Power fault detection and response method and system based on multi-source signal fusion

Abstract

The invention relates to the technical field of data processing of industrial internet, and provides a power failure detection and response method and system based on multi-source signal fusion. Acquiring working data of the power system through an intelligent sensor; preprocessing the working data to obtain a data tensor with unified time and space; performing multi-modal feature extraction on the data tensor to obtain a time-frequency feature corresponding to a time-frequency dimension, a temperature gradient field corresponding to a space dimension and a fractal dimension corresponding to a pulse dimension, and constructing a feature set based on the time-frequency feature, the temperature gradient field and the fractal dimension; features in the feature set are fused to generate a fused feature sequence, weighting processing based on an attention mechanism is carried out on the fused feature sequence, and fault probability distribution is generated; and generating a dynamic response strategy according to the fault probability distribution. Through multi-physical-quantity collaborative analysis, dynamic feature optimization and probabilistic decision, an intelligent electric power operation and maintenance system with adaptability, high precision and quick response is constructed.

Description

Technical Field

[0001] The present application relates to the field of data processing technology for the industrial Internet, and specifically to a method and system for detecting and responding to power faults based on multi-source signal fusion. Background Art

[0002] In industrial production applications, the safe and stable operation of power systems, as critical infrastructure, is directly related to social and economic activities. With the expansion of power grids and the high penetration of renewable energy, power equipment fault types are becoming increasingly complex, such as arc faults, insulation degradation, and localized overheating. Traditional detection methods based on single-source signals (such as current and voltage) face problems such as high missed detection rates, inaccurate positioning, and delayed response.

[0003] In existing technologies, data is collected through smart sensors on the Industrial Internet. Multi-source signal fusion technology is used to integrate multimodal data such as current, voltage, partial discharge, and infrared temperature to capture fault characteristics. However, existing methods have the problem of rigid feature fusion strategies and are unable to extract accurate features from the collected data, which in turn results in low accuracy in power fault detection. Summary of the Invention

[0004] The present application provides a method and system for power fault detection and response based on multi-source signal fusion, thereby solving the problem of low accuracy of power fault detection at least to a certain extent.

[0005] Other features and advantages of the present application will become apparent from the following detailed description, or may be learned in part by practice of the present application.

[0006] According to one aspect of the present application, a power fault detection and response method based on multi-source signal fusion is provided, including: acquiring operating data of the power system through intelligent sensors; preprocessing the operating data to obtain a data tensor unified in time and space; performing multimodal feature extraction on the data tensor to obtain time-frequency features corresponding to the time-frequency dimension, a temperature gradient field corresponding to the spatial dimension, and a fractal dimension corresponding to the pulse dimension, and constructing a feature set based on the time-frequency features, the temperature gradient field, and the fractal dimension; fusing the features in the feature set to generate a fused feature sequence, and performing weighted processing on the fused feature sequence based on an attention mechanism to generate a fault probability distribution; and generating a dynamic response strategy based on the fault probability distribution.

[0007] In the present application, based on the aforementioned scheme, the obtaining of the working data of the power system through intelligent sensors includes: collecting electrical signals through current and voltage sensors, obtaining the temperature of the power system equipment through infrared thermal imagers and generating temperature field distribution, and recording pulse waveforms through partial discharge sensors.

[0008] In the present application, based on the aforementioned scheme, the working data is preprocessed to obtain a data tensor that is unified in time and space, including: interpolation processing of multi-source working data, integrating the interpolation processed signals into a three-dimensional first tensor, the first tensor including information of type dimension, time dimension and space dimension; projecting the first tensor into a low-dimensional space to generate a second tensor; based on the information on the time dimension in the second tensor, dynamically aligning the second tensor to generate a data tensor.

[0009] In the present application, based on the aforementioned scheme, the multimodal feature extraction is performed on the data tensor to obtain the time-frequency features corresponding to the time-frequency dimension, the temperature gradient field corresponding to the spatial dimension, and the fractal dimension corresponding to the pulse dimension, including: extracting the time-frequency features of the current signal based on the data tensor, and the time-frequency features are used to determine whether there is an instantaneous arc fault in the power system; determining the temperature gradient field of the spatial temperature of the power system based on the data tensor, and the temperature gradient field is used to determine the fault hotspot of the insulation breakdown risk in the power system; determining the fractal dimension of the discharge pulse of the power system based on the data tensor, and the fractal dimension is used to evaluate the discharge mode of the power system.

[0010] In the present application, based on the aforementioned scheme, the features in the feature set are fused to generate a fused feature sequence, and the fused feature sequence is weighted based on the attention mechanism to generate a fault probability distribution, including: determining the information parameters corresponding to each feature based on the features in the feature set; fusing the features in the feature set according to the information parameters to generate a fused feature sequence; and generating a fault probability distribution based on the fused feature sequence.

[0011] In the present application, based on the aforementioned scheme, the fault probability distribution is generated according to the fused feature sequence, including: linearly transforming the fused feature sequence and the spatial position of the power system equipment to generate memory units of attention weights and spatial positions; performing weighted summation on the memory units of the spatial positions based on the attention weights to obtain comprehensive memory; inputting the comprehensive memory and the current input features into a preset gating unit to generate a hidden state representing historical memory; inputting the hidden state into a preset fully connected layer to output multiple mixed components, and determining the Gaussian probability density corresponding to the spatial position and the classification probability corresponding to the fault type based on the mixed components as the fault probability distribution.

[0012] In the present application, based on the aforementioned scheme, the generation of a dynamic response strategy according to the fault probability distribution includes: generating a dynamic response strategy according to the spatial position and probability distribution of the fault in the fault probability distribution through a preset fault strategy correspondence relationship; and sending the dynamic response strategy to the control terminal and the management terminal.

[0013] According to one aspect of the present application, a power fault detection and response system based on multi-source signal fusion is provided, comprising:

[0014] An acquisition unit, used for acquiring operating data of the power system through smart sensors;

[0015] A spatiotemporal unit, configured to preprocess the working data to obtain a data tensor unified in time and space;

[0016] an extraction unit, configured to perform multimodal feature extraction on the data tensor to obtain time-frequency features corresponding to the time-frequency dimension, a temperature gradient field corresponding to the spatial dimension, and a fractal dimension corresponding to the pulse dimension, and construct a feature set based on the time-frequency features, the temperature gradient field, and the fractal dimension;

[0017] a fusion unit, configured to fuse the features in the feature set to generate a fused feature sequence, and perform weighted processing on the fused feature sequence based on an attention mechanism to generate a fault probability distribution;

[0018] A strategy unit is used to generate a dynamic response strategy according to the fault probability distribution.

[0019] According to one aspect of the present application, a computer-readable medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the power fault detection and response method based on multi-source signal fusion as described in the above embodiment is implemented.

[0020] According to one aspect of the present application, an electronic device is provided, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the power fault detection and response method based on multi-source signal fusion as described in the above embodiments.

[0021] According to one aspect of the present application, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the power fault detection and response method based on multi-source signal fusion provided in the various optional implementations described above.

[0022] The technical solution of this application obtains the working data of the power system through intelligent sensors; preprocesses the working data to obtain a data tensor that is unified in time and space; performs multimodal feature extraction on the data tensor to obtain time-frequency features corresponding to the time-frequency dimension, the temperature gradient field corresponding to the space dimension, and the fractal dimension corresponding to the pulse dimension, and constructs a feature set based on the time-frequency features, the temperature gradient field, and the fractal dimension; fuses the features in the feature set to generate a fused feature sequence, and performs weighted processing on the fused feature sequence based on the attention mechanism to generate a fault probability distribution; and generates a dynamic response strategy based on the fault probability distribution. Through multi-source sensor fusion and unified time-space preprocessing, combined with multi-dimensional feature extraction such as time-frequency analysis, temperature gradient field, and fractal dimension of discharge, the attention mechanism is used to realize dynamic feature fusion and fault probability modeling, and finally generates an accurate response strategy. Through collaborative analysis of multiple physical quantities, dynamic feature optimization, and probabilistic decision-making, an intelligent power operation and maintenance system with adaptability, high precision, and rapid response is constructed.

[0023] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The accompanying drawings are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present application, and together with the specification, are used to explain the principles of the present application. Obviously, the drawings described below are only some embodiments of the present application, and those skilled in the art can derive other drawings based on these drawings without inventive effort.

[0025] Figure 1 The flowchart of a power fault detection and response method based on multi-source signal fusion in one embodiment of the present application is schematically shown.

[0026] Figure 2 The flowchart of generating a data tensor in one embodiment of the present application is schematically shown.

[0027] Figure 3 The following schematically illustrates a schematic diagram of a power fault detection and response system based on multi-source signal fusion in one embodiment of the present application.

[0028] Figure 4 A schematic diagram of the structure of a computer system suitable for implementing an electronic device according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0029] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this application will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art.

[0030] In addition, described feature, structure or characteristic can be combined in one or more embodiments in any suitable manner.In the following description, many specific details are provided so as to provide a full understanding of the embodiments of the present application. However, it will be appreciated by those skilled in the art that the technical scheme of the present application can be put into practice without one or more of the specific details, or other methods, components, devices, steps etc. can be adopted. In other cases, known methods, devices, implementations or operations are not shown or described in detail to avoid blurring the various aspects of the application.

[0031] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically separate entities. That is, these functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0032] The flowcharts shown in the accompanying drawings are for illustrative purposes only and do not necessarily include all contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may be decomposed, while others may be combined or partially combined. Therefore, the actual execution order may vary depending on the actual situation.

[0033] The implementation details of the technical solution of this application are described in detail below:

[0034] Figure 1 FIG2 shows a flow chart of a method for detecting and responding to power failures based on multi-source signal fusion according to an embodiment of the present application. Figure 1 As shown, the power fault detection and response method based on multi-source signal fusion includes at least steps S110 to S150, which are described in detail as follows:

[0035] In step S110 , operating data of the power system is acquired through smart sensors.

[0036] Intelligent sensors collect multi-source operating data from the power system in real time, including current, voltage, partial discharge pulse signals, and temperature field distribution. Optionally, in this embodiment, the intelligent sensors include high-frequency current transformers, infrared thermal imagers, and partial discharge detectors. Specifically, the current and voltage sensors capture electrical signals at a preset sampling rate; the infrared thermal imager acquires temperature information at preset time intervals and generates a device temperature field distribution; and the partial discharge sensor records the pulse waveform.

[0037] Optionally, working data is transmitted to edge computing nodes via communication protocols to ensure the timeliness and integrity of the original signal. At the same time, clocks between smart sensors are synchronized via industrial IoT protocols to ensure the timing consistency of the original data.

[0038] This process, through the parallel acquisition of heterogeneous data from multiple sources, covers multi-dimensional fault characteristics, including electrical, thermodynamic, and discharge characteristics, providing comprehensive input for subsequent fusion analysis. A high-precision synchronization mechanism avoids the signal phase misalignment caused by traditional asynchronous acquisition, particularly ensuring the integrity of high-frequency transient signals (such as arcs).

[0039] In step S120, the working data is preprocessed to obtain a data tensor that is unified in time and space.

[0040] The collected original multi-source data are subjected to spatiotemporal alignment processing, and dynamic time warping is used to compensate for the time offset of signals with different sampling rates. Single-point signals (such as current and voltage) are matched to the two-dimensional grid coordinates of the temperature field distribution through spatial interpolation. Subsequently, a three-dimensional data tensor is constructed, with dimensions corresponding to the signal type, unified time axis, and spatial grid, respectively. Missing data is filled by neighboring interpolation, and finally a spatiotemporally synchronized tensor structure is output, providing a standardized input format for subsequent feature fusion.

[0041] like Figure 2 As shown, in one embodiment of the present application, the working data is preprocessed to obtain a data tensor unified in time and space, including:

[0042] S210, performing interpolation processing on multi-source working data, integrating the interpolated signals into a three-dimensional first tensor, where the first tensor includes information on a type dimension, a time dimension, and a space dimension;

[0043] S220, projecting the first tensor into a low-dimensional space to generate a second tensor;

[0044] S230: Dynamically align the second tensor based on the information on the time dimension in the second tensor to generate a data tensor.

[0045] In one embodiment of the present application, interpolation processing is performed on working data from multiple sources. For example, spatial interpolation is performed on single-point signals (such as current and voltage) to map them to a two-dimensional grid consistent with the temperature field distribution. Linear interpolation is performed on low-sampling rate signals to align their time resolution with the current signal sampled at the microsecond level, so that signals with different sampling rates and spatial resolutions are unified into spatiotemporal grid data of the same format. Example: The temperature field distribution and microsecond current signal originally sampled at the second level are both presented at a sampling rate of 10kHz and a 100×100 grid after processing. For example, if the temperature field distribution is 100×100 pixels, the current signal needs to be interpolated to generate an equivalent value for each pixel.

[0046] The processed signals are integrated into a three-dimensional tensor structure as the first tensor, where the three dimensions include signal type dimension, time dimension, and space dimension. Specifically, the signal type dimension can include data corresponding to different signal types such as current, voltage, partial discharge, and temperature field distribution. The time dimension includes the length of the unified time series. The space dimension can include the pixel grid of the temperature field distribution or the data corresponding to the interpolated spatial position of other signals. For example, if there are 4 types of signals, 1000 time points, and a 100×100 spatial grid, the tensor size is 4×1000×100004×1000×10000.

[0047] The first tensor is projected into a low-dimensional space to compress the data and extract key features, generating a second tensor that is projected into a low-dimensional space to remove redundant information and retain key features. The second tensor retains the main spatiotemporal features of the signal. For example, the original tensor is reduced from 4×1000×10000 to 2×200×500, removing noise and redundancy.

[0048] To address timing offset issues between different signals, such as temperature changes lagging behind current changes, the second tensor is dynamically aligned based on the time dimension information in the second tensor to generate a data tensor. Specifically, the time factor matrix of the core tensor is calculated and the time evolution patterns of different signals are analyzed. The objective function is determined through dynamic programming as:

[0049]

[0050] Among them, S i 、S j represents the data segment of the signal to be aligned, i and j represent the generation time of the data judgment, λ represents the regularization coefficient, and W represents the nonlinear mapping relationship of the preset time axis.

[0051] Based on the objective function, the optimal time alignment path is found to align key events (such as the fault onset) of different signals on the time axis. The original signals are then resampled according to the alignment path to generate synchronized data on a unified time axis as a data tensor. The data tensor includes data features in a unified time dimension and a unified spatial grid dimension. By correcting for the spatial position deviation of the sensors, the temperature field distribution and the current signal are strictly aligned in the time and space dimensions, providing high-precision input for subsequent feature extraction.

[0052] The above process unifies the sampling rates of different sensors through interpolation. For example, the high-frequency sampling of the current sensor is aligned with the time axis of the low-frequency sampling of the infrared thermal imager. This constructs a three-dimensional tensor encompassing type, time, and space, addressing the heterogeneity of multi-source data. Mapping heterogeneous signals onto a standardized grid provides a consistent input format for subsequent multi-source fusion. Joint dimensionality reduction across multiple dimensions preserves the spatial and temporal correlation of the signals, avoiding the information loss associated with single-dimensional processing.

[0053] In step S130, multimodal feature extraction is performed on the data tensor to obtain time-frequency features corresponding to the time-frequency dimension, the temperature gradient field corresponding to the spatial dimension, and the fractal dimension corresponding to the pulse dimension, and a feature set is constructed based on the time-frequency features, the temperature gradient field, and the fractal dimension.

[0054] Multidimensional feature extraction is performed on the spatiotemporally aligned data tensor. For current and voltage signals, nonlinear time-frequency features are extracted through time-frequency analysis. The spatial temperature gradient field is calculated for temperature distribution data to identify areas of temperature anomalies. For partial discharge signals, the fractal dimension of the pulse waveform is calculated using a box counting method. The three types of features—time-frequency features, temperature gradient field, and fractal dimension—are then normalized and combined by feature type to construct a multidimensional feature vector set that incorporates spatiotemporal characteristics and signal complexity, providing standardized feature input for subsequent intelligent diagnosis.

[0055] In one embodiment of the present application, multimodal feature extraction is performed on the data tensor to obtain time-frequency features corresponding to the time-frequency dimension, a temperature gradient field corresponding to the spatial dimension, and a fractal dimension corresponding to the pulse dimension, including:

[0056] extracting time-frequency features of a current signal based on the data tensor, wherein the time-frequency features are used to determine whether a transient arc fault exists in the power system;

[0057] Determining a temperature gradient field of a spatial temperature of the power system based on the data tensor, wherein the temperature gradient field is used to determine a fault hotspot with an insulation breakdown risk in the power system;

[0058] A fractal dimension of a discharge pulse of the power system is determined based on the data tensor, and the fractal dimension is used to evaluate a discharge pattern of the power system.

[0059] In one embodiment of the present application, the time-frequency characteristics of the current signal are extracted based on the data tensor. Specifically, the time series is extracted from the data tensor obtained after alignment, and the time series is embedded in the phase space to construct the system differential equation. Then, based on the system differential equation, the long-term average value of the adjacent orbital exponential divergence rate is calculated to obtain the time-frequency characteristics C LE for:

[0060]

[0061] Among them, f′(x v ) indicates that the system is in state x v The derivative of the system differential equation at ; v represents the state type; t represents the working time of the power system.

[0062] In practical applications, the current in a power system exhibits periodicity during normal operation, while fault currents (such as arcs and short circuits) often induce nonlinear oscillations. This chaotic behavior can be quantified using system differential equations, manifesting as exponential divergence between adjacent orbits in phase space. By determining the system differential equation for the current signal, we can derive its time-frequency characteristics, which can be used to determine whether the power system is experiencing a transient arc fault—that is, whether it is in a chaotic state.

[0063] Based on the data tensor, the temperature gradient field of the spatial temperature of the power system is determined. Specifically, the spatial distribution is extracted from the temperature field distribution T in the aligned data tensor, and the gradient of each spatial point projected along the normal vector direction of the power system equipment surface is calculated; the rate of change of temperature over time is superimposed to generate a dynamic temperature gradient field G T for:

[0064]

[0065] Where T represents the temperature field distribution, is the spatial gradient of the temperature field distribution, which represents the rate of change of temperature in two-dimensional space; n represents the surface normal vector of the power system equipment, which projects the gradient to the direction perpendicular to the surface of the power system equipment. It indicates the rate of change of temperature over time, reflecting the dynamic accumulation or dissipation of heat.

[0066] The temperature gradient field calculated in this embodiment is used to identify fault hotspots that indicate insulation breakdown risk in the power system. Combining the spatial gradient and temporal rate of change of the temperature field distribution, the heat source is located and the heat diffusion process is tracked. If the calculated local temperature gradient increases abnormally, it indicates a fault in the power system equipment, such as poor contact or insulation breakdown, leading to localized Joule heating. The maximum value of the temperature gradient field directly corresponds to the fault hotspot.

[0067] The fractal dimension of the discharge pulse of the power system is determined based on the data tensor. Specifically, the pulse waveform is extracted from the aligned partial discharge signal; the box counting method is used to count the number of boxes with different side lengths required to cover the pulse waveform; and the fractal dimension D is calculated by double logarithmic linear regression as follows:

[0068]

[0069] Where ε is the side length of the square box used to cover the pulse waveform, and N(ε) is the number of boxes with a side length of ε required to cover the entire waveform.

[0070] In practical applications, when insulation materials degrade, the partial discharge path becomes tortuous, and the pulse waveform exhibits more irregular fluctuations, which increases the fractal dimension. The fractal dimension calculated in this embodiment is used to evaluate the discharge pattern of the power system, quantifying the geometric complexity of the discharge pulse. A larger fractal dimension indicates a more complex discharge waveform and significantly increased insulation degradation.

[0071] This process accurately locates instantaneous arc faults by extracting the time-frequency characteristics of the current signal, distinguishing normal operation from fault pulses. By calculating the spatial gradient of the temperature field distribution, areas at risk of insulation breakdown are identified, providing early warning of poor contact or localized overheating. The fractal dimension of the discharge pulse is calculated using the box counting method to quantify the complexity of the discharge pattern. An increase in the fractal dimension indicates the progression of partial discharge from benign to malignant, providing a basis for predicting fault evolution.

[0072] In step S140, the features in the feature set are fused to generate a fused feature sequence, and the fused feature sequence is weighted based on the attention mechanism to generate a fault probability distribution.

[0073] In one embodiment of the present application, the time-frequency features, temperature gradient field and fractal dimension in the feature set are adaptively entropy-weighted and fused, the weights of each feature are dynamically allocated and cross-terms are introduced to enhance nonlinear correlations, thereby generating a fused feature sequence; the sequence is then input into a spatiotemporal attention network, and the historical time series features and spatial position encodings are dynamically weighted through the attention mechanism to focus on key fault signs, and finally the hybrid density network jointly outputs the probability distribution of the fault type, three-dimensional position coordinates and their confidence, thereby achieving millisecond-level precise diagnosis and uncertainty quantification.

[0074] In one embodiment of the present application, the features in the feature set are fused to generate a fused feature sequence, and the fused feature sequence is weighted based on an attention mechanism to generate a fault probability distribution, including:

[0075] Determining information parameters corresponding to each feature based on the features in the feature set;

[0076] fusing the features in the feature set according to the information parameters to generate a fused feature sequence;

[0077] A fault probability distribution is generated according to the fused feature sequence.

[0078] In one embodiment of the present application, after generating a multimodal feature set, the feature set may include information such as time-frequency features, temperature gradient field principal components, and discharge fractal dimensions. Each feature vector is binned and counted. For example, the time-frequency feature C LE The value of is divided into 20 data intervals, and the probability p of each data interval is counted. k .

[0079] Based on the features in the feature set, determine the information parameter H corresponding to each feature i for:

[0080]

[0081] Where i represents the identifier of the feature in the feature set, k represents the identifier of the data interval, and K represents the number of preset data intervals. Represents the probability value of the i-th feature in the k-th data interval.

[0082] By quantifying the information validity of each feature, information parameters are generated to measure the uncertainty of the feature. For example, if the probability distribution of a feature is concentrated in a few data intervals, such as the time-frequency feature C LE When the fault occurs, the information parameter H is stable at around 0.5. i Low (close to 0); if the time-frequency feature C LE The more dispersed the distribution is, the higher the entropy value is (close to 1). The lower the information parameter, the more concentrated the features are, such as the sudden change signal during a fault, which is more important for fault detection; the higher the information parameter, the more dispersed the features are, such as background noise, which has a lower contribution to fault detection.

[0083] After the information parameters are determined, the characteristic parameters ω of the features in the same dimension are determined according to the information parameters. i for:

[0084]

[0085] Among them, α represents the preset sensitivity parameter, j represents the identifier of the feature involved in the fusion, and N represents the total number of features involved in the fusion. For example, if the three features of current, temperature, and discharge are fused, N is 3.

[0086] After determining the characteristic parameters, based on the characteristic parameters {ω i} for the feature set {F iSpecifically, each feature vector is multiplied by a weight and then added together to obtain a fused feature sequence, and then a fault probability distribution is generated based on the fused feature sequence.

[0087] In one embodiment of the present application, generating a fault probability distribution according to the fused feature sequence includes:

[0088] Performing a linear transformation on the fused feature sequence and the spatial position of the power system equipment to generate a memory unit of attention weight and spatial position;

[0089] Performing weighted summation on the memory units of the spatial positions based on the attention weights to obtain a comprehensive memory;

[0090] Inputting the comprehensive memory and current input features into a preset gating unit to generate a hidden state representing historical memory;

[0091] The hidden state is input into a preset fully connected layer, and multiple mixed components are output. The Gaussian probability density corresponding to the spatial position and the classification probability corresponding to the fault type are determined based on the mixed components as the fault probability distribution.

[0092] In one embodiment of the present application, a fused multi-modal fusion feature sequence is obtained, which includes current chaos, temperature gradient, discharge waveform complexity, etc., and at the same time, the encoded information of the spatial position of the power system equipment, such as the coordinates of the cable joint, is obtained. The fused feature sequence and the spatial position of the power system equipment are linearly transformed, and the importance score of each position is dynamically calculated. In this way, the model automatically pays attention to the spatiotemporal positions related to the current fault. Positions with high importance will receive higher weights, such as local hot spots or discharge areas, and attention will be focused on the spatial area near the discharge point.

[0093] Combined with attention weights, historical memory units are weighted and fused to generate a comprehensive memory. This comprehensive memory and the current input features are fed into a preset gating unit, which outputs a hidden state representing the historical memory. By fusing memories from different locations using attention weights, key fault modes, such as the temperature accumulation process during insulation breakdown, are preserved. In the event of a sudden fault, real-time features guide the update of the hidden state, improving response speed. For example, based on the current input features, the hidden state is updated within 10ms after an arc fault occurs, reducing response latency to less than 50ms.

[0094] For example, when an abnormal temperature is detected at a cable joint, the model automatically focuses on the historical data and current input features at that location. If the characteristics of an arc fault within the past 10 minutes are assigned a high weight, the hidden states representing these historical memories will be retained. Current input features (such as sudden current chaos) are directly input into the memory unit to generate hidden states representing historical memories, balancing the influence of historical information with real-time data.

[0095] The hidden state is input into a preset fully connected layer, which outputs multiple mixed components. These mixed components may include the weights corresponding to each type of fault, the mean corresponding to the fault location, and the probability distribution corresponding to each type of fault. Based on these mixed components, the Gaussian probability density of each spatial location and the classification probability of the fault type are determined as the fault probability distribution. For example, if the output has an 85% probability of a certain type of fault, the corresponding three-dimensional spatial position coordinates are (1.2, 3.4, 0.5) with a confidence interval of ±0.2 meters. By simultaneously constructing fault type identification based on discrete classification and fault location detection based on continuous regression, the limitations of traditional separation processing methods are overcome.

[0096] The above process adaptively adjusts feature weights through an attention mechanism. For example, in arc fault scenarios, the weight of time-frequency features is increased; in insulation aging scenarios, the weight of temperature gradient and fractal dimension is increased to suppress irrelevant feature noise. A gated recurrent unit (GRU) and spatial position memory are combined to establish a fault evolution history model. The hidden state records the historical state of power system equipment, such as the cumulative effects of long-term overheating. The fully connected layer outputs a mixed Gaussian probability density to achieve a joint probability distribution of fault location and type.

[0097] In step S150, a dynamic response strategy is generated according to the fault probability distribution.

[0098] In one embodiment of the present application, after generating the fault probability distribution, a dynamic response strategy is generated based on the spatial location and probability distribution of the fault in the fault probability distribution and through a preset fault strategy correspondence, and the dynamic response strategy is sent to the control terminal and the management terminal.

[0099] In one embodiment of the present application, the fault probability distribution output by the spatiotemporal attention detector is first analyzed to extract the fault type, spatial location, and confidence level. A response strategy is then dynamically matched based on a pre-defined fault strategy mapping, such as a knowledge base or rule engine.

[0100] For example, if it's an arc fault and the corresponding failure probability is greater than 90%, the "shut off the corresponding circuit breaker and initiate nitrogen spray fire extinguishing" command is triggered. If the spatial position covariance is large, indicating uncertainty in positioning, a "manual review" request is added. Ultimately, by invoking a message queue or data processing interface, the policy instructions are distributed to the control and management terminals, and the relevant information is displayed graphically in the terminal interface, achieving the dual guarantee of automated response and manual monitoring.

[0101] This process significantly improves response speed and accuracy through real-time data-driven decision-making. Dynamic policy matching supports multi-level responses, such as early warning, isolation, and fire extinguishing. Probabilistic outputs are combined to implement risk-tiered processing, avoiding over- or under-response. The control terminal executes millisecond-level actions, such as opening a circuit breaker, while the management terminal simultaneously receives visual alarms and fault analysis reports, creating a closed-loop management system. Field tests have shown that compared to traditional fixed-threshold strategies, this reduces the error rate by 35% and shortens fault isolation time to less than 200ms. Flexible collaboration with human intervention is also supported.

[0102] The technical solution of this application obtains the working data of the power system through intelligent sensors; preprocesses the working data to obtain a data tensor that is unified in time and space; performs multimodal feature extraction on the data tensor to obtain time-frequency features corresponding to the time-frequency dimension, the temperature gradient field corresponding to the space dimension, and the fractal dimension corresponding to the pulse dimension, and constructs a feature set based on the time-frequency features, the temperature gradient field, and the fractal dimension; fuses the features in the feature set to generate a fused feature sequence, and performs weighted processing on the fused feature sequence based on the attention mechanism to generate a fault probability distribution; and generates a dynamic response strategy based on the fault probability distribution. Through multi-source sensor fusion and unified time-space preprocessing, combined with multi-dimensional feature extraction such as time-frequency analysis, temperature gradient field, and fractal dimension of discharge, the attention mechanism is used to realize dynamic feature fusion and fault probability modeling, and finally generates an accurate response strategy. Through collaborative analysis of multiple physical quantities, dynamic feature optimization, and probabilistic decision-making, an intelligent power operation and maintenance system with adaptability, high precision, and rapid response is constructed.

[0103] The following describes an embodiment of the device of the present application, which can be used to execute the method for detecting and responding to power faults based on multi-source signal fusion in the above-mentioned embodiment of the present application. It is understood that the device can be a computer program (including program code) running on a computer device, for example, the device is an application software; the device can be used to execute the corresponding steps in the method provided in the embodiment of the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the embodiment of the method for detecting and responding to power faults based on multi-source signal fusion in the above-mentioned embodiment of the present application.

[0104] Figure 3A block diagram of a power fault detection and response system based on multi-source signal fusion according to an embodiment of the present application is shown.

[0105] Reference Figure 3 As shown, according to an embodiment of the present application, a power fault detection and response system based on multi-source signal fusion includes:

[0106] An acquisition unit 310 is configured to acquire operating data of the power system through an intelligent sensor;

[0107] A spatiotemporal unit 320 is configured to preprocess the working data to obtain a data tensor that is unified in time and space;

[0108] An extraction unit 330 is configured to perform multimodal feature extraction on the data tensor to obtain time-frequency features corresponding to the time-frequency dimension, a temperature gradient field corresponding to the spatial dimension, and a fractal dimension corresponding to the pulse dimension, and construct a feature set based on the time-frequency features, the temperature gradient field, and the fractal dimension;

[0109] A fusion unit 340 is configured to fuse the features in the feature set to generate a fused feature sequence, and perform weighted processing on the fused feature sequence based on an attention mechanism to generate a fault probability distribution;

[0110] The strategy unit 350 is configured to generate a dynamic response strategy according to the fault probability distribution.

[0111] In the present application, based on the aforementioned scheme, the obtaining of the working data of the power system through intelligent sensors includes: collecting electrical signals through current and voltage sensors, obtaining the temperature of the power system equipment through infrared thermal imagers and generating temperature field distribution, and recording pulse waveforms through partial discharge sensors.

[0112] In the present application, based on the aforementioned scheme, the working data is preprocessed to obtain a data tensor that is unified in time and space, including: interpolation processing of multi-source working data, integrating the interpolation processed signals into a three-dimensional first tensor, the first tensor including information of type dimension, time dimension and space dimension; projecting the first tensor into a low-dimensional space to generate a second tensor; based on the information on the time dimension in the second tensor, dynamically aligning the second tensor to generate a data tensor.

[0113] In the present application, based on the aforementioned scheme, the multimodal feature extraction is performed on the data tensor to obtain the time-frequency features corresponding to the time-frequency dimension, the temperature gradient field corresponding to the spatial dimension, and the fractal dimension corresponding to the pulse dimension, including: extracting the time-frequency features of the current signal based on the data tensor, and the time-frequency features are used to determine whether there is an instantaneous arc fault in the power system; determining the temperature gradient field of the spatial temperature of the power system based on the data tensor, and the temperature gradient field is used to determine the fault hotspot of the insulation breakdown risk in the power system; determining the fractal dimension of the discharge pulse of the power system based on the data tensor, and the fractal dimension is used to evaluate the discharge mode of the power system.

[0114] In the present application, based on the aforementioned scheme, the features in the feature set are fused to generate a fused feature sequence, and the fused feature sequence is weighted based on the attention mechanism to generate a fault probability distribution, including: determining the information parameters corresponding to each feature based on the features in the feature set; fusing the features in the feature set according to the information parameters to generate a fused feature sequence; and generating a fault probability distribution based on the fused feature sequence.

[0115] In the present application, based on the aforementioned scheme, the fault probability distribution is generated according to the fused feature sequence, including: linearly transforming the fused feature sequence and the spatial position of the power system equipment to generate memory units of attention weights and spatial positions; performing weighted summation on the memory units of the spatial positions based on the attention weights to obtain comprehensive memory; inputting the comprehensive memory and the current input features into a preset gating unit to generate a hidden state representing historical memory; inputting the hidden state into a preset fully connected layer to output multiple mixed components, and determining the Gaussian probability density corresponding to the spatial position and the classification probability corresponding to the fault type based on the mixed components as the fault probability distribution.

[0116] In the present application, based on the aforementioned scheme, the generation of a dynamic response strategy according to the fault probability distribution includes: generating a dynamic response strategy according to the spatial position and probability distribution of the fault in the fault probability distribution through a preset fault strategy correspondence relationship; and sending the dynamic response strategy to the control terminal and the management terminal.

[0117] The technical solution of this application obtains the working data of the power system through intelligent sensors; preprocesses the working data to obtain a data tensor that is unified in time and space; performs multimodal feature extraction on the data tensor to obtain time-frequency features corresponding to the time-frequency dimension, the temperature gradient field corresponding to the space dimension, and the fractal dimension corresponding to the pulse dimension, and constructs a feature set based on the time-frequency features, the temperature gradient field, and the fractal dimension; fuses the features in the feature set to generate a fused feature sequence, and performs weighted processing on the fused feature sequence based on the attention mechanism to generate a fault probability distribution; and generates a dynamic response strategy based on the fault probability distribution. Through multi-source sensor fusion and unified time-space preprocessing, combined with multi-dimensional feature extraction such as time-frequency analysis, temperature gradient field, and fractal dimension of discharge, the attention mechanism is used to realize dynamic feature fusion and fault probability modeling, and finally generates an accurate response strategy. Through collaborative analysis of multiple physical quantities, dynamic feature optimization, and probabilistic decision-making, an intelligent power operation and maintenance system with adaptability, high precision, and rapid response is constructed.

[0118] Figure 4 A schematic diagram of the structure of a computer system suitable for implementing an electronic device according to an embodiment of the present application is shown.

[0119] It should be noted that the computer system of the electronic device in this embodiment is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0120] In this embodiment, the computer system includes a central processing unit (CPU) 401, which can execute various appropriate actions and processes based on programs stored in a read-only memory (ROM) 402 or programs loaded from a storage unit 408 into a random access memory (RAM) 403. For example, this system can execute the power fault detection and response method based on multi-source signal fusion described in the above embodiments. The RAM 403 also stores various programs and data required for system operation. The CPU 401, ROM 402, and RAM 403 are interconnected via a bus 404. An input / output interface 405 is also connected to the bus 404.

[0121] The following components are connected to the input / output interface 405: an input section 406 including a keyboard, a mouse, and the like; an output section 407 including devices such as a cathode ray tube (CRT), a liquid crystal display (LCD), and a speaker; a storage section 408 including a hard disk and the like; and a communication section 409 including a network interface card such as a LAN (Local Area Network) card or a modem. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the input / output interface 405 as needed. Removable media 411, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, and the like, is installed in the drive 410 as needed so that computer programs read therefrom can be installed into the storage section 408 as needed.

[0122] In particular, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product that includes a computer program carried on a computer-readable medium, the computer program including a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 409 and / or installed from a removable medium 411. When the computer program is executed by the central processing unit 401, the various functions defined in the system of the present application are performed.

[0123] It should be noted that the computer-readable medium shown in the embodiments of the present application can be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media can include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, which carries a computer-readable computer program. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. A computer program embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wired, or any suitable combination thereof.

[0124] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. Among them, each box in the flowchart or block diagram can represent a module, program segment, or part of the code, and the above-mentioned module, program segment, or part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0125] The units involved in the embodiments described in this application may be implemented by software or hardware, and the units described may also be set in a processor. In some cases, the names of these units do not constitute limitations on the units themselves.

[0126] According to one aspect of the present application, a computer program product or computer program is provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the various optional implementations described above.

[0127] As another aspect, the present application also provides a computer-readable medium, which may be included in the electronic device described in the above embodiments, or may exist independently without being incorporated into the electronic device. The computer-readable medium carries one or more programs, and when the one or more programs are executed by the electronic device, the electronic device implements the power fault detection and response method based on multi-source signal fusion described in the above embodiments.

[0128] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiment of the application, the features and functions of two or more modules or units described above can be concretized in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into multiple modules or units to be concretized.

[0129] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solution according to the embodiments of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the embodiments of the present application.

[0130] Those skilled in the art will readily conceive of other embodiments of the present application after considering the specification and practicing the embodiments disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present application that follow the general principles of this application and include common knowledge or customary techniques in the art that are not disclosed herein.

[0131] It should be understood that the present application is not limited to the exact structures described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A power fault detection and response method based on multi-source signal fusion, characterized in that: include: Obtain power system operating data through smart sensors; Preprocessing the working data to obtain a data tensor that is unified in time and space; Performing multimodal feature extraction on the data tensor to obtain time-frequency features corresponding to the time-frequency dimension, a temperature gradient field corresponding to the spatial dimension, and a fractal dimension corresponding to the pulse dimension, and constructing a feature set based on the time-frequency features, the temperature gradient field, and the fractal dimension; Fusing the features in the feature set to generate a fused feature sequence, and performing weighted processing on the fused feature sequence based on an attention mechanism to generate a fault probability distribution; A dynamic response strategy is generated according to the fault probability distribution.

2. The power fault detection and response method based on multi-source signal fusion according to claim 1 is characterized in that: The method of obtaining the operating data of the power system through the intelligent sensor includes: Electrical signals are collected through current and voltage sensors, the temperature of power system equipment is obtained through infrared thermal imagers and temperature field distribution is generated, and pulse waveforms are recorded through partial discharge sensors.

3. The power fault detection and response method based on multi-source signal fusion according to claim 1 is characterized in that: The preprocessing of the working data to obtain a data tensor unified in time and space includes: Performing interpolation processing on multi-source working data, and integrating the interpolated signals into a three-dimensional first tensor, where the first tensor includes information on a type dimension, a time dimension, and a space dimension; Projecting the first tensor into a low-dimensional space to generate a second tensor; Based on the information on the time dimension in the second tensor, the second tensor is dynamically aligned to generate a data tensor.

4. The power fault detection and response method based on multi-source signal fusion according to claim 1 is characterized in that: The multimodal feature extraction is performed on the data tensor to obtain the time-frequency feature corresponding to the time-frequency dimension, the temperature gradient field corresponding to the spatial dimension, and the fractal dimension corresponding to the pulse dimension, including: extracting time-frequency features of a current signal based on the data tensor, wherein the time-frequency features are used to determine whether a transient arc fault exists in the power system; Determining a temperature gradient field of a spatial temperature of the power system based on the data tensor, wherein the temperature gradient field is used to determine a fault hotspot with an insulation breakdown risk in the power system; A fractal dimension of a discharge pulse of the power system is determined based on the data tensor, and the fractal dimension is used to evaluate a discharge pattern of the power system.

5. The power fault detection and response method based on multi-source signal fusion according to claim 1 is characterized in that: The fusing of the features in the feature set to generate a fused feature sequence, and performing weighted processing on the fused feature sequence based on an attention mechanism to generate a fault probability distribution includes: Determining information parameters corresponding to each feature based on the features in the feature set; fusing the features in the feature set according to the information parameters to generate a fused feature sequence; A fault probability distribution is generated according to the fused feature sequence.

6. The power fault detection and response method based on multi-source signal fusion according to claim 5 is characterized in that: Generating a fault probability distribution according to the fused feature sequence includes: Performing a linear transformation on the fused feature sequence and the spatial position of the power system equipment to generate a memory unit of attention weight and spatial position; Performing weighted summation on the memory units of the spatial positions based on the attention weights to obtain a comprehensive memory; Inputting the comprehensive memory and current input features into a preset gating unit to generate a hidden state representing historical memory; The hidden state is input into a preset fully connected layer, and multiple mixed components are output. The Gaussian probability density corresponding to the spatial position and the classification probability corresponding to the fault type are determined based on the mixed components as the fault probability distribution.

7. The power fault detection and response method based on multi-source signal fusion according to claim 1, characterized in that: Generating a dynamic response strategy according to the fault probability distribution includes: Generate a dynamic response strategy based on the spatial location and probability distribution of the fault in the fault probability distribution and a preset fault strategy correspondence relationship; The dynamic response strategy is sent to the control terminal and the management terminal.

8. A power fault detection and response system based on multi-source signal fusion, characterized in that: include: An acquisition unit, used for acquiring operating data of the power system through smart sensors; A spatiotemporal unit, configured to preprocess the working data to obtain a data tensor unified in time and space; an extraction unit, configured to perform multimodal feature extraction on the data tensor to obtain time-frequency features corresponding to the time-frequency dimension, a temperature gradient field corresponding to the spatial dimension, and a fractal dimension corresponding to the pulse dimension, and construct a feature set based on the time-frequency features, the temperature gradient field, and the fractal dimension; a fusion unit, configured to fuse the features in the feature set to generate a fused feature sequence, and perform weighted processing on the fused feature sequence based on an attention mechanism to generate a fault probability distribution; A strategy unit is used to generate a dynamic response strategy according to the fault probability distribution.

9. The power fault detection and response system based on multi-source signal fusion according to claim 8, characterized in that: The method of obtaining the operating data of the power system through the intelligent sensor includes: Electrical signals are collected through current and voltage sensors, the temperature of power system equipment is obtained through infrared thermal imagers and temperature field distribution is generated, and pulse waveforms are recorded through partial discharge sensors.

10. The power fault detection and response system based on multi-source signal fusion according to claim 8, characterized in that: The preprocessing of the working data to obtain a data tensor unified in time and space includes: Performing interpolation processing on multi-source working data, and integrating the interpolated signals into a three-dimensional first tensor, where the first tensor includes information on a type dimension, a time dimension, and a space dimension; Projecting the first tensor into a low-dimensional space to generate a second tensor; Based on the information on the time dimension in the second tensor, the second tensor is dynamically aligned to generate a data tensor.

Images