]

[0033] Where Linear is a linear embedding layer, F i Let be the high-dimensional feature corresponding to the i-th time step.

[0034] Finally, the features from multiple time steps are sequentially and cyclically input into a gated recurrent unit (GRU), with the temporal hidden state of the previous time step serving as the input to the next GRU, thus completing the extraction of historical temporal features. The specific operation method is shown in the following equation:

[0035] H i+1 =GRU(F i H i )

[0036] Among them, H i H represents the temporal hidden state corresponding to the i-th time step. i+1 This represents the temporal hidden state corresponding to the (i+1)th time step.

[0037] After extracting historical time-series features, state value prediction is performed. The purpose of this stage is to predict the transformer state value for future time steps using historical time-series features. A fully connected layer is added after the GRU unit, taking the temporal hidden state as input and outputting the predicted transformer state value for future time steps. The specific operation is shown in the following equation:

[0038] V i+1 =FC v (H i+1 )

[0039] Among them, FC v V is the fully connected layer responsible for state numerical prediction. i+1 This represents the transformer state value at the (i+1)th time step. It's worth noting that this prediction process is repeated multiple times to obtain the state value for the expected future time step, denoted as V. i+1 As input, the state value V corresponding to the (i+2)th time step can be predicted. i+2 And so on.

[0040] Simultaneously with state value prediction, fault category prediction is also performed. The purpose of this stage is to predict the transformer fault category for future time steps using historical time series features. An additional fully connected layer is added after the GRU unit, which also takes the temporal hidden state as input and outputs the transformer fault category for future time steps. The specific operation is shown in the following equation:

[0041] C i+1 =FC c (H i+1 )

[0042] Among them, FC c C is the fully connected layer responsible for fault category prediction.i+1 This represents the transformer fault category corresponding to the (i+1)th time step. Similarly, to obtain the fault category for the expected future time step, this prediction process is repeated multiple times, using the predicted state value V. i+1 Given the input, we can predict the fault category C corresponding to the (i+2)th time step. i+2 And so on. Specifically, in this embodiment, as... Figure 3 As shown, historical time series data of length h is obtained. For the prediction task of the next f time steps, in each time step, historical time series features (time series hidden states) are extracted through a gated recurrent unit. The extracted historical time series features are then input into two parallel fully connected layers for full connection, thereby performing state value prediction and fault category prediction respectively, and finally obtaining the fault prediction result.

[0043] The established model is trained. The model loss is a weighted sum of the state numerical prediction loss and the fault category prediction loss. The state numerical prediction and fault category prediction tasks correspond to regression loss and classification loss, respectively. Both losses are simultaneously optimized end-to-end for the entire model through multi-task joint optimization. The overall loss function is calculated as follows:

[0044] Loss=α*L reg +(1-α)*L cla

[0045] Among them, L reg For regression loss, L cla The loss is the classification loss, and α is the weight that balances the two losses.

[0046]

[0047] Where MSE is the mean squared error loss, Y i Let f be the actual state value of the transformer corresponding to the i-th future time step, and f be the number of future time steps.

[0048]

[0049] Where CE is the cross-entropy loss, and T i Let i be the actual fault category of the transformer corresponding to the i-th future time step.

[0050] The fault prediction model is iteratively optimized using the loss function Loss to minimize the loss function, improve the model's performance on the training and validation sets, and ultimately achieve accurate prediction of substation faults.

[0051] Example 2

[0052] This embodiment provides a power equipment fault prediction system, including:

[0053] The data acquisition module is configured to acquire historical operating data of the power equipment;

[0054] The prediction module is configured to input historical operating data into the fault prediction model to obtain fault prediction results; wherein, the fault prediction model performs state value prediction and fault category prediction at each time step of fault prediction.

[0055] It should be noted that each module in this embodiment corresponds one-to-one with the steps of the method in Embodiment 1, and their specific implementation processes are the same, so they will not be repeated here.

[0056] Example 3

[0057] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to complete the steps of the method in Embodiment 1.

[0058] Example 4

[0059] This embodiment provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, complete the steps of the method in Embodiment 1.

[0060] Example 5

[0061] This embodiment provides a computer program product, including a computer program / instructions, characterized in that the computer program / instructions, when executed by a processor, implement the steps of the method in Embodiment 1.

[0062] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A power transformer equipment failure prediction method, characterized by, The method comprises: obtaining historical operation data of the power transformation equipment; inputting the historical operation data into a fault prediction model to obtain a fault prediction result; wherein the fault prediction model performs state value prediction and fault category prediction in each time step of fault prediction.

2. A power transformer failure prediction method as claimed in claim 1, characterized by, The fault prediction model extracts historical time sequence features according to the input historical operation data, and performs state value prediction and fault category prediction according to the historical time sequence features.

3. A power transformer failure prediction method as claimed in claim 2, characterized in that, The fault prediction model extracts historical time sequence features through a gated recurrent unit, and performs state value prediction and fault category prediction through two fully connected layers respectively.

4. A power transformer failure prediction method as claimed in claim 1, characterized in that, The loss of the fault prediction model is a weighted sum of state value prediction loss and fault category prediction loss.

5. A power transformer failure prediction method as claimed in claim 4, characterized in that, The state value prediction loss is a regression loss, and a mean square error loss is used for calculation.

6. A power transformer failure prediction method as claimed in claim 4, characterized in that, The fault category prediction loss is a classification loss, and a cross-entropy loss is used for calculation.

7. A power transformer failure prediction system characterized by, The method comprises: a data acquisition module configured to obtain historical operation data of the power transformation equipment; a prediction module configured to input the historical operation data into a fault prediction model to obtain a fault prediction result; wherein the fault prediction model performs state value prediction and fault category prediction in each time step of fault prediction.

8. An electronic device, comprising: A computer program product comprising a memory, a processor, and a computer program stored on the memory, wherein the processor executes the computer program to complete the steps of the method of any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, A computer program product for storing computer instructions, wherein the computer instructions are executed by a processor to complete the steps of the method of any one of claims 1-6.

10. A computer program product comprising computer programs / instructions, characterized in that, The computer program / instructions are executed by the processor to implement the steps of the method of any one of claims 1-6.

Images