Substation dynamic fault probability prediction method and device based on meteorological-equipment multi-physics field coupling

Abstract

The invention discloses a substation dynamic fault probability prediction method and device based on meteorological-equipment multi-physics field coupling, and the method comprises the steps: collecting the surrounding meteorological parameters of a substation and the physics field parameters of key equipment of the substation, sequentially carrying out the noise filtering, data normalization and time-space alignment processing of the collected original data, and obtaining the standardized data; key features are extracted from the standardized data, a comprehensive coupling influence coefficient of response is output according to a coupling model library, and based on the comprehensive coupling influence coefficient, a dynamic linear model and a proportional risk model are updated by adopting Bayesian, and minute-level calculation of fault probability is carried out; the equipment fault probability and influence factors are displayed in real time, and when the fault probability exceeds a dynamic early warning threshold value, sound-light alarm and remote early warning are given out. The method can dynamically predict the fault probability of the transformer substation with high precision, and guarantees the operation safety of the transformer substation.

Description

Technical Field

[0001] This invention belongs to the field of substation safety technology, specifically relating to a method and device for predicting the dynamic fault probability of substations based on the coupling of meteorological and equipment multi-physical fields. Background Technology

[0002] In substation operation and maintenance, accurate prediction of equipment failure probability is crucial for ensuring the stable operation of the power system. However, existing substation failure probability prediction methods have several shortcomings: First, most methods only consider the equipment's own operating parameters or meteorological factors, ignoring the coupling effect between meteorological factors and the equipment's multi-physical fields, which leads to deviations between the prediction results and the actual operating conditions. Secondly, traditional forecasting methods are mostly based on static analysis of historical data, which cannot respond in real time to dynamic changes in meteorological conditions and real-time fluctuations in equipment physical field parameters, making it difficult to meet the needs of real-time operation and maintenance decision-making. Third, meteorological data and equipment data sources for multiple physical fields are scattered, and existing devices lack an effective data fusion mechanism, which can easily affect the accuracy of predictions due to problems such as inconsistent data formats and noise. Furthermore, due to insufficient consideration of complex coupling relationships and dynamic characteristics, existing prediction models have low accuracy in predicting faults in special scenarios, making it difficult to support refined operation and maintenance management. Summary of the Invention

[0003] To address the shortcomings of existing technologies, such as neglecting the multi-physical field coupling effect of meteorology and equipment, insufficient dynamic prediction capability, low data fusion degree, and limited prediction accuracy, this invention provides a substation dynamic fault probability prediction method and device based on meteorology-equipment multi-physical field coupling, which has a reasonable structure, can realize multi-field coupling analysis, timely dynamic updates, high prediction accuracy, and strong adaptability.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: A method for predicting the dynamic fault probability of substations based on meteorological-equipment multi-physics coupling includes: multi-source data acquisition, collecting meteorological parameters around the substation through a meteorological sensor group, including temperature, humidity, wind speed, precipitation, lightning intensity, and air pressure; collecting physical field parameters of key equipment in the substation through an equipment sensor group, including temperature field, electric field, mechanical stress field, humidity field, partial discharge, and insulation resistance; data preprocessing, sequentially performing noise filtering, data normalization, and spatiotemporal alignment on the collected raw data to obtain standardized data; coupling analysis, extracting key features from the standardized data and outputting the comprehensive coupling influence coefficient of the response based on the coupling model library; dynamic fault probability calculation, based on the comprehensive coupling influence coefficient, using a Bayesian update dynamic linear model and a proportional hazards model to perform minute-level calculation of fault probability; model update, using an incremental learning and federated learning framework, periodically optimizing and updating the prediction model using new operational data, fault cases, and multimodal inspection data; result output and early warning, displaying the equipment fault probability and influencing factors in real time, and issuing audible and visual alarms and remote warnings when the fault probability exceeds the dynamic early warning threshold.

[0005] Optionally, the data acquisition frequency in the multi-source data acquisition step can be dynamically adjusted, including: meteorological data is acquired once every 5 minutes under normal operating conditions, and equipment physical field data is acquired once every 2 minutes; when a sudden change in meteorological parameters or abnormal equipment parameters are detected, the acquisition frequency is increased to 10 seconds / time.

[0006] Optionally, in the data preprocessing step, noise filtering employs an adaptive filtering algorithm based on a sliding window, as shown in the formula: ,in w For the width of the sliding window, w j For adaptive weights; data normalization uses a linear normalization method, the formula is as follows: Time alignment is achieved using linear interpolation, and spatial alignment is based on the influence weights calculated using the sensor's spatial coordinates. The formulas are as follows: , .

[0007] Optionally, in the coupling analysis step, the coupling model library contains multiple sets of coupling relationship models to construct the coupling effect equations for extreme disasters. ,in, For dynamic weights, For disaster coupling function, For meteorological intensity parameters, The calculation of the comprehensive coupling influence coefficient includes: calling the matching model according to the substation type and operating scenario, analyzing the coupling influence relationship between meteorological factors and the multiple physical fields of the equipment, and outputting the comprehensive coupling influence coefficient.

[0008] Optionally, in the dynamic failure probability calculation, an improved Monte Carlo sampling method is used to correct the comprehensive coupling coefficient, and the number of samplings is adjusted. N ≥1000, generate the fault probability spatiotemporal matrix, the formula is as follows: .

[0009] Optionally, in the dynamic failure probability calculation step, after the Bayesian update of the dynamic linear model passes, the failure probability is calculated by verifying the probability, as shown in the formula: The proportional risk model formula is: , ,in As the baseline failure rate, ß For the correlation coefficient of the running parameters, X t These are real-time operating parameters.

[0010] Optionally, in the model update step, incremental learning uses mini-batch parameter updates, as shown in the formula: Federated learning aggregates the global model through sample size weighting, as shown in the formula: .

[0011] Optionally, in the result output and early warning steps, the early warning threshold is dynamically set, using the following formula: ,in Based on the threshold, This represents the equipment importance coefficient. This is the normalized coupling effect coefficient.

[0012] A substation dynamic fault probability prediction device based on meteorological-equipment multi-physics coupling includes: a multi-source data acquisition module, comprising a meteorological sensor group for collecting meteorological parameters around the substation and an equipment sensor group for collecting physical field parameters of key equipment in the substation; a data preprocessing module, comprising a noise filtering unit, a data normalization unit, and a spatiotemporal alignment unit, used to sequentially perform noise filtering, data normalization, and spatiotemporal alignment processing on the acquired raw data to obtain standardized data; and a coupling analysis module, comprising a feature extraction unit and a coupling model library. The feature extraction unit is used to extract key features from the preprocessed standardized data, and the coupling model library contains multiple sets of coupling relationship models, which are configured according to the substation type. The system includes a model matching the operational scenario, which analyzes the coupling influence between meteorological factors and equipment multi-physics fields, and outputs a comprehensive coupling influence coefficient; a dynamic fault probability module, which uses Bayesian updates to the dynamic linear model and proportional risk model based on the comprehensive coupling influence coefficient to perform minute-level calculations of fault probability; a model update module, which uses incremental learning and federated learning frameworks to periodically optimize and update the prediction model using new operational data, fault cases, and inspection multimodal data; and an output and early warning module, which includes a display unit and an early warning unit. The display unit is used to display the equipment fault probability and influencing factors in real time, and the early warning unit is used to issue audible and visual alarms and remote warnings when the fault probability exceeds the dynamic early warning threshold.

[0013] Compared with the prior art, the present invention has at least the following beneficial effects: This invention fully considers the coupling effect between meteorological factors and the multi-physical fields of equipment. By establishing a coupling model, the fault probability prediction is made more consistent with the actual operating scenario, which effectively improves the prediction accuracy and provides a more reliable decision-making basis for substation operation and maintenance. This invention patent realizes dynamic updating of fault probability, which can respond in real time to changes in meteorological conditions and fluctuations in equipment physical field parameters, solves the lag problem of traditional static prediction methods, and meets the needs of real-time operation and maintenance. This invention patent sets up a dedicated data preprocessing module to perform noise reduction, normalization and spatiotemporal alignment on multi-source data, which improves the effectiveness and fusion of data and lays a good data foundation for subsequent coupled analysis and predictive calculations. The coupling model library of this invention patent supports the addition of custom models, which can adapt to the needs of different types of substations and different operating environments, and has strong flexibility and adaptability, with a wide range of applications; This invention patent, through accurate fault probability prediction and timely early warning functions, helps to reduce blind inspection work, lower operation and maintenance costs, and at the same time can avoid potential fault risks in advance, thereby improving the safety and reliability of substation operation. The output module of this invention patent can clearly display the probability of failure and key influencing factors, enabling maintenance personnel to quickly understand the equipment operating status and the source of failure risks, facilitating the development of targeted maintenance measures and improving maintenance efficiency. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a flowchart illustrating the substation dynamic fault probability prediction method based on meteorological-equipment multi-physics field coupling according to Embodiment 1 of the present invention. Figure 2 This is a schematic diagram of the substation dynamic fault probability prediction device based on meteorological-equipment multi-physics field coupling according to Embodiment 2 of the present invention. Figure 3 This is a schematic diagram of the substation dynamic fault probability prediction system based on meteorological-equipment multi-physics field coupling according to Embodiment 2 of the present invention. Detailed Implementation

[0016] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of the present invention. However, the present invention can be practiced in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of the present invention. Therefore, the present invention is not limited to the basic embodiments disclosed below.

[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0018] Example 1 Reference Figure 1 The substation dynamic fault probability prediction method based on meteorological-equipment multi-physics coupling of the present invention includes the following steps: Step S1: Multi-source data acquisition. Meteorological parameters around the substation are collected through the meteorological sensor group, including temperature, humidity, wind speed, precipitation, lightning intensity, air pressure, and snow thickness. Physical field parameters of key equipment in the substation are collected through the equipment sensor group, including temperature field, electric field, mechanical stress field, humidity field, partial discharge, insulation resistance, and dissolved gas content in oil.

[0019] In this step, the data acquisition frequency can be dynamically adjusted, including: meteorological data is collected once every 5 minutes under normal operating conditions, and equipment physical field data is collected once every 2 minutes; when a sudden change in meteorological parameters or an abnormality in equipment parameters is detected, the acquisition frequency is increased to once every 10 seconds. The meteorological sensor group includes multiple sensors for collecting multiple meteorological parameters, which are deployed in suitable areas around the substation; the equipment sensor group includes multiple sensors for collecting multiple physical field parameters, which are installed at characteristic locations of key equipment in the substation; each sensor transmits the collected meteorological parameter data and physical field parameter data to the data preprocessing module through the data transmission unit.

[0020] Step S2: Data preprocessing. The collected raw data is subjected to noise filtering, data normalization, and spatiotemporal alignment in sequence to obtain standardized data.

[0021] Specifically, noise filtering is used to denoise the original data and eliminate interference signals. It employs an adaptive filtering algorithm based on a sliding window, with the following formula: ,in, These are the filtered data points. w For the width of the sliding window, For adaptive weights, This represents the mean of the data within the window; data normalization is used to convert different types of parameters into the [0,1] interval, unifying the magnitude range. Linear normalization is used, and the formula is... in, The historical statistical extreme values ​​of the parameters are extracted from the substation operation and maintenance history database; spatiotemporal alignment is used to match the temporal and spatial dimensions of meteorological parameter data and equipment physical field parameter data. Time alignment uses linear interpolation to complete missing time-series data, as shown in the formula. To ensure that meteorological data and equipment data timestamps are synchronized; spatial alignment is based on calculating the influence weights using sensor spatial coordinates, the formula is as follows: This enables precise spatial matching between meteorological data and equipment physical field data.

[0022] Step S3: Coupling analysis, extract key features from standardized data, and output the comprehensive coupling influence coefficient of the response based on the coupling model library.

[0023] Specifically, key features such as parameter mutation points, time series trends, and extreme values ​​are extracted from standardized data. Core features strongly correlated with coupling effects are screened through mutual information entropy, and redundant information is eliminated. The coupling model library has built-in coupling models such as strong wind-mechanical stress, heavy rainfall-humidity field, and lightning-electric field. The model is automatically matched and adapted according to the substation type, equipment type, and operating scenario to analyze the coupling influence relationship between meteorological factors and equipment multi-physical fields, and output the comprehensive coupling influence coefficient.

[0024] The coupling model library contains multiple sets of coupling relationship models. It employs an improved Monte Carlo scene sampling method and constructs extreme disaster coupling equations based on high-resolution numerical weather prediction data and equipment multiphysics parameters. ,in, For dynamic weights ( (Dynamically adjusted based on the frequency of disaster occurrence) For meteorological intensity parameters, For the physical field parameters of the equipment, The disaster coupling function includes This is the wind-mechanical stress coupling function; This is the coupling function of heavy rainfall and humidity fields. The lightning-electric field coupling function is defined as , and a1, b1, etc., are equipment characteristic correlation coefficients, preset based on equipment type and material. This method can dynamically reflect the spatiotemporal evolution characteristics of coupling effects under extreme weather conditions, significantly improving model accuracy.

[0025] Step S4: Dynamic failure probability calculation. Based on the comprehensive coupling influence coefficient, the failure probability is calculated in minutes using a Bayesian updated dynamic linear model, a proportional risk model, or a hybrid model.

[0026] Improved Monte Carlo sampling is used to correct the coupling coefficient, and the number of samplings is adjusted. N ≥1000, through more than 1000 random scenario samplings, simulating uncertainties such as weather fluctuations and equipment degradation, the formula is: ( obey Generate a device-level fault probability spatiotemporal matrix .

[0027] The Yesse update dynamic linear model achieves minute-level fault probability estimation through optimal filtering of time-series data. The posterior probability formula is: After verification, the failure probability is calculated using the following formula: The proportional risk model integrates historical equipment lifespan data with real-time coupled stress; the instantaneous failure probability formula is as follows: ,in The baseline failure rate is the Weibull distribution. ß For the correlation coefficient of the running parameters,X t These are real-time operating parameters; the probability of failure within the [t,T] time window is... .

[0028] Step S5: Model update. Using an incremental learning and federated learning framework, the prediction model is regularly optimized and updated by utilizing new operational data, failure cases, and multimodal inspection data.

[0029] Incremental learning optimization includes employing mini-batch training and experience replay techniques to dynamically update model parameters based on newly collected operational data and failure cases. The formula is as follows: , η =0.01 is the learning rate. B For small batch sample sets; federated learning aggregation includes each substation edge node uploading its encrypted local model update to the central cloud platform, and the central nodes weighting and aggregating the data according to sample size to generate a global model, as shown in the formula. This achieves a collaborative optimization closed loop of "receiving-learning-updating-contributing".

[0030] Step S6: Result output and early warning. Real-time display of equipment failure probability and influencing factors. When the failure probability exceeds the dynamic early warning threshold, sound and light alarms and remote early warnings are issued.

[0031] It displays real-time information such as equipment failure probability curves and key influencing factors (e.g., dominant meteorological factors, abnormal physical field parameters); the warning threshold is dynamically set, using the formula: ,in Based on the threshold, This represents the equipment importance coefficient. This is the normalized coupling effect coefficient. When the failure probability exceeds... When the alarm is triggered, an audible and visual alarm will be activated, and a warning message and maintenance suggestions will be sent to the maintenance platform via remote communication.

[0032] Example 2 Reference Figure 2 The substation dynamic fault probability prediction device based on meteorological-equipment multi-physics coupling of the present invention includes: The multi-source data acquisition module includes a meteorological sensor group for collecting meteorological parameters around the substation and an equipment sensor group for collecting physical field parameters of key equipment in the substation; it also includes a data transmission unit to transmit the collected meteorological parameters and physical field parameters to the data preprocessing module. The data preprocessing module includes a noise filtering unit, a data normalization unit, and a spatiotemporal alignment unit, which are used to sequentially perform noise filtering, data normalization, and spatiotemporal alignment on the collected raw data to obtain standardized data. The coupling analysis module includes a feature extraction unit and a coupling model library. The feature extraction unit is used to extract key features from preprocessed standardized data. The coupling model library has multiple sets of coupling relationship models built in. According to the substation type and operation scenario, the matching model is called to analyze the coupling influence relationship between meteorological factors and equipment multi-physical fields, and output the comprehensive coupling influence coefficient. The dynamic failure probability module is used to perform minute-level calculations of failure probability based on the comprehensive coupling influence coefficient, using a Bayesian update dynamic linear model and a proportional risk model. The model update module is used to periodically optimize and update the prediction model using incremental learning and federated learning frameworks, leveraging new operational data, fault cases, and inspection multimodal data. The output and early warning module includes a display unit and an early warning unit. The display unit is used to display the probability of equipment failure and influencing factors in real time, and the early warning unit is used to issue audible and visual alarms and remote early warnings when the probability of failure exceeds the dynamic early warning threshold.

[0033] Reference Figure 3 The substation dynamic fault probability prediction system based on meteorological-equipment multi-physics coupling of the present invention includes a cloud platform, a federated learning coordinator, a global model distribution mechanism to edge nodes, a substation module, aggregated and encrypted model updates, and generation of a new generation of global models. In the substation module, the global model is first received from the central cloud platform as the initial basis for local learning. Subsequently, the model enters the local incremental learning engine, where it is continuously fine-tuned and optimized using multi-source data generated in real time at the substation through small-batch training and experience replay techniques. After this process, the system updates the local model to better adapt it to the current operating environment and risk characteristics of the substation. Next, the substation calculates the model update, that is, extracts the parameter changes generated by local training. Finally, the encrypted model update is uploaded to the center to participate in the federated aggregation of the global model, thus completing a complete local collaborative optimization closed loop of "receiving-learning-updating-contributing".

[0034] It should be noted that, depending on the implementation needs, the steps described in the embodiments of the present invention can be broken down into more steps, or two or more steps or parts of the steps can be combined into new steps to achieve the purpose of the embodiments of the present invention.

[0035] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A method for predicting the dynamic fault probability of substations based on meteorological-equipment multi-physics coupling, characterized in that, include: Multi-source data acquisition: Meteorological parameters around the substation are collected through a group of meteorological sensors, including temperature, humidity, wind speed, precipitation, lightning intensity, and air pressure. The physical field parameters of key equipment in the substation are collected through the equipment sensor group, including temperature field, electric field, mechanical stress field, humidity field, partial discharge quantity, and insulation resistance. Data preprocessing involves sequentially performing noise filtering, data normalization, and spatiotemporal alignment on the collected raw data to obtain standardized data. Coupling analysis extracts key features from standardized data and outputs the comprehensive coupling influence coefficient of the response based on the coupling model library; Dynamic failure probability calculation is based on the comprehensive coupling influence coefficient, and uses a Bayesian updated dynamic linear model and a proportional risk model to perform minute-level calculations of failure probability. Model updates employ an incremental learning and federated learning framework, utilizing new operational data, fault cases, and multimodal inspection data to periodically optimize and update the prediction model; Results output and early warning: Real-time display of equipment failure probability and influencing factors; When the failure probability exceeds the dynamic early warning threshold, sound and light alarms and remote early warning are issued.

2. The method for predicting the dynamic fault probability of a substation based on meteorological-equipment multi-physics coupling according to claim 1, characterized in that, The data acquisition frequency can be dynamically adjusted in the multi-source data acquisition process, including: meteorological data is acquired once every 5 minutes under normal operating conditions, and equipment physical field data is acquired once every 2 minutes; when meteorological parameters change suddenly or equipment parameters are abnormal, the acquisition frequency is increased to 10 seconds / time.

3. The method for predicting the dynamic fault probability of a substation based on meteorological-equipment multi-physics coupling according to claim 1, characterized in that, In the data preprocessing step, Noise filtering employs an adaptive filtering algorithm based on a sliding window, the formula of which is: in w For the width of the sliding window, w j For adaptive weights; Data normalization uses linear normalization, and the formula is: Time alignment is achieved using linear interpolation, and spatial alignment is based on the influence weights calculated using the sensor's spatial coordinates. The formulas are as follows: , .

4. The method for predicting the dynamic fault probability of a substation based on meteorological-equipment multi-physics coupling according to claim 1, characterized in that, In the coupling analysis step, the coupling model library contains multiple sets of coupling relationship models to construct the coupling effect equations for extreme disasters. ,in, For dynamic weights, For disaster coupling function, For meteorological intensity parameters, The calculation of the comprehensive coupling influence coefficient includes: calling the matching model according to the substation type and operating scenario, analyzing the coupling influence relationship between meteorological factors and the multiple physical fields of the equipment, and outputting the comprehensive coupling influence coefficient.

5. The method for predicting the dynamic fault probability of a substation based on meteorological-equipment multi-physics coupling according to claim 1, characterized in that, In the dynamic failure probability calculation, an improved Monte Carlo sampling method is used to correct the comprehensive coupling coefficient, and the number of samplings is adjusted. N ≥1000, generate the fault probability spatiotemporal matrix, the formula is as follows: .

6. The method for predicting the dynamic fault probability of a substation based on meteorological-equipment multi-physics coupling according to claim 1, characterized in that, In the dynamic failure probability calculation step After the Bayesian updated dynamic linear model passes the validation probability calculation, the failure probability is calculated using the following formula: ; The formula for the proportional risk model is: , ,in As the baseline failure rate, ß For the correlation coefficient of the running parameters, X t These are the parameters for real-time operation.

7. The method for predicting the dynamic fault probability of a substation based on meteorological-equipment multi-physics coupling according to claim 1, characterized in that, In the model update step, Incremental learning uses mini-batch parameter updates, as shown in the formula: ; Federated learning aggregates the global model using a weighted average of sample sizes, as shown in the formula: .

8. The method for predicting the dynamic fault probability of a substation based on meteorological-equipment multi-physics coupling according to claim 1, characterized in that, In the results output and early warning steps The warning threshold is set dynamically, and the formula is as follows: ,in Based on the threshold, This represents the equipment importance coefficient. This is the normalized coupling effect coefficient.

9. A substation dynamic fault probability prediction device based on meteorological-equipment multi-physics coupling, characterized in that, include: The multi-source data acquisition module includes a meteorological sensor group for collecting meteorological parameters around the substation and an equipment sensor group for collecting physical field parameters of key equipment in the substation. The data preprocessing module includes a noise filtering unit, a data normalization unit, and a spatiotemporal alignment unit, which are used to sequentially perform noise filtering, data normalization, and spatiotemporal alignment on the collected raw data to obtain standardized data. The coupling analysis module includes a feature extraction unit and a coupling model library. The feature extraction unit is used to extract key features from preprocessed standardized data. The coupling model library has multiple sets of coupling relationship models built in. According to the substation type and operation scenario, the matching model is called to analyze the coupling influence relationship between meteorological factors and equipment multi-physical fields, and output the comprehensive coupling influence coefficient. The dynamic failure probability module is used to perform minute-level calculations of failure probability based on the comprehensive coupling influence coefficient, using a Bayesian update dynamic linear model and a proportional risk model. The model update module is used to periodically optimize and update the prediction model using incremental learning and federated learning frameworks, leveraging new operational data, fault cases, and inspection multimodal data. The output and early warning module includes a display unit and an early warning unit. The display unit is used to display the probability of equipment failure and influencing factors in real time, and the early warning unit is used to issue audible and visual alarms and remote early warnings when the probability of failure exceeds the dynamic early warning threshold.

Images