Solid-state power controller fault diagnosis system and method based on data analysis

Abstract

The present application provides a solid-state power controller fault diagnosis system and method based on data analysis, which relates to the field of intelligent equipment management. It uses artificial intelligence-based data analysis technology to organize and time-series encode the load electrical characteristic data of the solid-state power controller, and then responds to the characteristics after the time-series encoding, and embeds the type label of the load to encode it, so as to intelligently predict the production impulse voltage value and generate corresponding fault warning prompts based on the fast semantic matching query representation information between the timing response association characteristics of each load electrical characteristic and the embedded characteristics of the load type. In this way, it can adapt to different types of load changes more flexibly, and can more accurately reflect the effect of the load type on the load electrical characteristics, thereby helping to improve the accuracy of fault prediction.

Description

Technical Field

[0001] The present application relates to the field of intelligent equipment management, and more specifically, to a solid-state power controller fault diagnosis system and method based on data analysis. Background Art

[0002] Solid-state power controllers (SSPCs) are widely used in aerospace, industrial automation, power distribution systems and other fields to control and protect power loads. Compared with traditional mechanical relays or circuit breakers, SSPCs have faster response speeds, higher reliability and longer service life. As the complexity of power systems increases, the risk of failure faced by SSPCs also increases accordingly, which may lead to serious economic losses and safety hazards.

[0003] Patent CN118277852A discloses a fault diagnosis method and device for a solid-state power controller, which receives data from current, voltage, temperature, fault status, load condition and magnetic field sensors, and uses them to train a CNN-LSTM model after filtering and noise reduction, effective data extraction and sliding window sampling preprocessing to achieve fault classification, and combines the current and voltage characteristics analysis of resistive, capacitive and inductive loads to predict faults in advance.

[0004] In this patent, the fault result is obtained by combining the current, voltage and load type through the trained CNN-LSTM model. Although the CNN-LSTM model is used for fault diagnosis in the patent, it obtains and analyzes current and voltage data based on the conditions of the determined load type. Although this method is simple, it has some limitations. Specifically, the method assumes that the load type remains unchanged throughout the monitoring process. However, in actual applications, the load type may change dynamically, such as switching from a resistive load to an inductive or capacitive load. If the load type changes and the system fails to update its status in time, the fault diagnosis result may be inaccurate. In addition, the method fails to deeply explore the intrinsic connection and influence between the load type and the load electrical characteristics. Due to the lack of in-depth analysis of the load characteristics, the model may not be able to fully capture the complex behaviors and potential failure modes under various load conditions, which may affect the accuracy and reliability of fault diagnosis when dealing with diverse actual application scenarios.

[0005] Therefore, an optimized solid-state power controller fault diagnosis scheme is desired. Summary of the invention

[0006] In order to solve the above technical problems, the present application is proposed. The embodiments of the present application provide a solid-state power controller fault diagnosis system and method based on data analysis.

[0007] According to one aspect of the present application, a solid-state power controller fault diagnosis system based on data analysis is provided, which includes:

[0008] A load electrical characteristic data acquisition module, used to acquire a time set of load electrical characteristic data of the solid-state power controller, wherein the load electrical characteristic data includes a load current value and a load voltage value;

[0009] A load electrical characteristic data encoding module, used for grouping the time set of the load electrical characteristic data and encoding the electrical characteristic time sequence to obtain a sequence of load current time sequence characteristics and a sequence of load voltage time sequence characteristics;

[0010] A load electrical characteristic data response association module, used for performing response association coding on each group of corresponding load current timing characteristics and load voltage timing characteristics in the sequence of load current timing characteristics and the sequence of load voltage timing characteristics to obtain a sequence of load electrical characteristic timing response association characteristics;

[0011] A load type acquisition module is used to obtain a load type label;

[0012] A load type embedding coding module, used for embedding and coding the type label of the load to obtain a load type embedding feature;

[0013] A load type query encoding module, used for performing a fast scanning semantic matching query on the sequence of the load electrical characteristic timing response associated features and the load type embedded features to obtain a load type query electrical characteristic timing semantic representation;

[0014] The prediction result generating module is used to query the electrical characteristic time series semantic representation based on the load type to obtain the prediction result, and determine whether to generate a fault warning prompt based on the prediction result.

[0015] According to another aspect of the present application, a solid-state power controller fault diagnosis method based on data analysis is provided, which includes:

[0016] A time set for collecting load electrical characteristic data of the solid-state power controller, wherein the load electrical characteristic data includes a load current value and a load voltage value;

[0017] Grouping the time set of the load electrical characteristic data and performing electrical characteristic time series encoding to obtain a sequence of load current time series characteristics and a sequence of load voltage time series characteristics;

[0018] Performing response association coding on each group of corresponding load current timing characteristics and load voltage timing characteristics in the sequence of load current timing characteristics and the sequence of load voltage timing characteristics to obtain a sequence of load electrical characteristic timing response association characteristics;

[0019] Get the type tag of the payload;

[0020] Embedding and encoding the type label of the load to obtain a load type embedding feature;

[0021] Performing a fast scanning semantic matching query on the sequence of the load electrical characteristic timing response associated features and the load type embedded features to obtain a load type query electrical characteristic timing semantic representation;

[0022] The electrical characteristic time series semantic representation is queried based on the load type to obtain a prediction result, and based on the prediction result, it is determined whether to generate a fault warning prompt.

[0023] Compared with the prior art, the solid-state power controller fault diagnosis system and method based on data analysis provided by the present application adopts artificial intelligence-based data analysis technology to organize and time-series encode the load electrical characteristic data of the solid-state power controller, and then responds to the characteristics after time-series encoding, and embeds the type label of the load to encode, so as to intelligently predict the production impact voltage value and generate corresponding fault warning prompts based on the fast semantic matching query representation information between the timing response association characteristics of each load electrical characteristic and the embedded characteristics of the load type. In this way, it can adapt to different types of load changes more flexibly, and can more accurately reflect the effect of load type on load electrical characteristics, and then deeply understand the complex behavior under different load conditions, which helps to improve the accuracy of fault prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0025] Figure 1 It is a system block diagram of a solid-state power controller fault diagnosis system based on data analysis according to an embodiment of the present application.

[0026] Figure 2 It is a block diagram of a load electrical characteristic data encoding module in a solid-state power controller fault diagnosis system based on data analysis according to an embodiment of the present application.

[0027] Figure 3 It is a block diagram of a load type query coding module in a solid-state power controller fault diagnosis system based on data analysis according to an embodiment of the present application.

[0028] Figure 4It is a flow chart of a solid-state power controller fault diagnosis method based on data analysis according to an embodiment of the present application.

[0029] Figure 5 It is an electronic circuit diagram of the practical application of the solid-state power controller according to other embodiments of the present application. DETAILED DESCRIPTION

[0030] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0031] Solid-state power controllers (SSPCs) are used in many fields to control and protect power loads. Compared with traditional mechanical relays or circuit breakers, they have faster response speeds, higher reliability, and longer service life. Patent CN118277852A proposes a fault diagnosis method and device for a solid-state power controller, which implements fault classification and predicts faults by analyzing data such as current, voltage, and temperature. However, this fault diagnosis method has limitations because it assumes that the load type remains unchanged throughout the monitoring process, but in fact the load type may change dynamically. If the load type changes and the system fails to update its status in time, the fault diagnosis results will be inaccurate. In addition, the method fails to deeply analyze the relationship between load characteristics and types, which also affects the accuracy and reliability of fault diagnosis.

[0032] Based on this, the present application proposes a solid-state power controller fault diagnosis system based on data analysis, which collects the time set of load electrical characteristic data (load current value and load voltage value) of the solid-state power controller, obtains the type label of the load, and uses data analysis and diagnosis algorithms based on artificial intelligence to sort and time-series encode the load electrical characteristic data, then responds to the characteristics after time-series encoding, and embeds the type label of the load at the same time, so as to intelligently predict and generate the impulse voltage value according to the fast semantic matching query representation information between the timing response association characteristics of each load electrical characteristic and the embedded characteristics of the load type, and generate corresponding fault warning prompts. In this way, it can adapt to different types of load changes more flexibly, solve the problem of assuming that the load type is unchanged in the patent method, thereby enhancing the adaptive ability of the system. In addition, it can more accurately reflect the effect of the load type on the load electrical characteristics, so as to deeply understand the complex behavior under different load conditions, which helps to improve the accuracy of fault prediction.

[0033] Figure 1 It is a system block diagram of a solid-state power controller fault diagnosis system based on data analysis according to an embodiment of the present application. Figure 1As shown, in the solid-state power controller fault diagnosis system 100 based on data analysis, it includes: a load electrical characteristic data acquisition module 110, which is used to collect the time set of load electrical characteristic data of the solid-state power controller, and the load electrical characteristic data includes a load current value and a load voltage value; a load electrical characteristic data encoding module 120, which is used to group the time set of the load electrical characteristic data and encode the electrical characteristic time sequence to obtain a sequence of load current timing characteristics and a sequence of load voltage timing characteristics; a load electrical characteristic data response association module 130, which is used to associate each group of corresponding load current timing characteristics and load voltage timing characteristics in the sequence of load current timing characteristics and the sequence of load voltage timing characteristics A response association coding is performed to obtain a sequence of load electrical characteristic timing response association features; a load type acquisition module 140 is used to obtain a type tag of the load; a load type embedding coding module 150 is used to embed the type tag of the load to obtain a load type embedded feature; a load type query coding module 160 is used to perform a fast scanning semantic matching query on the sequence of load electrical characteristic timing response association features and the load type embedded feature to obtain a load type query electrical characteristic timing semantic representation; a prediction result generation module 170 is used to obtain a prediction result based on the load type query electrical characteristic timing semantic representation, and based on the prediction result, determine whether to generate a fault warning prompt.

[0034] In an embodiment of the present application, the load electrical characteristic data acquisition module 110 is used to collect the time set of the load electrical characteristic data of the solid-state power controller, and the load electrical characteristic data includes the load current value and the load voltage value. It should be understood that the time set of the load electrical characteristic data of the solid-state power controller includes the load current value and the load voltage value recorded continuously over a period of time. In actual operation, the operating state of the load is a dynamic process, and only obtaining the current and voltage values ​​at a certain moment cannot accurately grasp its overall operating state. By collecting the time set, the electrical characteristics of the load under different working conditions can be obtained. For example, a large impact current will appear at the moment of equipment startup, while the current and voltage are relatively stable during the stable operation stage. By analyzing the time set of the load electrical characteristic data, the current and voltage change trends of the load can be captured, which can provide a basis for the subsequent accurate diagnosis of faults.

[0035] In the process of collecting the time collection of the load electrical characteristic data of the solid-state power controller, it is necessary to comprehensively consider various factors to ensure that the collected data is accurate, complete and can truly reflect the operating status of the load. The following is a detailed description of its implementation process:

[0036] First of all, the selection and configuration of hardware equipment is the basis. In order to accurately measure the load current value and load voltage value, it is necessary to install a high-performance sensor in the circuit connecting the solid-state power controller and the load. For current measurement, sensors based on the Hall effect principle are usually selected. This sensor has the advantages of high precision, wide measurement range and little impact on the measured circuit. It can accurately sense the size and change of current without interfering with the normal operation of the load. For voltage measurement, resistive voltage divider or capacitive voltage divider sensors can be used. These sensors can convert high voltage signals into low voltage signals suitable for collection and processing in proportion, and maintain the original characteristics of the voltage signal to a certain extent. The installation position of the sensor is very important and should be as close to the load end as possible, so as to minimize the influence of line resistance, inductance and other factors on the measurement results. Because the resistance and inductance in the line may cause voltage drop and current distortion, so that the collected data cannot accurately reflect the actual electrical characteristics of the load. For example, in a long-distance transmission line, if the sensor is installed far away from the load, the line resistance may make the measured voltage value lower than the actual end voltage of the load, which will cause errors in subsequent analysis.

[0037] At the same time, the data acquisition card is also a key device. The selected data acquisition card should have a multi-channel analog input function so that it can simultaneously receive the analog signals output by the current sensor and the voltage sensor. Its sampling frequency must be high enough to meet the needs of capturing the rapid changes in the load electrical characteristics. In some power systems, the state of the load may change very quickly, such as the start and stop process of the motor, and the instantaneous current and voltage will change dramatically. If the sampling frequency is too low, these key change information may be missed, resulting in a misjudgment of the load operation status. Generally speaking, for many loads in the field of industrial automation, the sampling frequency should reach several thousand hertz or even higher. In addition, the data acquisition card should also have a high resolution, which means that it can accurately distinguish between small current and voltage changes. In practical applications, even small changes in electrical characteristics may indicate potential problems, such as small changes in leakage current that may be caused by insulation aging. High-resolution data acquisition cards can detect these subtle changes in time. The data acquisition card also needs to have data caching and transmission functions to ensure that the collected data can be transmitted to the subsequent data processing unit in a timely and stable manner to avoid data loss or transmission delays.

[0038] In the data acquisition process, time synchronization setting is an important link. Establishing a unified time base is essential for accurately analyzing the timing relationship of the load electrical characteristic data. The data acquisition card usually has a clock synchronization mechanism inside to ensure that the data collected by each channel is strictly aligned in time. Alternatively, a high-precision clock source can be connected externally to further improve the accuracy of time synchronization. For example, when analyzing the phase relationship between load current and voltage, accurate time synchronization can ensure that the measured current and voltage changes occur at the same time, thereby accurately calculating the phase difference between them. If there is an error in time synchronization, it may lead to errors in the calculation of the phase relationship, which in turn affects the judgment of the load properties and may mislead the analysis of the cause of the fault during fault diagnosis.

[0039] After starting the data acquisition system, the data acquisition card works continuously at the set sampling frequency. It converts the analog signal output by the sensor into a digital signal and stores it in chronological order. The collected time set should cover the entire operation process of the load, including the startup phase, stable operation phase, dynamic change phase, and stop phase. At the moment of load startup, a large impact current often occurs, which is caused by the sudden change of the equivalent impedance of the load at startup. Continuous data collection can fully record the current changes in this process, providing an important basis for subsequent analysis of the load startup characteristics and judging whether there are hidden faults during the startup process. In the stable operation stage, the current and voltage of the load are relatively stable, but there may be some subtle changes, such as current fluctuations caused by resistance changes caused by changes in the internal temperature of the load. These changes also need to be accurately recorded. When the load changes dynamically, such as the load's working mode switching or a sudden change in the load amount, the current and voltage will change accordingly, and the collected data can reflect the electrical characteristics of the load under different working conditions.

[0040] In order to ensure the quality of the collected data, a series of data processing operations are required. During the collection process, real-time data verification is essential. By pre-setting a reasonable data range, check whether the collected data is within the range, and promptly discover and process abnormal values ​​or erroneous values. For example, if the collected current value exceeds the current range of the load during normal operation by several times, this is likely caused by sensor failure or strong electromagnetic interference during the collection process. For this type of abnormal data, it is necessary to mark it for further verification and correction. At the same time, digital filtering technology is used to filter the collected data to remove noise interference. Common filtering methods include mean filtering, median filtering, and low-pass filtering. Mean filtering smoothes the data by calculating the average value of the data in a certain time window, which is suitable for removing random noise; median filtering sorts the data in the window by size and takes the middle value as the filtering result, which is better for removing pulse noise; low-pass filtering can filter out high-frequency noise and retain the low-frequency components of the signal, which has a good inhibitory effect on high-frequency noise caused by factors such as power supply ripple. For example, in an industrial environment, electromagnetic interference from surrounding equipment may introduce high-frequency noise into the collected current and voltage signals. Low-pass filtering can effectively remove this noise and make the data smoother, thereby improving the accuracy of subsequent analysis of the load electrical characteristics.

[0041] In addition, it is necessary to regularly check whether the collected data is complete. Make sure that there is no data loss during the entire collection time range, because any data loss may lead to a misunderstanding of the load behavior and inaccurate fault diagnosis. If data loss is found, it is necessary to analyze the cause in depth, such as data acquisition card failure, transmission line interruption, or full storage device. Take appropriate measures according to the specific cause, such as re-collecting data for the lost time period, or using methods such as data interpolation to supplement when the amount of data loss is small and the data changes are relatively smooth. Data interpolation can estimate the value of the missing data point based on the rules of known data points, such as using linear interpolation, polynomial interpolation and other methods. However, it should be noted that the interpolation method should be selected according to the characteristics of the data and the actual application scenario to minimize the impact of interpolation errors on the data analysis results.

[0042] In summary, the time collection of load electrical characteristic data of solid-state power controllers is a complex process involving hardware equipment selection and configuration, data acquisition process control, and data quality assurance. Only by fully and meticulously implementing each link can high-quality load electrical characteristic data be obtained, thereby providing a solid data foundation for subsequent fault diagnosis and system operation status analysis.

[0043] In the embodiment of the present application, the load electrical characteristic data encoding module 120 is used to group the time set of the load electrical characteristic data and perform electrical characteristic time sequence encoding to obtain a sequence of load current time sequence characteristics and a sequence of load voltage time sequence characteristics. Specifically, Figure 2 FIG. 1 is a block diagram of a load electrical characteristic data encoding module in a solid-state power controller fault diagnosis system based on data analysis according to an embodiment of the present application. Figure 2 As shown, the load electrical characteristic data encoding module 120 includes: a load electrical characteristic data grouping unit 121, which is used to use a data grouper to group the time set of the load electrical characteristic data according to the load electrical characteristic parameter sample dimension to obtain a load current timing vector and a load voltage timing vector; a load current and load voltage timing feature generating unit 122, which is used to pass the load current timing vector and the load voltage timing vector respectively through an electrical characteristic timing encoder based on a 1D-FCN model to obtain a sequence of load current timing feature vectors as the sequence of load current timing features and a sequence of load voltage timing feature vectors as the sequence of load voltage timing features.

[0044] In an embodiment of the present application, the load electrical characteristic data grouping unit 121 is used to use a data grouper to group the time set of the load electrical characteristic data according to the load electrical characteristic parameter sample dimension to obtain a load current timing vector and a load voltage timing vector. Accordingly, considering that both the load current value and the load voltage value have timing characteristics on a time scale, different load electrical data have different characteristics in the change on a time scale. Therefore, in order to understand and analyze each parameter more finely granularly, in the technical solution of the present application, a data grouper is used to group the time set of the load electrical characteristic data according to the load electrical characteristic parameter sample dimension to obtain a load current timing vector and a load voltage timing vector. In this way, the timing change trend and fluctuation of each specific electrical information in time can be understood and processed more clearly.

[0045] In an embodiment of the present application, the load current load voltage timing feature generation unit 122 is used to pass the load current timing vector and the load voltage timing vector through an electrical characteristic timing encoder based on a 1D-FCN model to obtain a sequence of load current timing feature vectors as a sequence of load current timing features and a sequence of load voltage timing feature vectors as a sequence of load voltage timing features. It should be understood that in order to further capture and mine complex timing feature information from each load electrical characteristic parameter, such as local pattern fluctuations of load current values ​​in a short period of time and global timing trends of load voltage values ​​between long spans. In the technical solution of the present application, the load current timing vector and the load voltage timing vector are respectively passed through an electrical characteristic timing encoder based on a 1D-FCN model to obtain a sequence of load current timing feature vectors and a sequence of load voltage timing feature vectors. It can be understood that the 1D-FCN model is good at processing time series data and can simultaneously capture local patterns (such as fluctuations in a short period of time) and global trends (such as long-term trend changes) in the data. This is particularly important for the complex and changeable current and voltage data in the power system. Specifically, the 1D-FCN model uses a deconvolution layer to upsample the features extracted by the last convolution layer to increase its resolution to the same as the original input data. This processing not only retains the temporal details of the input data, but also ensures that the model output can be accurately mapped back to the corresponding position of the input data. This is crucial for fault diagnosis, because only by accurately determining the specific location of the problem can the problem be effectively solved. In addition, the 1D-FCN model introduces jump connections to fuse the features of the earlier convolutional layers with the upsampled features. This strategy not only retains the global features of the data, but also improves the accuracy of the division of local areas, which helps to capture more detailed information. In fault diagnosis, this means that the model can not only identify overall trend changes, but also detect local minor anomalies, thereby greatly improving the accuracy of the diagnosis results.

[0046] The process of passing the load current timing vector and the load voltage timing vector through the electrical characteristic timing encoder based on the 1D-FCN model to obtain the sequence of load current timing feature vectors as the sequence of load current timing features and the sequence of load voltage timing feature vectors as the sequence of load voltage timing features is a complex process involving model architecture principles, data processing flow, model training optimization, etc. The following is a detailed description of its implementation process:

[0047] The core architecture and principles of the 1D-FCN model play a key role in processing the load current timing vector and the load voltage timing vector. The model is mainly composed of convolutional layers, deconvolutional layers, and jump connections. The convolutional layer uses convolution kernels of different sizes (such as 3x1, 5x1, etc.) to slide calculations on the input vector to extract local features. Multiple convolutional layers can be stacked to gradually obtain higher-level abstract features, such as capturing short-term fluctuations and long-term trends from the load current timing vector. The deconvolution layer upsamples the feature vector output by the last convolutional layer, and restores the resolution to the same as the original input according to the set parameters (such as the deconvolution kernel size, step size, etc.), ensuring that the time details are restored. The jump connection fuses the features of the early convolutional layer with the upsampled features, retaining global and local features and improving the accuracy of local area division.

[0048] When processing the load current time series vector, it first passes through the convolution layer, and the neurons are operated according to the convolution kernel weights to obtain the feature map, and multiple convolution kernel operations generate the output feature vector. Then, the last convolution layer outputs the feature vector and enters the deconvolution layer, and the features are reconstructed by upsampling according to the rules to restore the time scale correspondence. Finally, the features are fused through jump connections to obtain a sequence of load current time series feature vectors to fully reflect the current characteristics.

[0049] For the load voltage time series vector, the processing flow is similar. First, the convolution layer extracts features at different levels, such as instantaneous changes and periodic fluctuations. Then the deconvolution layer upsamples to restore the resolution, and reconstructs the interpolation based on the voltage characteristics and learning parameters. Finally, through jump connection fusion, a sequence of load voltage time series feature vectors is obtained to accurately describe the voltage change characteristics.

[0050] In order for the model to accurately extract features, training data preparation is crucial. It is necessary to cover the current and voltage time series data of different types of loads under various working conditions, and perform normalization preprocessing to ensure data quality and diversity. During training, a suitable loss function (such as the mean square error function) is used to measure the difference between prediction and reality, and the weight parameters are adjusted through the back propagation algorithm. The optimization algorithm (such as Adam) is combined to accelerate convergence and avoid local optimality. At the same time, regularization techniques (such as L2 regularization) are used to prevent overfitting and enhance the generalization ability of the model. After multiple iterative training, the 1D-FCN model can accurately encode the load electrical characteristic data and output a high-quality time series feature vector sequence.

[0051] In an embodiment of the present application, the load electrical characteristic data response association module 130 is used to perform response association encoding on each group of corresponding load current timing characteristics and load voltage timing characteristics in the sequence of the load current timing characteristics and the sequence of the load voltage timing characteristics to obtain a sequence of load electrical characteristic timing response association characteristics. Specifically, in an embodiment of the present application, the load electrical characteristic data response association module is used to: calculate the position point multiplication between each group of corresponding load current timing characteristic vectors and load voltage timing characteristic vectors in the sequence of the load current timing characteristic vectors and the sequence of the load voltage timing characteristic vectors to obtain a sequence of load electrical characteristic timing response association characteristic vectors as the sequence of the load electrical characteristic timing response association characteristics. It should be understood that there is a mutual response association relationship between the load current timing characteristics and the load voltage timing characteristics at each time point, such as voltage fluctuations will cause corresponding current changes. Therefore, in the technical solution of the present application, each group of corresponding load current timing characteristic vectors and load voltage timing characteristic vectors in the sequence of the load current timing characteristic vectors and the sequence of the load voltage timing characteristic vectors are subjected to response association coding to capture and reveal the timing dynamic interaction and intrinsic connection between the load current and voltage to obtain a sequence of load electrical characteristic timing response association characteristic vectors. In particular, in a specific embodiment of the present application, the vector obtained by calculating the position point multiplication between each group of corresponding load current timing characteristic vectors and load voltage timing characteristic vectors can be used as the load electrical characteristic timing response association characteristic vector between the two. In other words, the position point multiplication can directly capture the interaction effect between the current and voltage characteristics at the same time point. This interaction may be nonlinear, reflecting the complex behavior patterns in the power system. For example, in some cases, even if the current does not change much, the sharp fluctuation of the voltage may also indicate potential problems. In this way, the timing interaction effect between the current and the voltage can be effectively captured without increasing the complexity, providing strong support for the fault diagnosis of the solid-state power controller.

[0052] In the embodiment of the present application, the load type acquisition module 140 is used to obtain the type label of the load. In particular, the type label of the load is such as resistive load, capacitive load, inductive load, etc. It should be understood that the type label of the load contains the basic electrical property information of the load and is a classification mark of the load characteristics. There are various types of loads, such as resistive, capacitive and inductive loads, and their electrical characteristics are essentially different. Specifically, the resistive load current is in phase with the voltage and follows Ohm's law; the capacitive load current leads the voltage, which will produce capacitor charging and discharging phenomena; the inductive load current lags behind the voltage, and there is an inductive energy storage and release process. Different types of loads show different modes and characteristics when a fault occurs. For example, a capacitive load may generate a large surge current due to capacitor charging at the moment of power-on. If it exceeds the bearing capacity of the solid-state power controller, it may cause overcurrent protection action; the inductive load will generate overvoltage due to the back electromotive force of the inductor when the power is off, which may damage the controller or other equipment. Here, obtaining the type label of the load helps to more accurately determine the fault type and cause of the solid-state power controller.

[0053] In an embodiment of the present application, the load type embedding coding module 150 is used to embed the type label of the load to obtain a load type embedding feature. Specifically, in an embodiment of the present application, the load type embedding coding module is used to: embed the type label of the load using a load type label embedding matrix to obtain a load type embedding coding vector as the load type embedding feature. It should be understood that the type label of the load contains key semantic information of different load types, and the key semantic information contains different load electrical characteristics and operating mode information. Therefore, in order to provide a richer and more meaningful load type representation and better capture the association between each type, in the technical solution of the present application, the type label of the load is embedded and encoded to obtain a load type embedding coding vector, so that the model can be understood and explained at a higher level. In particular, in a specific embodiment of the present application, the type label of the load can be embedded and encoded using a load type label embedding matrix to map the load type label to a low-dimensional continuous space, thereby better capturing the semantic association relationship.

[0054] In an embodiment of the present application, the load type query encoding module 160 is used to perform a fast scanning semantic matching query on the sequence of the load electrical characteristic timing response associated features and the load type embedded features to obtain the load type query electrical characteristic timing semantic representation. In particular, considering that the sequence of the load electrical characteristic timing response associated features captures the current and voltage changes of the load at different time points, it reflects the timing dynamic evolution characteristics of the electrical characteristics during the operation of the load. The load type embedded features represent the semantic information of the category of each load, including the working mode and relationship between the load current and the load voltage. Therefore, in order to extract the semantic information and features most closely related to the load type embedded features from the sequence of the load electrical characteristic timing response associated features, so as to accurately select highly relevant load electrical characteristics. In the technical solution of the present application, a fast scanning semantic matching query is performed on the sequence of the load electrical characteristic timing response associated features and the load type embedded features to obtain the load type query electrical characteristic timing semantic representation. In this way, not only the dynamic electrical behavior and static category information are integrated, but also the model's ability to understand data is enhanced, enabling it to analyze and reason at a higher level of abstraction. The fast scanning mechanism ensures that similar patterns or anomalies can be efficiently found even in large-scale data sets, improving the system's response speed and efficiency.

[0055] Specifically, Figure 3 FIG. 1 is a block diagram of a load type query encoding module in a solid-state power controller fault diagnosis system based on data analysis according to an embodiment of the present application. Figure 3 As shown, the load type query encoding module 160 includes: a load type-load electrical characteristic semantic maximum value calculation unit 161, which is used to calculate the first load type-load electrical characteristic semantic maximum value and the second load type-load electrical characteristic semantic maximum value of the sequence of the load electrical characteristic timing response associated feature vector based on the load type embedded coding vector; a load electrical characteristic fast matching unit 162, which is used to determine a subsequence of the load electrical characteristic timing response associated feature vector that is quickly matched from the sequence of the load electrical characteristic timing response associated feature vector based on the first load type-load electrical characteristic semantic maximum value and the second load type-load electrical characteristic semantic maximum value; a load type query electrical characteristic timing semantic representation generation unit 163, which is used to perform a linear transformation on the load type embedded coding vector to obtain a load type query vector and a load type value vector, and use the subsequence of the load electrical characteristic timing response associated feature vector as a subsequence of the key vector, and perform a converter-based cross-domain query encoding on the load type query vector, the load type value vector and the subsequence of the key vector to obtain a load type query electrical characteristic timing semantic representation vector as the load type query electrical characteristic timing semantic representation.

[0056] Specifically, in the embodiment of the present application, the load type-load electrical characteristic semantic maximum value calculation unit is used to calculate the mutual information between the load type embedded coding vector and each load electrical characteristic timing response associated feature vector in the sequence of the load electrical characteristic timing response associated feature vector to obtain a sequence of load type-load electrical characteristic semantic fast semantic matching factors. The process can be expressed as follows:

[0057] V={v 21 ,v 22 ,...,v 2i ,...,v 2n}

[0058] x1 = sigmoid(v1)

[0059] x 2i =sigmoid(v 2i )

[0060] MI i =H(x1)+H(x 2i )-H(x1,x 2i )

[0061] Wherein, V is the sequence of the associated characteristic vectors of the load electrical characteristic timing response, v 21 , v 22 , v 2i and v 2n are the first, second, i-th and n-th load electrical characteristic timing response associated feature vectors in the sequence of the load electrical characteristic timing response associated feature vectors, v1 is the load type embedding coding vector, sigmoid(·) is the sigmoid function, x1 is the load type embedding coding transformation feature vector, x 2i Yes 2i The corresponding load electrical characteristic timing response associated transformation feature vector, H(x1) represents the information entropy of x1, H(x 2i ) represents x 2i The information entropy of H(x1,x 2i ) represents x1 and x 2i The joint entropy between i represents x1 and x 2i The load type-load electrical characteristic semantics fast semantic matching factor between them;

[0062] The first load type-load electrical characteristic semantic maximum value and the second load type-load electrical characteristic semantic maximum value are identified from the sequence of the load type-load electrical characteristic semantic fast semantic matching factors. The process can be expressed as:

[0063]

[0064] in, is a sequence of load type-load electrical characteristic semantic fast semantic matching factors, n is the number of load type-load electrical characteristic semantic fast semantic matching factors in the sequence of load type-load electrical characteristic semantic fast semantic matching factors, max(·) is the maximum value in the sequence, MI max1 The first load type - load electrical characteristics semantic maximum value, MI max2 It is the second load type - the maximum value of the load electrical characteristic semantics.

[0065] It should be understood that, first, by calculating the mutual information between the load type embedded coding vector and each load electrical characteristic timing response associated feature vector in the sequence of the load electrical characteristic timing response associated feature vector, it can be used to evaluate the information overlap between the load type information and the load electrical characteristics at different times. The larger the calculated mutual information value, the more highly correlated the load type information is with the load electrical characteristic data at a certain moment, which can help identify the load electrical characteristic timing response associated feature vector that is most relevant to the load type embedded coding vector in the sequence of the load electrical characteristic timing response associated feature vector. In addition, the mutual information value is also used as a quantified data to provide an objective standard for subsequent fast matching. Next, it is necessary to identify the first load type-load electrical characteristic semantic maximum value and the second load type-load electrical characteristic semantic maximum value from the sequence of the load type-load electrical characteristic semantic fast semantic matching factor. This process is actually looking for the two load electrical characteristic timing response associated feature vectors that are most semantically relevant to the load type embedded coding vector, so that the load type can be more accurately matched with the load electrical characteristics at a specific moment, thereby helping to understand the key behavior pattern of the load at a specific moment.

[0066] Specifically, in the embodiment of the present application, the load electrical characteristic fast matching unit is used to determine a subsequence of the load electrical characteristic timing response associated feature vectors that are fast matched from the sequence of the load electrical characteristic timing response associated feature vectors based on the positions of the first load type-load electrical characteristic semantic maximum value and the second load type-load electrical characteristic semantic maximum value in the sequence of the load type-load electrical characteristic semantic fast semantic matching factors. The above process can be expressed as:

[0067]

[0068] {k sub}={v 2arg max1 ,...,v 2j ,...,v 2arg max2}

[0069] in, is a sequence of load type-load electrical characteristic semantic fast semantic matching factors, n is the number of load type-load electrical characteristic semantic fast semantic matching factors in the sequence of load type-load electrical characteristic semantic fast semantic matching factors, MI max1 is the first load type - the maximum value of the load electrical characteristic semantics, arg max·) is the position of the maximum value in the sequence, arg max1 is the first determined fast matching position, arg max2 is the second determined fast matching position, {k sub} is a subsequence of the associated characteristic vector of the load electrical characteristic timing response, v 2arg max1 、v 2j 、v 2arg max2 Respectively represent the load electrical characteristic timing response associated feature vectors of the first determined fast matching position, the jth position, and the second determined fast matching position in the subsequence of the load electrical characteristic timing response associated feature vector.

[0070] It should be understood that based on the positions of the first load type-load electrical characteristic semantic maximum value and the second load type-load electrical characteristic semantic maximum value in the sequence of the load type-load electrical characteristic semantic fast semantic matching factors, a subsequence of the load electrical characteristic timing response associated feature vectors that is quickly matched is determined from the sequence of the load electrical characteristic timing response associated feature vectors, and this process can be regarded as a preliminary feature selection. In this way, computing resources can be concentrated to process those load electrical characteristic timing response associated feature vectors that have meaningful interactions with the load type embedded coding vector, which can improve the efficiency of model data processing while also improving the quality of subsequent controller fault diagnosis.

[0071] Finally, the load type embedded coding vector is linearly transformed to obtain a load type query vector and a load type value vector, and a subsequence of the load electrical characteristic timing response associated feature vector is used as a subsequence of the key vector, and the load type query vector, the load type value vector and the subsequence of the key vector are subjected to converter-based cross-domain query encoding to obtain a load type query electrical characteristic timing semantic representation vector as the load type query electrical characteristic timing semantic representation. The above process can be expressed as:

[0072] v q =v1W q +b q

[0073] v v =v1W v +b v

[0074]

[0075] Wherein, v1 is the payload type embedding coding vector, arg max1 is the first determined fast matching position, argmax2 is the second determined fast matching position, W q and b q are the query embedding matrix and the query bias vector, respectively, v q is the load type query vector, W v and b v are the value embedding matrix and value bias vector, respectively, v v is the load type value vector, v 2j T Yes 2j The transposed vector of v 2j The length of , softmax(·) is the softmax function, is the matrix multiplication, v r It is the load type query electrical characteristic timing semantic representation vector.

[0076] It should be understood that by performing a linear transformation on the load type embedding coding vector to obtain a load type query vector and a load type value vector, and using a subsequence of the load electrical characteristic timing response associated feature vector as a subsequence of the key vector, the load type query vector, the load type value vector, and the subsequence of the key vector are subjected to converter-based cross-domain query encoding, and the self-attention mechanism at the core of the converter structure can be used to capture and utilize the long-distance dependency between the load type embedding coding vector and the subsequence of the load electrical characteristic timing response associated feature vector to generate an accurate load type query electrical characteristic timing semantic representation vector. Here, the load type query vector is mainly responsible for finding the key vector most relevant to itself in the subsequence of the load electrical characteristic timing response associated feature vector, while the load type value vector carries the specific information of the load type. Through this encoding method, the timing changes of the load electrical characteristic data can be more accurately grasped, the load behavior can be more comprehensively understood, and a more reliable basis can be provided for subsequent fault diagnosis.

[0077] In an embodiment of the present application, the prediction result generation module 170 is used to query the electrical characteristic timing semantic representation based on the load type to obtain a prediction result, and based on the prediction result, determine whether to generate a fault warning prompt. Specifically, in an embodiment of the present application, the prediction result generation module is used to: pass the load type query electrical characteristic timing semantic representation vector through a decoder-based closed impulse voltage predictor to obtain the prediction result, and the prediction result is used to represent the predicted decoded value of the impulse voltage; based on the comparison between the predicted decoded value of the impulse voltage and a preset threshold, determine whether to generate a fault warning prompt. It should be understood that the impulse voltage is an important parameter in the operation of the solid-state power controller, and its size is directly related to the safe and stable operation of the controller and the load equipment. Excessive impulse voltage can cause equipment damage, so accurate prediction of the impulse voltage is crucial for fault diagnosis and system protection. Accordingly, considering that the load type query electrical characteristic time series semantic representation vector contains the correlation information between the load type, load current and voltage parameters, in order to parse and process the feature information in this high-dimensional space to obtain the key indicator data value of the impulse voltage generated at the moment of load closing, the technical solution of the present application needs to input the load type query electrical characteristic time series semantic representation vector into the closed impulse voltage predictor based on the decoder. The decoder is a neural network model that can reversely convert the encoded feature information through training to generate the final prediction result, that is, the predicted decoded value of the impulse voltage. After obtaining the predicted decoded value of the impulse voltage, it is compared with the preset threshold value so that the system can actively monitor the possible fault risks under the load operation state. The preset threshold value is a critical value determined based on factors such as the design specifications of the solid-state power controller, the tolerance of the load equipment, and the system safety operation requirements. When the predicted decoded value exceeds the threshold, it means that the load may generate an impulse voltage that is harmful to the equipment during future operation. At this time, it is necessary to issue a fault warning prompt in time to provide a clear basis for action for relevant operation and maintenance personnel. For example, when it is determined that a fault warning prompt needs to be generated, a common way is to start an audible and visual alarm device. In the circuit where the solid-state power controller is located, sound and light alarm components such as indicator lights (such as red LED lights) and buzzers are installed. When the impulse voltage prediction decoding value exceeds the threshold, the control circuit will light up the red LED light, and the buzzer will emit a loud alarm sound. In addition to the sound and light alarm, the system can also push fault warning information to relevant operation and maintenance personnel. Through network communication technologies such as Ethernet and wireless communication, the warning information is sent to the monitoring terminal (such as computers, mobile phones, etc.) of the operation and maintenance personnel. The warning information includes the location where the fault may occur (such as the specific solid-state power controller number), the expected impulse voltage value, the time of occurrence and other details. At the same time, the system will automatically record the warning event, including the warning time, relevant parameter values ​​and other information, for subsequent fault analysis and troubleshooting.

[0078] In particular, in an embodiment of the present application, a possible implementation method of passing the load type query electrical characteristic timing semantic representation vector through a decoder-based closed impulse voltage predictor to obtain a prediction result for representing a predicted decoding value of the impulse voltage is: multiplying the decoding weight matrix of the decoder with the load type query electrical characteristic timing semantic representation vector to obtain a decoded load type query electrical characteristic timing semantic representation vector, and accumulating and summing all eigenvalues ​​of the decoded load type query electrical characteristic timing semantic representation vector to obtain the predicted decoding value of the impulse voltage.

[0079] Here, when the sequence of the load electrical characteristic timing response associated feature vector and the load type embedded coding vector respectively represent the load current value-voltage value timing associated response coding feature and the embedded coding feature of the load type label, after performing semantic query encoding based on the fast scanning mechanism, the load type query electrical characteristic timing semantic representation vector will also have significant semantic query coding distribution space structure differences due to the differences in the feature scanning mechanism of heterogeneous data, affecting the convergence consistency of the decoder, thereby affecting the accuracy of the prediction results obtained by the closed impulse voltage predictor.

[0080] Therefore, in one example, the load type query electrical characteristic time series semantic representation vector is optimized, and the optimization process includes:

[0081] The absolute value sum of all eigenvalues ​​of the load type query electrical characteristic time series semantic representation vector and the square root of the square sum are calculated to obtain the first load type query electrical characteristic time series semantic space structure value and the second load type query electrical characteristic time series semantic space structure value. The process can be expressed as:

[0082]

[0083] Among them, v i represents the characteristic value of the ith position of the load type query electrical characteristic time series semantic representation vector, w1 represents the first load type query electrical characteristic time series semantic space structure value, and w2 represents the second load type query electrical characteristic time series semantic space structure value;

[0084] Determine the number of eigenvalues ​​of all eigenvalues ​​of the load type query electrical characteristic time series semantic representation vector, that is, the length L of the load type query electrical characteristic time series semantic representation vector;

[0085] For each eigenvalue of the load type query electrical characteristic time series semantic representation vector, the first load type query electrical characteristic time series semantic space structure value minus the product of the eigenvalue and the number of eigenvalues ​​is calculated to obtain the first load type query electrical characteristic time series semantic long-range dependency value. The process can be expressed as:

[0086] x i =w1-v i ×L

[0087] Among them, w1 represents the time-series semantic space structure value of the first load type query electrical characteristics, v i represents the eigenvalue of the i-th position of the load type query electrical characteristic time series semantic representation vector, L represents the length of the load type query electrical characteristic time series semantic representation vector, x i Indicates that the first load type queries the electrical characteristic timing semantic long-range dependency value;

[0088] The product of the square root of the number of eigenvalues ​​multiplied by the eigenvalues ​​minus the second load type query electrical characteristic temporal semantic space structure value is calculated to obtain the second load type query electrical characteristic temporal semantic long-range dependency value. The process can be expressed as:

[0089]

[0090] Where L represents the length of the load type query electrical characteristic time series semantic representation vector, v i represents the eigenvalue of the i-th position of the load type query electrical characteristic time series semantic representation vector, w2 represents the second load type query electrical characteristic time series semantic space structure value, y i Indicates the second load type query electrical characteristic timing semantic long-range dependency value;

[0091] The exponential value calculated by taking the first load type query electrical characteristic time series semantic long-range dependency value as the exponent of the natural constant and the inverse of the second load type query electrical characteristic time series semantic long-range dependency value are weighted summed to obtain the optimized eigenvalue corresponding to each eigenvalue. The process can be expressed as:

[0092]

[0093] Among them, x i Indicates the long-range dependency value of the first load type query electrical characteristic timing semantics, y i represents the long-range dependency value of the electrical characteristic timing semantics of the second load type query, α and β represent weighted hyperparameters, and v' i The optimized eigenvalue of the eigenvalue of the i-th position of the load type query electrical characteristic time series semantic representation vector;

[0094] The optimized characteristic values ​​are combined into an optimized load type query electrical characteristic timing semantic representation vector.

[0095] That is, in view of the possible spatial structure loss of the feature set of the load type query electrical characteristic time series semantic representation vector in the high-dimensional space, which causes the decoder weight to implicitly infer the spatial structure information based on the features, resulting in inconsistent convergence. A long-distance feature dependency relationship is established based on the overall feature scale of the load type query electrical characteristic time series semantic representation vector relative to the spatial structure representation of the load type query electrical characteristic time series semantic representation vector, so as to establish the feature local connectivity of the load type query electrical characteristic time series semantic representation vector, and capture the spatial ambiguous information of the object feature value through the unstructured feature value point prediction of the load type query electrical characteristic time series semantic representation vector, thereby improving the spatial inductive deviation perception ability of the feature set of the load type query electrical characteristic time series semantic representation vector, improving the convergence consistency of the decoder, and improving the accuracy of the prediction result obtained by the decoder-based closed impulse voltage predictor for the load type query electrical characteristic time series semantic representation vector.

[0096] In summary, a solid-state power controller fault diagnosis system 100 based on data analysis according to an embodiment of the present application is explained, which uses artificial intelligence-based data analysis technology to organize and time-series encode the load electrical characteristic data of the solid-state power controller, and then performs response association on the time-series encoded features, while embedding and encoding the type label of the load, so as to intelligently predict the production impact voltage value and generate corresponding fault warning prompts based on the fast semantic matching query representation information between the timing response association features of each load electrical characteristic and the embedded features of the load type. In this way, it can adapt to different types of load changes more flexibly, and can more accurately reflect the effect of the load type on the load electrical characteristics, and then deeply understand the complex behavior under different load conditions, which helps to improve the accuracy of fault prediction.

[0097] Figure 4 FIG. 1 is a flow chart of a solid-state power controller fault diagnosis method based on data analysis according to an embodiment of the present application. Figure 4As shown, in the solid-state power controller fault diagnosis method based on data analysis, it includes: S110, collecting the time set of load electrical characteristic data of the solid-state power controller, the load electrical characteristic data including the load current value and the load voltage value; S120, grouping the time set of the load electrical characteristic data and encoding the electrical characteristic time sequence to obtain a sequence of load current timing characteristics and a sequence of load voltage timing characteristics; S130, encoding each group of corresponding load current timing characteristics and load voltage timing characteristics in the sequence of load current timing characteristics and the sequence of load voltage timing characteristics Perform response association coding to obtain a sequence of load electrical characteristic timing response association features; S140, obtain the type label of the load; S150, embed the type label of the load to obtain a load type embedded feature; S160, perform a fast scanning semantic matching query on the sequence of load electrical characteristic timing response association features and the load type embedded feature to obtain a load type query electrical characteristic timing semantic representation; S170, obtain a prediction result based on the load type query electrical characteristic timing semantic representation, and determine whether to generate a fault warning prompt based on the prediction result.

[0098] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned solid-state power controller fault diagnosis method based on data analysis have been referred to above. Figures 1 to 3 The description of the solid-state power controller fault diagnosis system based on data analysis has been introduced in detail, and therefore, its repeated description will be omitted.

[0099] In summary, a solid-state power controller fault diagnosis method based on data analysis based on the embodiment of the present application is explained, which uses artificial intelligence-based data analysis technology to organize and time-series encode the load electrical characteristic data of the solid-state power controller, and then responds to the characteristics after the time-series encoding, and embeds the type label of the load to encode, so as to intelligently predict the production impact voltage value and generate corresponding fault warning prompts based on the fast semantic matching query representation information between the timing response association characteristics of each load electrical characteristic and the embedded characteristics of the load type. In this way, it can adapt to different types of load changes more flexibly, and can more accurately reflect the effect of load type on load electrical characteristics, and then deeply understand the complex behavior under different load conditions, which helps to improve the accuracy of fault prediction.

[0100] In other embodiments of the present application, a solid-state power controller is also provided. Figure 5 FIG. 1 is an electronic circuit diagram of a solid-state power controller according to another embodiment of the present application. Figure 5As shown in the figure, the Vline on the far left is the input power line, which provides power for the entire circuit; Ven is the control power input, which is used to drive the control circuit; the power supply symbol represents the power supply device in the circuit, such as a transformer, which is responsible for providing the required power; the control power supply provides the necessary power for the control chip; K1 is a switch used to control the switch of the circuit and perform the reset function; Vcc is the power input pin of the chip, which is used to provide the required operating voltage for the circuit inside the chip; EN is the enable pin of the chip, which is used to control whether the chip is in working state; GND is the ground pin of the chip, which provides a reference potential for the chip to ensure the normal operation of the internal circuit of the chip; Trip is the protection pin of the chip Pins are usually used for overcurrent, overvoltage or overheating protection; CANRX and CANTX are the receiving and transmitting pins of the controller area network (CAN), allowing the chip to communicate with other CAN network devices; Vline and Vout are the input voltage pin and output voltage pin of the chip, respectively, which are used to receive the voltage signal of the external input and output the voltage signal after being processed by the internal circuit of the chip; HL1 is a light-emitting diode, which is usually used to indicate the working status of the circuit, such as power status or fault status; R1 is a resistor, which is used to limit the current flowing through the light-emitting diode to protect the light-emitting diode from being damaged by excessive current; the load is the output end of the circuit and is connected to the device that needs power supply. The working principle of the whole circuit is: Vline provides power, which is converted into the required voltage and current by the power supply. The control power supply powers the control circuit, and the K1 switch controls the opening or closing of the circuit. The chip is powered by Vcc and GND, and communicates with other devices through CANRX and CANTX. When the circuit is working normally, HL1 will light up to indicate the circuit status. If the circuit detects an abnormality, the Trip signal may be triggered to notify to take protective measures. R1 is used to protect HL1 and ensure that it works within a safe current range. The load receives power through Vout for its normal operation.

Claims

1. A solid-state power controller fault diagnosis system based on data analysis, characterized in that: include: A load electrical characteristic data acquisition module, used to acquire a time set of load electrical characteristic data of the solid-state power controller, wherein the load electrical characteristic data includes a load current value and a load voltage value; A load electrical characteristic data encoding module, used for grouping the time set of the load electrical characteristic data and encoding the electrical characteristic time sequence to obtain a sequence of load current time sequence characteristics and a sequence of load voltage time sequence characteristics; A load electrical characteristic data response association module, used for performing response association coding on each group of corresponding load current timing characteristics and load voltage timing characteristics in the sequence of load current timing characteristics and the sequence of load voltage timing characteristics to obtain a sequence of load electrical characteristic timing response association characteristics; A load type acquisition module is used to obtain a load type label; A load type embedding coding module, used for embedding and coding the type label of the load to obtain a load type embedding feature; A load type query encoding module, used for performing a fast scanning semantic matching query on the sequence of the load electrical characteristic timing response associated features and the load type embedded features to obtain a load type query electrical characteristic timing semantic representation; The prediction result generating module is used to query the electrical characteristic time series semantic representation based on the load type to obtain the prediction result, and determine whether to generate a fault warning prompt based on the prediction result.

2. The solid-state power controller fault diagnosis system based on data analysis according to claim 1 is characterized in that: The load electrical characteristic data encoding module includes: A load electrical characteristic data grouping unit, configured to use a data grouper to group the time set of the load electrical characteristic data according to the load electrical characteristic parameter sample dimension to obtain a load current timing vector and a load voltage timing vector; A load current and load voltage timing feature generating unit is used to pass the load current timing vector and the load voltage timing vector through an electrical characteristic timing encoder based on a 1D-FCN model respectively to obtain a sequence of load current timing feature vectors as a sequence of load current timing features and a sequence of load voltage timing feature vectors as a sequence of load voltage timing features.

3. The solid-state power controller fault diagnosis system based on data analysis according to claim 2 is characterized in that: The load electrical characteristic data response association module is used to: calculate the position point multiplication between each group of corresponding load current timing characteristic vectors and load voltage timing characteristic vectors in the sequence of the load current timing characteristic vectors and the sequence of the load voltage timing characteristic vectors to obtain a sequence of load electrical characteristic timing response association characteristic vectors as the sequence of the load electrical characteristic timing response association characteristics.

4. The solid-state power controller fault diagnosis system based on data analysis according to claim 3 is characterized in that: The load type embedding coding module is used to: use the load type label embedding matrix to embed the type label of the load to obtain a load type embedding coding vector as the load type embedding feature.

5. The solid-state power controller fault diagnosis system based on data analysis according to claim 4 is characterized in that: The load type query encoding module includes: A load type-load electrical characteristic semantic maximum value calculation unit, used to calculate a first load type-load electrical characteristic semantic maximum value and a second load type-load electrical characteristic semantic maximum value of a sequence of characteristic vectors associated with the load electrical characteristic timing response based on the load type embedded coding vector; A load electrical characteristic fast matching unit, configured to determine a subsequence of load electrical characteristic timing response associated feature vectors for fast matching from the sequence of load electrical characteristic timing response associated feature vectors based on the first load type-load electrical characteristic semantic maximum value and the second load type-load electrical characteristic semantic maximum value; A load type query electrical characteristic timing semantic representation generation unit is used to perform a linear transformation on the load type embedded coding vector to obtain a load type query vector and a load type value vector, and use a subsequence of the load electrical characteristic timing response associated feature vector as a subsequence of the key vector, perform converter-based cross-domain query encoding on the load type query vector, the load type value vector and the subsequence of the key vector to obtain a load type query electrical characteristic timing semantic representation vector as the load type query electrical characteristic timing semantic representation.

6. The solid-state power controller fault diagnosis system based on data analysis according to claim 5, characterized in that: The load type-load electrical characteristic semantic maximum value calculation unit is used to: Calculating the mutual information between the load type embedding code vector and each load electrical characteristic timing response associated feature vector in the sequence of the load electrical characteristic timing response associated feature vector to obtain a sequence of load type-load electrical characteristic semantic fast semantic matching factors; The first load type-load electrical characteristic semantic maximum value and the second load type-load electrical characteristic semantic maximum value are identified from the sequence of the load type-load electrical characteristic semantic fast semantic matching factors.

7. The solid-state power controller fault diagnosis system based on data analysis according to claim 6, characterized in that: The load electrical characteristic fast matching unit is used to determine a subsequence of the load electrical characteristic timing response associated feature vectors that are quickly matched from the sequence of the load electrical characteristic timing response associated feature vectors based on the positions of the first load type-load electrical characteristic semantic maximum value and the second load type-load electrical characteristic semantic maximum value in the sequence of the load type-load electrical characteristic semantic fast semantic matching factors.

8. The solid-state power controller fault diagnosis system based on data analysis according to claim 7, characterized in that: The prediction result generating module is used for: The load type query electrical characteristic time series semantic representation vector is passed through a decoder-based closed impulse voltage predictor to obtain the prediction result, wherein the prediction result is used to represent a predicted decoded value of the impulse voltage; Based on the comparison between the predicted decoded value of the impulse voltage and a preset threshold, it is determined whether to generate a fault warning prompt.

9. A solid-state power controller fault diagnosis method based on data analysis, characterized in that: include: A time set for collecting load electrical characteristic data of the solid-state power controller, wherein the load electrical characteristic data includes a load current value and a load voltage value; Grouping the time set of the load electrical characteristic data and performing electrical characteristic time series encoding to obtain a sequence of load current time series characteristics and a sequence of load voltage time series characteristics; Performing response association coding on each group of corresponding load current timing characteristics and load voltage timing characteristics in the sequence of load current timing characteristics and the sequence of load voltage timing characteristics to obtain a sequence of load electrical characteristic timing response association characteristics; Get the type tag of the payload; Embedding and encoding the type label of the load to obtain a load type embedding feature; Performing a fast scanning semantic matching query on the sequence of the load electrical characteristic timing response associated features and the load type embedded features to obtain a load type query electrical characteristic timing semantic representation; The electrical characteristic time series semantic representation is queried based on the load type to obtain a prediction result, and based on the prediction result, it is determined whether to generate a fault warning prompt.

10. The solid-state power controller fault diagnosis method based on data analysis according to claim 9, characterized in that: The time set of the load electrical characteristic data is grouped and the electrical characteristic time sequence is encoded to obtain a sequence of load current time sequence characteristics and a sequence of load voltage time sequence characteristics, including: Using a data grouper to group the time set of the load electrical characteristic data according to the load electrical characteristic parameter sample dimension to obtain a load current timing vector and a load voltage timing vector; The load current timing vector and the load voltage timing vector are respectively passed through an electrical characteristic timing encoder based on a 1D-FCN model to obtain a sequence of load current timing feature vectors as a sequence of load current timing features and a sequence of load voltage timing feature vectors as a sequence of load voltage timing features.

Images