Charging equipment fault prediction method and device, equipment, storage medium and product

Abstract

The invention provides a charging equipment fault prediction method and device, equipment, a storage medium and a product. The method comprises the following steps: performing feature extraction on current data and temperature data of charging equipment to obtain corresponding feature data; inputting the feature data into a quantum model, and mapping the feature data into corresponding quantum state data; the quantum state data is formed by encoding the feature data and is represented by a superposition state and an entanglement state of a plurality of quantum bits; according to a preset feature importance weight, adjusting a quantum gate in the quantum model to extract a feature vector of the quantum state data; the quantum gate is used for transforming quantum state data and adjusting a transformation rule of probability amplitude distribution of the quantum state data; and determining the occurrence probability corresponding to each fault type of the charging equipment based on the feature vector. Based on the fault probability generated by the feature vector, the recognition accuracy of the composite fault is remarkably improved.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method, apparatus, device, storage medium and product for predicting faults in charging equipment. Background Technology

[0002] With the rapid development of the electric vehicle industry and the large-scale deployment of charging infrastructure, public charging stations have become an important part of urban energy networks. Currently, charging systems generally use IoT technology to remotely monitor the charging process and handle payment. However, in complex operating environments, charging equipment may experience problems such as voltage fluctuations, hardware aging, and communication anomalies, leading to abnormal charging interruptions and affecting the reliability of the charging process.

[0003] In existing technologies, data difference analysis between vehicles and charging stations is typically relied upon for centralized cloud processing. While this can achieve a certain degree of fault identification, the long cloud processing link results in end-to-end latency that cannot meet the millisecond-level real-time control requirements of fast charging scenarios. Furthermore, existing solutions heavily rely on communication with the cloud, leading to a high false alarm rate, insufficient fault identification capability, and a low fault identification rate when communication is abnormal. Summary of the Invention

[0004] This application provides a method, apparatus, device, storage medium, and product for predicting faults in charging equipment, aiming to solve the technical problem of low fault identification rate in related technologies.

[0005] In a first aspect, this application provides a method for predicting faults in charging equipment, the method comprising: Feature extraction is performed on the current and temperature data of the charging device to obtain the corresponding feature data; The feature data is input into the quantum model, and the feature data is mapped to the corresponding quantum state data; the quantum state data is encoded from the feature data and is represented by the superposition and entanglement states of multiple qubits. According to the preset feature importance weights, the quantum gates in the quantum model are adjusted to extract the feature vectors of the quantum state data; the quantum gates are transformations of the quantum state data, adjusting the transformation rules of the probability amplitude distribution of the quantum state data; Based on the feature vector, the probability of occurrence of each fault type of the charging device is determined.

[0006] In some possible implementations, the feature extraction of the current and temperature data of the charging device to obtain corresponding feature data includes: Based on the current data of the charging equipment, the corresponding current effective value characteristics, harmonic distortion rate characteristics, and pulse density characteristics are calculated. Based on the temperature data of preset temperature measurement points in the temperature data of the charging device, the corresponding temperature characteristics are obtained. The characteristic data are obtained by combining the current RMS value characteristics, the harmonic distortion rate characteristics, the pulse density characteristics, and the temperature characteristics.

[0007] In some possible implementations, the calculation of the corresponding current RMS characteristics, harmonic distortion rate characteristics, and pulse density characteristics based on the current data of the charging device includes: Based on the current data of the charging device, the root mean square value of the current is calculated to obtain the characteristics of the effective value of the current. Based on the current data of the charging device, the proportion of harmonic components in the current waveform is calculated to obtain the harmonic distortion rate characteristics. Based on the current data of the charging device, the number of times the current value exceeds a preset threshold is calculated to obtain the pulse density characteristics.

[0008] In some possible implementations, before performing feature extraction on the current and temperature data of the charging device to obtain the corresponding feature data, the method further includes: Collect raw current and temperature image sequences of the charging device within a preset time window; The original current sequence is filtered to obtain the current data of the charging device. The temperature image sequence is spatiotemporally aligned to obtain the temperature data of the charging device.

[0009] In some possible implementations, after determining the occurrence probability of each fault type of the charging device based on the feature vector, the method further includes: The quantum prediction score of the charging device is determined based on the probability of occurrence of each fault type of the charging device. The comprehensive safety score of the charging device is obtained based on the quantum prediction score, the current data, and the temperature data. Based on the comprehensive safety score, the charging power of the charging device is adjusted.

[0010] In some possible implementations, after obtaining the comprehensive safety score of the charging device based on the quantum prediction score, the current data, and the temperature data, the method further includes: Based on the comprehensive safety score, a display signal is sent to the user terminal so that the user terminal displays visual content corresponding to the charging device. Upon receiving a control command from the user terminal, the corresponding function is executed according to the control command.

[0011] Secondly, this application provides a quantum model training method, the method comprising: Obtain a training sample set; the training sample set includes multiple training samples, each of which includes historical operating data of the sample charging device and a corresponding fault category label; the historical operating data includes current data and temperature data of the sample charging device during operation; For each training sample, perform the following steps: Calculate the feature importance of each feature in the historical data to obtain the feature importance vector; Feature extraction is performed on the historical operation data to obtain the corresponding historical feature data; The historical feature data is input into the initial quantum model, and the historical feature data is mapped to the corresponding quantum state historical data; Based on the feature importance vector, the quantum gates in the quantum model are adjusted to extract the historical feature vector of the quantum state history data; Based on the historical feature vectors, the corresponding fault category prediction results are determined; Based on the fault category prediction results and the fault category labels, the loss function value of the initial quantum model is determined; Determine whether the loss function value of the initial quantum model meets the preset training stopping condition; If the conditions are not met, the model parameters of the initial quantum model are adjusted, and the adjusted initial quantum model is trained using the training sample set until the preset training stop condition is met, thus obtaining the trained quantum model.

[0012] In some possible implementations, adjusting the model parameters of the initial quantum model when the conditions are not met includes: If the conditions are not met, calculate the quantum Fisher information matrix corresponding to the model parameters of the current initial quantum model; The model parameters of the initial quantum model are adjusted based on the quantum Fisher information matrix and the gradient of the preset loss function.

[0013] In some possible implementations, the step of calculating the quantum Fisher information matrix corresponding to the model parameters of the current initial quantum model when the condition is not met includes: If the conditions are not met, determine the corresponding parameter quantum state of the model parameters of the current initial quantum model; Calculate the matrix elements corresponding to each of the quantum states of the parameters according to the parameter translation rule; By combining the elements of each matrix, the quantum Fisher information matrix is ​​obtained.

[0014] Thirdly, this application provides a charging equipment fault prediction device, the device comprising: The extraction module is used to extract features from the current and temperature data of the charging device to obtain the corresponding feature data. The mapping module is used to input the feature data into the quantum model and map the feature data into corresponding quantum state data; the quantum state data is encoded from the feature data and is represented by the superposition and entanglement states of multiple qubits. An adjustment module is used to adjust the quantum gates in the quantum model according to preset feature importance weights in order to extract the feature vectors of the quantum state data; the quantum gates are transformation rules for the quantum state data, which are used to adjust the probability amplitude distribution of the quantum state data. The determination module is used to determine the probability of occurrence of each fault type of the charging device based on the feature vector.

[0015] Fourthly, this application provides a charging device fault prediction device, the device comprising: a processor, and a memory storing computer program instructions; the processor reads and executes the computer program instructions to implement the charging device fault prediction method as described above.

[0016] Fifthly, this application provides a computer-readable storage medium storing computer program instructions, which, when executed by a processor, implement the charging device fault prediction method described above.

[0017] Sixthly, this application provides a computer program product in which instructions, when executed by a processor of an electronic device, cause the electronic device to perform the charging device fault prediction method described above.

[0018] The charging device fault prediction method, apparatus, device, storage medium, and product provided in this application provide a comprehensive and high-quality input foundation for the model by extracting features from current and temperature data; by mapping feature data to quantum states, quantum parallelism is utilized to achieve efficient representation of complex data patterns; furthermore, the quantum gate is dynamically adjusted according to the preset feature importance weights, enabling the model to adaptively focus on key fault symptoms, thereby extracting quantum feature vectors with higher information density and stronger discriminative power from the quantum states; finally, the fault probability generated based on this feature vector significantly improves the accuracy of identifying complex faults. Attached Figure Description

[0019] This application can be better understood from the following description of specific embodiments in conjunction with the accompanying drawings, wherein: Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings, wherein the same or similar reference numerals denote the same or similar features.

[0020] Figure 1 This is a flowchart of a charging device fault prediction method provided in one embodiment of this application; Figure 2 This is a flowchart of a charging device fault prediction method provided in another embodiment of this application; Figure 3 This is a flowchart of a quantum model training method provided in one embodiment of this application; Figure 4 This is a flowchart illustrating the specific process of augmented reality interaction with a user, provided in one embodiment of this application. Figure 5 This is a schematic diagram of the structure of a charging equipment fault prediction device provided in one embodiment of this application; Figure 6 This is a schematic diagram of the hardware structure of the charging equipment fault prediction device provided in the embodiments of this application. Detailed Implementation

[0021] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0022] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0023] Public charging stations typically connect to a backend management system via the internet to provide electric vehicle charging services. However, the common charging process involves plugging the charging station's plug into the desired charging device, scanning a QR code to access a charging app or mini-program, following the provided instructions, making a prepayment, and starting charging. Charging stops when complete or when the prepayment is insufficient, and the app or mini-program sends a notification that charging is finished. With the rapid increase in the number of electric vehicles, charging station equipment operating in complex environments for extended periods is prone to malfunctions and safety issues.

[0024] During the charging process, voltage fluctuations and hardware malfunctions may occur, causing charging to be interrupted. In such cases, the charging head must be unplugged and plugged back in, and the code must be scanned again to continue charging. These abnormalities greatly affect the user experience.

[0025] Existing technology may lead to incorrect vehicle identification due to reading errors or identification malfunctions, resulting in incorrect or non-initiated charging, which affects normal user operation.

[0026] Machine learning-based solutions have significant limitations. Relying on vehicle-to-charging-pile data difference analysis, they are poorly adaptable to the transient characteristics of high-power fast charging scenarios and cannot support the deployment of older equipment without modification. They also exhibit significant shortcomings in real-time performance, accuracy, and applicability. With the rapid popularization of electric vehicles and smart devices, higher demands are placed on the reliability and intelligence of charging infrastructure. Current charging systems have significant shortcomings in areas such as insufficient fault prediction capabilities, lagging real-time response, unintuitive human-computer interaction, and poor user experience. Traditional models rely on cloud processing, resulting in high latency and security risks. Quantum computing, due to its powerful computing capabilities, is expected to play a crucial role in training complex models. Combining this with edge computing enables real-time reasoning and response, while augmented reality enhances the immersion and operability of user interaction.

[0027] To address the problems of the prior art, embodiments of this application provide a method, apparatus, device, storage medium, and product for predicting charging device faults. The method for predicting charging device faults provided in this application embodiment will be described first below.

[0028] Figure 1 A flowchart illustrating a charging device fault prediction method according to an embodiment of this application is shown. Figure 1 As shown, the method includes the following steps: S101 to S104.

[0029] S101: Extract features from the current and temperature data of the charging device to obtain the corresponding feature data.

[0030] S102: Input the feature data into the quantum model and map the feature data to the corresponding quantum state data. The quantum state data is formed by encoding the feature data and is represented by the superposition and entanglement states of multiple qubits.

[0031] S103: Adjust the quantum gates in the quantum model according to the preset feature importance weights to extract the feature vectors of the quantum state data. The quantum gate is a transformation rule that transforms the quantum state data and adjusts the probability amplitude distribution of the quantum state data.

[0032] S104: Based on the feature vector, determine the probability of occurrence of each fault type of the charging equipment.

[0033] In the specific implementation of S101, real-time current and temperature data are first acquired from the charging device. Current data is typically collected by a current sensor, while temperature data is obtained by a temperature sensor that monitors the internal or external temperature state of the charging device in real time. This data is usually continuous time-series data and may contain noise and fluctuations. For subsequent analysis, feature extraction is required from this raw data to extract meaningful patterns and information from the raw current and temperature data. Feature extraction methods include calculating the average, standard deviation, maximum, minimum values, and rate of change within a time window. These features allow the capture of key characteristics during the charging process. For example, a sudden increase in current may indicate a malfunction in the charging device, while a sharp rise in temperature may indicate overheating.

[0034] In the specific implementation of S102, feature data is passed as input to the quantum model, which then maps the quantum state data. The quantum model is a model based on quantum computing theory, representing data through qubits. Feature data is first encoded into qubits. During encoding, each element of the feature data is mapped to the corresponding state of the qubit using a quantum algorithm. The states of qubits are typically represented through superposition and entanglement, enabling the quantum model to process large amounts of information in parallel and providing a more efficient data representation method.

[0035] In the specific implementation of S103, the importance weights of each feature in the feature data are determined. Each feature may have a different impact on the fault diagnosis results, thus requiring a weight value to be assigned to it. Based on these weights, the quantum gates in the quantum model will be adjusted accordingly. Quantum gates are fundamental operations in quantum computing, through which qubits can be manipulated, such as rotated and flipped. These operations can adjust the probability amplitude of the quantum state, thereby affecting the final result of the quantum state. The adjustment of the quantum gates, by controlling the strength and manner of their action, enables the quantum model to better extract key feature vectors from the quantum state data.

[0036] In the specific implementation of S104, the extracted feature vectors are used to calculate the probability of charging device failure. The feature vectors contain key feature descriptions of the charging device's state. After processing by the quantum model, these features can reflect whether the device has a failure, as well as the type and probability of the failure.

[0037] The charging device fault prediction method provided in this application provides a comprehensive and high-quality input foundation for the model by extracting features from current and temperature data; by mapping feature data to quantum states, it utilizes quantum parallelism to achieve efficient representation of complex data patterns; furthermore, it dynamically adjusts quantum gates according to preset feature importance weights, enabling the model to adaptively focus on key fault symptoms, thereby extracting quantum feature vectors with higher information density and stronger discriminative power from the quantum states; finally, the fault probability generated based on this feature vector significantly improves the accuracy of identifying complex faults.

[0038] To construct suitable feature data, the above S101 may include the following steps: S1011 to S1013.

[0039] S1011: Calculate the corresponding current RMS characteristics, harmonic distortion rate characteristics, and pulse density characteristics based on the current data of the charging device.

[0040] S1012: Based on the temperature data of the preset temperature measurement point in the temperature data of the charging device, obtain the corresponding temperature characteristics.

[0041] S1013: Combine the characteristics of the effective value of the current, the harmonic distortion rate, the pulse density, and the temperature to obtain characteristic data.

[0042] In the specific implementation of S1011, current data is acquired from the charging device. This data is real-time current time-series data collected by a current sensor. The corresponding RMS current characteristics, harmonic distortion rate characteristics, and pulse density characteristics are calculated. The RMS current characteristics reflect the strength of the current signal and are often used to assess the load condition of electrical equipment; the harmonic distortion rate characteristics reflect the harmonic components in the current signal, referring to the degree of nonlinear distortion of different frequency components in the current signal; and the pulse density characteristics represent the periodicity and pulse characteristics of the current waveform.

[0043] In order to accurately calculate the current characteristics, the above S1011 may include the following steps: S10111 to S10113.

[0044] S10111: Calculate the root mean square value of the current based on the current data of the charging device to obtain the effective value characteristics of the current.

[0045] S10112: Based on the current data of the charging device, calculate the proportion of harmonic components in the current waveform to obtain the harmonic distortion rate characteristics.

[0046] S10113: Based on the current data of the charging device, calculate the number of times the current value exceeds the preset threshold to obtain the pulse density characteristics.

[0047] In the specific implementation of S10111, the current data of the charging device is acquired. The current data is usually collected in real time by a current sensor, reflecting the current fluctuations during the charging process. The current value at each sampling time is squared, the average of these squared values ​​is calculated, and finally the square root of the average value is taken.

[0048] In the specific implementation of S10112, the current signal of the charging device is converted into frequency domain data through discrete Fourier transform. The amplitude values ​​of different frequencies in the current waveform can be obtained through Fourier transform. The harmonic distortion rate characteristics are determined by calculating the ratio of the total energy of higher harmonics to the fundamental frequency energy.

[0049] In the specific implementation of S10113, a current threshold is set to determine whether the current exceeds the normal operating range. Then, for the collected current data, the current value at each time point is checked to see if it exceeds the threshold. If the current value exceeds the threshold, it is recorded as a pulse event. By calculating the number of pulses exceeding the threshold per unit time, the pulse density characteristic is obtained.

[0050] The embodiments described above in this application calculate the root mean square value of the current based on the current data of the charging device to obtain the effective value characteristic of the current. Based on the current data of the charging device, the proportion of harmonic components in the current waveform is calculated to obtain the harmonic distortion rate characteristic. Furthermore, based on the current data of the charging device, the number of times the current value exceeds a preset threshold is calculated to obtain the pulse density characteristic, thereby accurately calculating the current characteristics.

[0051] In the specific implementation of S1012, to monitor the operating status of the charging device, multiple temperature measurement points are typically preset inside or outside the device. These measurement points are located in critical parts of the device, such as the rechargeable battery, power module, and transformer. Each measurement point collects real-time temperature data at that location via a temperature sensor. Useful temperature features are extracted from these measurement points. These features may include the current temperature value at each measurement point, the rate of temperature change, and the range of temperature fluctuations. By analyzing these temperature features, the system can identify whether the device is overheating and thus determine if there is a risk of malfunction.

[0052] In the specific implementation of S1013, the features extracted from S1011 and S1012 are combined into a complete feature dataset. Feature combination can employ vector concatenation, where each feature value is treated as part of a data vector, forming a feature dataset with multiple dimensions. These features can also be weighted or normalized to ensure consistent scale across different features, thereby preventing certain features from having an excessively large or small impact on the combined result.

[0053] The above-described embodiments of this application calculate the corresponding current effective value characteristics, harmonic distortion rate characteristics, and pulse density characteristics based on the current data of the charging device. Then, based on the temperature data corresponding to the preset temperature measurement point in the temperature data of the charging device, the corresponding temperature characteristics are obtained. By combining the current effective value characteristics, harmonic distortion rate characteristics, pulse density characteristics, and temperature characteristics, feature data is obtained, thereby constructing suitable feature data.

[0054] To obtain accurate current and temperature data, refer to Figure 2 Before S101, the following steps may be included: S201 to S203.

[0055] S201: Collect the original current sequence and temperature image sequence of the charging device within a preset time window.

[0056] S202: Filter the original current sequence to obtain the current data of the charging device.

[0057] S203: Perform spatiotemporal alignment on the temperature image sequence to obtain the temperature data of the charging device.

[0058] In the specific implementation of S201, data is acquired from the charging device, including current data and temperature image data. Current data is acquired using a current sensor or current probe, while temperature image data is acquired using an infrared temperature sensor or a thermal imaging camera. The raw current sequence consists of current value data within a predetermined time window, which is continuously measured by the sensor and acquired at regular time intervals. The temperature image sequence is a series of temperature images continuously captured by the thermal imaging camera within the time window, showing the temperature distribution of the charging device at different points in time.

[0059] In the specific implementation of S202, the original current sequence is filtered to remove unnecessary noise and obtain more accurate current data. Filtering methods can include low-pass filtering, high-pass filtering, and band-pass filtering. The appropriate filter is selected based on the actual requirements. For example, a low-pass filter can remove high-frequency noise, retaining only low-frequency signals, and can eliminate high-frequency interference in current acquisition.

[0060] In the specific implementation of S203, the acquired temperature image sequence is spatiotemporally aligned. Spatiotemporal alignment involves precisely matching each frame of the image sequence with its corresponding time point, ensuring that the image data reflects the temperature changes of the device at different times and spatial points.

[0061] The above-described embodiments of this application acquire the original current sequence and temperature image sequence of the charging device within a preset time window, then filter the original current sequence to obtain the current data of the charging device, and then perform spatiotemporal alignment of the temperature image sequence to obtain the temperature data of the charging device. By processing the original data, the current data and temperature data are accurately obtained.

[0062] To adjust the charging power, the steps S104 and S1043 may be included after the above steps.

[0063] S1041: Determine the quantum prediction score of the charging equipment based on the occurrence probability of each fault type of the charging equipment.

[0064] S1042: Based on the quantum prediction score, current data, and temperature data, a comprehensive safety score for the charging device is obtained.

[0065] S1043: Adjust the charging power of the charging device based on the comprehensive safety score.

[0066] In the specific implementation of S1041, a quantum prediction score is calculated based on the fault type of the charging device and its corresponding probability of occurrence. The probability of fault occurrence is quantified using a quantum computing model or a probabilistic model to derive a comprehensive evaluation index, which serves as the quantum prediction score for the charging device. Fault type refers to different fault events that may occur during the use of the charging device, such as battery overheating, unstable current, and charging interruption. Each fault has a certain probability of occurrence.

[0067] In the specific implementation of S1042, a comprehensive safety score for the charging device is calculated by combining quantum prediction scores, current data, and temperature data. The safety of the device is comprehensively assessed through the fusion of multi-dimensional data. The quantum prediction score reflects the probability of failure, while the current and temperature data provide real-time monitoring information on the device's current operating status. The current and temperature data are preprocessed, such as through filtering and noise reduction, to ensure that the data accurately reflects the device's current state. Then, the processed current and temperature data are combined with the quantum prediction score to obtain the comprehensive safety score. This data fusion can be achieved using a weighted average method or a multivariate regression model.

[0068] To enable user interaction, the steps following S1042 may also include S10421 to S10422.

[0069] S10421: Based on the comprehensive safety score, a display signal is sent to the user terminal so that the user terminal displays visual content corresponding to the charging device.

[0070] S10422: Upon receiving a control command from the user terminal, execute the corresponding function according to the control command.

[0071] In the specific implementation of S10421, a display signal is sent to the user terminal based on the comprehensive safety score of the charging device, thereby causing the user terminal to display visual content corresponding to the charging device. The safety of the charging device is assessed based on the previously calculated comprehensive safety score. Depending on the score, it is determined what information needs to be sent to the user terminal.

[0072] In the specific implementation of S10422, the corresponding function is executed based on the control command sent by the received user terminal. The user sends the control command through their terminal. The content of the control command may include, but is not limited to, changing the charging power, starting or pausing the charging process, and adjusting the device's operating mode. Specifically, the user terminal inputs the control command through the corresponding user interface. Once the system receives the control command sent by the user terminal, it parses the command and executes the corresponding operation according to the command.

[0073] The above-described embodiments of this application send a display signal to the user terminal based on a comprehensive safety score, so that the user terminal displays visual content corresponding to the charging device. Then, upon receiving a control command sent by the user terminal, the corresponding function is executed according to the control command to realize user interaction.

[0074] In the specific implementation of S1043, the charging power of the charging device is adjusted based on the comprehensive safety score to ensure safety during the charging process. If the comprehensive safety score exceeds a preset safety threshold, it indicates that the device is highly safe and the charging power can be maintained. If the comprehensive safety score is lower than the threshold, it indicates that the device is less safe and the charging power needs to be reduced or charging needs to be paused to avoid malfunctions or accidents.

[0075] The above-described embodiments of this application determine the quantum prediction score of the charging device based on the occurrence probability corresponding to each fault type of the charging device. Then, based on the quantum prediction score, current data, and temperature data, a comprehensive safety score of the charging device is obtained. Based on the comprehensive safety score, the charging power of the charging device is adjusted, and the charging power is adjusted in real time according to the state of the charging device.

[0076] As another embodiment, a quantum model training method is provided, referring to... Figure 3 The quantum model training method includes the following steps: S301: Obtain the training sample set. The training sample set includes multiple training samples, each of which includes historical operating data of the sample charging device and its corresponding fault category label. The historical operating data includes current and temperature data of the sample charging device during operation.

[0077] For each training sample, perform the following steps: S302: Calculate the feature importance of each feature in the historical running data to obtain the feature importance vector.

[0078] S303: Extract features from historical operational data to obtain corresponding historical feature data.

[0079] S304: Input historical feature data into the initial quantum model and map the historical feature data to the corresponding quantum state historical data.

[0080] S305: Adjust the quantum gates in the quantum model based on the feature importance vector to extract the historical feature vector of the quantum state history data.

[0081] S306: Based on historical feature vectors, determine the corresponding fault category prediction results.

[0082] S307: Determine the loss function value of the initial quantum model based on the fault category prediction results and fault category labels.

[0083] S308: Determine whether the loss function value of the initial quantum model meets the preset training stopping condition.

[0084] S309: If the conditions are not met, adjust the model parameters of the initial quantum model and train the adjusted initial quantum model using the training sample set until the preset training stopping condition is met, thus obtaining the trained quantum model.

[0085] In the specific implementation of S301, a training sample set is obtained, which contains multiple training samples. Each training sample includes historical operating data of the sample charging device and its corresponding fault category label. The historical operating data includes current and temperature data of the charging device during operation. This data is collected in real time over a long period of time through the device's sensors and monitoring system, and can reflect the operating characteristics of the charging device under different working conditions.

[0086] In the specific implementation of S302, the feature importance corresponding to each feature is calculated based on different characteristics in the historical operating data. Specifically, a feature selection algorithm is used to evaluate each feature and generate a feature importance vector.

[0087] In the specific implementation of S303, a sliding window method is used to extract data features (such as mean, standard deviation, maximum, minimum, volatility, etc.) over a period of time from the raw current and temperature data. These extracted features will be used as historical feature data for further training of the quantum model. By extracting features from the historical operating data of each training sample, historical feature data related to fault prediction is extracted.

[0088] In the specific implementation of S304, historical feature data is input into the initial quantum model and mapped to corresponding quantum state historical data. The input historical feature data is encoded through quantum gate operations to generate qubits, which are then mapped to quantum state historical data. These quantum states will be used for subsequent model training and feature extraction.

[0089] In the specific implementation of S305, the quantum gates in the quantum model are adjusted based on the previously calculated feature importance vector. Each quantum gate transforms a qubit, thereby affecting the evolution of the quantum state. For example, if the current feature is of high importance, quantum gate operations on the current data are added to the quantum model to strengthen the representation of this feature in the quantum state. Similarly, if the temperature feature is of low importance, the system can reduce the number of quantum gate operations on the temperature feature. Through this adjustment, more discriminative historical quantum state data can be extracted more precisely, and the model's ability to learn important features can be effectively enhanced.

[0090] In the specific implementation of S306, fault category prediction is performed based on the adjusted historical feature vector. The historical feature vector is data obtained after processing through a quantum model, containing the equipment's operating status and possible fault modes. A classification algorithm is used to classify the historical feature vector to predict the equipment's fault category.

[0091] In the specific implementation of S307, the loss function value of the initial quantum model is calculated based on the obtained fault category prediction results and the actual fault category labels. The loss function is a standard used to measure the model's prediction error; the smaller the value, the more accurate the model prediction. The loss function can take the form of cross-entropy loss function, mean squared error loss function, etc. For example, if the prediction result is "overload" while the actual fault label is "normal," the value of the loss function will be high, indicating a larger prediction error.

[0092] In the specific implementation of S308, it checks whether the loss function value of the initial quantum model meets the preset training stopping condition. The training stopping condition can be based on the change in the loss function value or the setting of the training epochs. If the loss function value reaches a certain predetermined threshold, or if the improvement of the loss function after multiple training epochs is less than the preset tolerance, training will stop, and the model is considered to have converged.

[0093] In the specific implementation of S309, the charging equipment fault prediction method provided in this application embodiment enables the model to adaptively focus on key fault symptoms, thereby extracting quantum feature vectors with higher information density and stronger discriminative power from the quantum state; ultimately, the fault probability generated based on this feature vector significantly improves the accuracy of identifying complex faults. For example, if the loss function value does not converge, the parameters in the quantum model can be adjusted by gradient descent or other optimization algorithms, and the loss function can be re-evaluated after each round of training until the loss function reaches a suitable level, thus obtaining a trained quantum model.

[0094] In order to make reasonable parameter adjustments, the above S309 may include the following steps: S3091 to S3092.

[0095] S3091: If the conditions are not met, calculate the quantum Fisher information matrix corresponding to the model parameters of the current initial quantum model.

[0096] S3092: Adjust the model parameters of the initial quantum model based on the quantum Fisher information matrix and the preset loss function gradient.

[0097] In the specific implementation of S3091, the quantum Fisher information matrix corresponding to the model parameters of the current initial quantum model is calculated. First, the density matrix, i.e., the state representation of the quantum system under given parameters, is calculated from the parameterized quantum states of the quantum model. Then, the influence of each model parameter on the quantum state change is calculated. The derivative of the quantum state with respect to each model parameter is calculated; this derivative is used to construct Fisher information, quantifying the degree of influence of the parameter on the quantum state measurement results.

[0098] In order to construct the quantum Fisher information matrix in a reasonable manner, the above S3091 may include the following steps: S30911 to S30913.

[0099] S30911: If the conditions are not met, determine the corresponding parameter quantum state of the model parameters of the current initial quantum model.

[0100] S30912: Calculate the matrix elements corresponding to each parameter quantum state according to the parameter translation rule.

[0101] S30913: Combine the elements of each matrix to obtain the quantum Fisher information matrix.

[0102] In a specific implementation of S30911, a series of quantum operations (such as rotation gates, measurements, etc.) are applied to the quantum system, and the current state of the quantum system is obtained based on these operations. This quantum state can be described by mathematical expressions of quantum mechanics (such as density matrices or state vectors). For example, in the qubit model, given a rotation angle parameter, the corresponding quantum state representation can be obtained through the rotation matrix.

[0103] In the specific implementation of S30912, the gradient of the quantum state with respect to each model parameter first needs to be calculated. This can be obtained by differentiating the wave function or density matrix of the quantum model. During the calculation, the translation rule is usually used to describe how to calculate the direction and magnitude of the quantum state change through small perturbations of the model parameters. Based on these calculation results, we can obtain the quantum state's response to changes in each model parameter, ultimately forming the corresponding matrix elements.

[0104] In the specific implementation of S30913, the matrix elements obtained in the previous step are combined to finally construct the quantum Fisher information matrix. The quantum Fisher information matrix is ​​a symmetric matrix that quantifies the estimation accuracy of each parameter in the quantum model, reflecting the degree of influence of each parameter change on the quantum state observation results. Specifically, the matrix elements of all individual parameters calculated in step S30912 are aggregated together, and the entire matrix is ​​constructed according to the definition rules of the Fisher information matrix. Each element of this matrix typically represents the degree of mutual influence between two parameters; that is, the diagonal elements in the matrix represent the influence of a single parameter, while the off-diagonal elements represent the correlation between parameters.

[0105] The embodiments described above in this application calculate the quantum Fisher information matrix through analysis of the quantum model, thereby providing important information for subsequent quantum model optimization. By determining the parameter quantum states, calculating the matrix elements corresponding to the parameter quantum states, and combining them into the final quantum Fisher information matrix, the quantum Fisher information matrix is ​​constructed in a reasonable manner.

[0106] In the specific implementation of S3092, the model parameters of the initial quantum model are adjusted based on the quantum Fisher information matrix and the gradient of a preset loss function. The sensitivity of each model parameter is calculated using the quantum Fisher information matrix; these sensitivities reflect the magnitude of quantum state changes when each parameter is adjusted in the parameter space. Then, using the preset loss function gradient and the information in the quantum Fisher information matrix, the model parameters are adjusted through gradient descent or a similar optimization algorithm.

[0107] The embodiments described above in this application evaluate the current quantum model using a quantum Fisher information matrix to understand the degree of influence of parameters on the quantum state, and then optimize the model parameters by combining the gradient of the loss function. Through this method, the parameters of the quantum model can be gradually adjusted during the optimization process, thereby improving its performance and ultimately achieving the expected results, thus enabling reasonable parameter adjustment.

[0108] In one embodiment of this application, a quantum hybrid model is first trained, and the input data for the training is the historical operation data matrix X∈R of the charging device. N×D This includes specific data values ​​for current, voltage, and temperature; as well as the charging protocol type, connector status, and the specific charging stage. Correspondingly, there is tag data indicating the fault category.

[0109] Next, outlier handling is performed based on the 3σ principle, discarding data points that exceed the mean by more than three times the standard deviation, such as those with current >150A or <10A. Then, time series interpolation is performed, and missing values ​​are imputed using linear interpolation. Finally, data normalization and encoding are performed, specifically using max-min normalization.

[0110] Next, a decision tree model is used to calculate feature importance. In the decision tree model, the number of decision trees T=100, the maximum depth=6, the learning rate=0.1, and the objective function of the model is weighted cross-entropy loss. Feature importance score vectors are generated by the decision tree model and used for quantum circuit design.

[0111] In quantum circuit design, the number of qubits n is Here, D represents the total dimension of the input features, i.e., the length of the feature vector input into the quantum model after data preprocessing and feature encoding. For example, when D=10, n=4 qubits. Then, a single-qubit rotation gate is used to map the normalized feature values ​​to quantum states. Next, an adaptive entanglement layer dynamically generates (controlled non-NOT gate, CNOT) gate connections, followed by random initialization of the secondary rotation gate and optimization through training.

[0112] Specifically, in the implementation of the adaptive entanglement layer, during the dynamic CNOT connection process, a complete graph corresponding to the qubits is first constructed. Then, the 2n edges with the highest weights in the complete graph are selected (e.g., 8 edges when n=4), and a CNOT gate is added to each edge, with the control bit being the node with the higher weight. Afterwards, quantum natural gradient descent optimization is performed. Specifically, the Quantum Fisher Information Matrix (QFIM) is calculated using the parameter translation rule. The formula for calculating the elements of the QFIM matrix is: in, Let represent the element in the i-th row and j-th column of the quantum Fisher information matrix; Re represents taking the real part of the complex number; | It is the inner product of the partial derivatives of two quantum states; | | It is a correction term used to ensure that the metric remains unchanged under a global phase transformation.

[0113] When updating parameters, the specific steps are as follows: in, F represents the learning rate; + This represents the pseudo-inverse of QFIM, which is calculated through singular value decomposition; This represents the gradient of the loss function.

[0114] Different weights are assigned to different samples for updating. During the calculation, the loss function formula is: Where L(θ) represents the loss function value; N represents the total number of training samples; and K represents the total number of fault categories; This represents the weight of the i-th training sample; Indicate whether i is a fault; This is the flag state corresponding to the fault.

[0115] During the model usage process, data is collected and then current signal processing is performed. A sliding window is used to perform low-pass filtering on the current data to remove high-frequency noise. Within each window, the root mean square (RMS) value of the current is calculated, reflecting the average current intensity, thus obtaining the effective current value. The proportion of harmonic components in the current waveform is analyzed to detect circuit anomalies, yielding the harmonic distortion rate. The number of pulses exceeding the threshold within each window is counted to capture instantaneous overcurrent phenomena, obtaining the transient pulse density.

[0116] Temperature data from key locations such as the charging gun terminals and cable connectors were acquired using an infrared thermal imager at a frequency of 30 frames per second. Spatiotemporal and temporal alignment were then performed. Since the current data sampling rate is much higher than the temperature data, the temperature data needed to be upsampled. A cubic spline interpolation algorithm was used to interpolate the temperature data to a 1kHz timestamp to ensure synchronization with the current data. The spatial relationship between the thermal imager and the charging equipment was calibrated using a calibration board. Perspective transformation was used to map the temperature measurement points in the thermal imager images to physical coordinates, eliminating viewing angle errors. Temperature values ​​were extracted from seven key locations, including the charging gun terminals and cable connectors, as input to the model.

[0117] Based on the input data described above, normalization is performed to obtain a 10-dimensional feature vector. A 10-qubit circuit is simulated, and quantum gate operations are applied to update the CNOT connection topology in real time according to the issued feature importance weights. Based on the model output, the first 128 probability amplitudes are extracted as quantum feature vectors. Based on the quantum feature vectors, a normalized exponential function (Softmax) is applied to obtain the fault probability.

[0118] During use, a weekly update mechanism distributes the new weight matrix trained in the cloud to edge devices. If the update fails, the previously cached weights are automatically used to ensure the continuous operation of the scoring system.

[0119] In addition, current stability and temperature change rate scores are considered. The current stability score is calculated by collecting a sequence of effective current values ​​within a 1-second time window, preprocessing the data, and then using a sliding window to calculate the standard deviation of the effective current values, mapping this standard deviation to the stability score. The temperature sequence within the time window is collected, preprocessed, and then fitted using the least squares method to determine the slope of the temperature trend, representing the temperature change rate. A comprehensive safety score is then calculated by weighted summation of these scores. The weights can be dynamically adjusted based on actual conditions.

[0120] Based on the final safety score and its numerical range, different control commands and user feedback are precisely mapped. Specifically, it is divided into three core intervals: when S≥0.85, the system performs full-power continuous charging and informs the user of the normal status through a solid green LED and an AR interface safety indicator; when 0.6≤S<0.85, the system triggers current limiting to 80% of the rated current, simultaneously pushes a notification to the user's APP, and provides a warning locally through a flashing yellow LED and voice prompts; when S<0.6, the system immediately performs an emergency stop, disconnects the charging circuit, activates an audible and visual alarm, and automatically pushes a maintenance work order to the maintenance system. To ensure smooth control, power adjustment is gradual, and current limiting is completed slowly within 200 milliseconds to avoid sudden current surges damaging the hardware. To ensure reliable command issuance, control commands are sent synchronously via a dual-channel CAN bus and RS-485, forming redundant communication.

[0121] Secondly, at the hardware control implementation level, the above logic is executed through specific power regulation units and user feedback devices. For full-power continuous charging, the digital signal processor sets the PWM duty cycle to 100% and activates current closed-loop control, causing the current to rise to the rated value within 50 milliseconds. For current-limiting mode, the system dynamically adjusts the parameters of the PI controller in the DSP, including reducing the proportional coefficient Kp by 20% to weaken the response intensity and setting the current loop reference value to 80% of the rated value. During this process, the system monitors the IGBT junction temperature in real time; if it exceeds 125℃, it automatically enters derating mode, and the current is reduced by another 20%. For emergency stop, the system uses a mechanical relay to cut off the main circuit and a solid-state relay to shut down the auxiliary power supply, while storing the current / voltage waveform 5 seconds before the stop for post-event analysis. The user feedback device uses a WS2812B smart LED, displaying specific colors and flashing frequencies according to different states. The audible and visual alarm ensures a sound pressure level ≥85dB at 1 meter, and the red LED flashing duty cycle is limited to 20% to meet safety standards.

[0122] At the software control logic level, a state switching debouncing mechanism is introduced to prevent malfunctions caused by score fluctuations. This mechanism employs hysteresis switching logic: switching from full power to current limiting requires three consecutive scores S < 0.85 to trigger; returning from current limiting to full power requires five consecutive scores S ≥ 0.85 to release the limit; and switching from current limiting to emergency stop requires an S < 0.6 state for 500ms to confirm danger. After an emergency stop occurs, a manual physical reset is required, and the quantum scoring system must reassess the safety status (S ≥ 0.9) before recovery can proceed, ensuring restart safety. For control signal conflicts, the system sets the emergency stop command as the highest priority, executing it immediately upon receipt and recording the conflict event in the log. For communication interruptions, when the CAN bus times out by 100ms, the system automatically switches to the RS-485 channel to ensure continuous command transmission.

[0123] Furthermore, refer to Figure 4 By using augmented reality technology, the results of quantum diagnostics are presented to users in an intuitive and visual way, enabling efficient and seamless recharging. Specifically, this includes the following steps S401 to S408.

[0124] S401: The quantum inference engine maps the 128-dimensional quantum feature vector generated by edge inference to three-dimensional spatial coordinates. Based on the comprehensive safety score S calculated in real time, it dynamically generates different AR effects and labels on the digital twin model of the charging pile: when S≥0.85, a green particle flow effect (flow rate 0.5m / s) is displayed with the label "System normal, can continue charging"; when 0.6≤S<0.85, a yellow rotating warning sign (speed 1rpm) is displayed with the label "Caution: Current limiting mode (80%)"; when S<0.6, explosion fragments and red arrows are displayed with the label "Danger! Click to view repair plan", thus transforming the abstract risk level into an intuitive visual perception.

[0125] S402: AR instruction generator. Based on the safety score and fault location information output by the quantum inference engine, combined with the preset visualization rule base, it generates a standardized set of AR rendering instructions, including effect types, spatial coordinates, color parameters and interactive hot zone definitions, to realize the structured transformation of quantum data into visualization elements.

[0126] S403: Determines device type, automatically detects the operating system type of the user terminal device, obtains the platform identifier by reading the system API, provides a basis for subsequent selection of native AR rendering engine, and ensures cross-platform compatibility.

[0127] S404: When the system determines that the device is an iOS platform, it uses the ARkit rendering engine to convert AR commands into iOS-optimized 3D rendering commands.

[0128] S405: When the device is identified as an Android platform, it uses the ARcore rendering engine to perform rendering tasks and achieves stable spatial positioning through motion tracking and environment understanding functions.

[0129] S406: Detection / Engine (D / E) continuously monitors the device type status and dynamically allocates rendering tasks to the corresponding rendering pipeline based on the platform's identification results.

[0130] S407: Dynamic AR scene generation. Based on AR commands and real-time rendering engine output, it dynamically constructs a 3D scene that blends virtual and reality. It switches visual elements such as particle flow, warning signs, and fault arrows according to safety scores, and maintains persistent alignment between virtual content and physical objects through spatial calculation.

[0131] S408: Interactive event processor, which captures the user's gesture trajectory and voice stream in real time, and performs feature analysis using dynamic time warping algorithm and Mel frequency cepstral coefficients respectively, and maps the recognition results into system control commands to realize multimodal interactive functions such as gaze triggering and gesture confirmation.

[0132] For gesture control, the system captures user gestures using the phone's camera and performs trajectory recognition using a dynamic time warping algorithm. When the DTW distance is less than 0.2, a corresponding command is triggered, such as the two-finger pinch-to-zoom model. For voice control, the system uses Mel-frequency cepstral coefficients to extract voice features.

[0133] After the user opens the mobile app, the AR interface automatically loads a digital twin model of the charging station and generates a floating 3D "One-Click Recharge" button on the charging gun handle. First, facial recognition is performed for identity verification; second, the payment token is decrypted; finally, a specific recharge command data frame is sent to the charging controller via the CAN bus. This process allows users to resume charging within 2 seconds without needing to scan the code again after an abnormal interruption.

[0134] Based on the charging equipment fault prediction method provided in the above embodiments, this application also provides specific implementation methods of the charging equipment fault prediction device. Please refer to the following embodiments.

[0135] First see Figure 5 The charging equipment fault prediction device 500 provided in this application embodiment includes the following modules: The extraction module 501 is used to extract features from the current data and temperature data of the charging device to obtain the corresponding feature data.

[0136] The mapping module 502 is used to input feature data into the quantum model and map the feature data to corresponding quantum state data. Quantum state data is data encoded from feature data and represented by the superposition and entanglement states of multiple qubits.

[0137] The adjustment module 503 is used to adjust the quantum gates in the quantum model according to preset feature importance weights in order to extract the feature vectors of the quantum state data. The quantum gate is a transformation rule that transforms the quantum state data and adjusts the probability amplitude distribution of the quantum state data.

[0138] The determination module 504 is used to determine the probability of occurrence of each fault type of the charging device based on the feature vector.

[0139] As one implementation of this application, the extraction module 501 includes: The calculation unit is used to calculate the corresponding current effective value characteristics, harmonic distortion rate characteristics, and pulse density characteristics based on the current data of the charging device.

[0140] The determining unit is used to obtain the corresponding temperature characteristics based on the temperature data of the preset temperature measurement point in the temperature data of the charging device.

[0141] The combination unit is used to combine the RMS current characteristics, harmonic distortion rate characteristics, pulse density characteristics, and temperature characteristics to obtain characteristic data.

[0142] As one implementation of this application, the computing unit includes: The calculation subunit is used to calculate the root mean square value of the current based on the current data of the charging device, and obtain the effective value characteristics of the current.

[0143] The calculation subunit is also used to calculate the proportion of harmonic components in the current waveform based on the current data of the charging device, and obtain the harmonic distortion rate characteristics.

[0144] The calculation subunit is also used to calculate the number of times the current value exceeds a preset threshold based on the current data of the charging device, and to obtain the pulse density characteristics.

[0145] As one implementation of this application, the charging equipment fault prediction device 500 further includes: The acquisition module is used to acquire the raw current sequence and temperature image sequence of the charging device within a preset time window.

[0146] The filtering module is used to filter the original current sequence to obtain the current data of the charging device.

[0147] The alignment module is used to perform spatiotemporal alignment on the temperature image sequence to obtain the temperature data of the charging device.

[0148] As one implementation of this application, the charging equipment fault prediction device 500 further includes: The determination module is used to determine the quantum prediction score of the charging equipment based on the occurrence probability corresponding to each fault type of the charging equipment.

[0149] The determination module is also used to obtain a comprehensive safety score for the charging device based on quantum prediction scores, current data, and temperature data.

[0150] The adjustment module is used to adjust the charging power of the charging device based on the comprehensive safety score.

[0151] As one implementation of this application, the charging equipment fault prediction device 500 further includes: The sending module is used to send a display signal to the user terminal based on the comprehensive safety score, so that the user terminal can display the visual content corresponding to the charging device.

[0152] The execution module is used to execute the corresponding function according to the control command sent by the user terminal.

[0153] Each module in the charging equipment fault prediction device provided in this application embodiment can implement each step in the above-mentioned charging equipment fault prediction method and achieve the corresponding effect. For the sake of brevity, it will not be described in detail here.

[0154] Figure 6 A schematic diagram of the structure of the charging device fault prediction hardware provided in an embodiment of this application is shown.

[0155] The charging equipment fault prediction device may include a processor 601 and a memory 602 storing computer program instructions.

[0156] Specifically, the processor 601 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0157] Memory 602 may include mass storage for data or instructions. For example, and not limitingly, memory 602 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 602 may include removable or non-removable (or fixed) media. Where appropriate, memory 602 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, memory 602 is non-volatile solid-state memory.

[0158] The memory may include read-only memory (ROM), random access memory (RAM), disk storage media devices, optical storage media devices, flash memory devices, and electrical, optical, or other physical / tangible memory storage devices. Therefore, typically, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the charging device fault prediction method according to any embodiment of this disclosure.

[0159] The processor 601 reads and executes computer program instructions stored in the memory 602 to implement any of the charging device fault prediction methods in the above embodiments.

[0160] In one example, the charging equipment fault prediction device may further include a communication interface 603 and a bus 610. Wherein, as Figure 6As shown, the processor 601, memory 602, and communication interface 603 are connected through bus 610 and complete communication with each other.

[0161] The communication interface 603 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0162] Bus 610 includes hardware, software, or both, that couples components of an online data traffic metering device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 610 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, any suitable bus or interconnect is contemplated herein.

[0163] Furthermore, in conjunction with the charging device fault prediction method in the above embodiments, this application embodiment can provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the charging device fault prediction methods in the above embodiments.

[0164] This application also provides a computer program product, including a computer program that, when executed by a processor, implements any of the charging device fault prediction methods described in the above embodiments.

[0165] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0166] The functional blocks shown in the above-described structural diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0167] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0168] The aspects of this disclosure have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowchart illustrations, and combinations of blocks in the block diagrams and / or flowchart illustrations, can also be implemented by an FPGA performing the specified functions or actions, or can be implemented by a combination of an FPGA and computer instructions.

[0169] The above description is merely a specific implementation of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A method for predicting faults in charging equipment, characterized in that, The method includes: Feature extraction is performed on the current and temperature data of the charging device to obtain the corresponding feature data; The feature data is input into the quantum model, and the feature data is mapped to the corresponding quantum state data; the quantum state data is encoded from the feature data and is represented by the superposition and entanglement states of multiple qubits. According to the preset feature importance weights, the quantum gates in the quantum model are adjusted to extract the feature vectors of the quantum state data; the quantum gates are transformations of the quantum state data, adjusting the transformation rules of the probability amplitude distribution of the quantum state data; Based on the feature vector, the probability of occurrence of each fault type of the charging device is determined.

2. The charging equipment fault prediction method according to claim 1, characterized in that, The process of extracting features from the current and temperature data of the charging device to obtain corresponding feature data includes: Based on the current data of the charging equipment, the corresponding current effective value characteristics, harmonic distortion rate characteristics, and pulse density characteristics are calculated. Based on the temperature data of preset temperature measurement points in the temperature data of the charging device, the corresponding temperature characteristics are obtained. The characteristic data are obtained by combining the current RMS value characteristics, the harmonic distortion rate characteristics, the pulse density characteristics, and the temperature characteristics.

3. The charging equipment fault prediction method according to claim 2, characterized in that, The calculation of the corresponding current RMS characteristics, harmonic distortion rate characteristics, and pulse density characteristics based on the current data of the charging device includes: Based on the current data of the charging device, the root mean square value of the current is calculated to obtain the characteristics of the effective value of the current. Based on the current data of the charging device, the proportion of harmonic components in the current waveform is calculated to obtain the harmonic distortion rate characteristics. Based on the current data of the charging device, the number of times the current value exceeds a preset threshold is calculated to obtain the pulse density characteristics.

4. The charging equipment fault prediction method according to claim 1, characterized in that, Before performing feature extraction on the current and temperature data of the charging device to obtain the corresponding feature data, the method further includes: Collect raw current and temperature image sequences of the charging device within a preset time window; The original current sequence is filtered to obtain the current data of the charging device. The temperature image sequence is spatiotemporally aligned to obtain the temperature data of the charging device.

5. The charging equipment fault prediction method according to claim 1, characterized in that, After determining the probability of occurrence of each fault type of the charging device based on the feature vector, the method further includes: The quantum prediction score of the charging device is determined based on the probability of occurrence of each fault type of the charging device. The comprehensive safety score of the charging device is obtained based on the quantum prediction score, the current data, and the temperature data. Based on the comprehensive safety score, the charging power of the charging device is adjusted.

6. The charging equipment fault prediction method according to claim 5, characterized in that, After obtaining the comprehensive safety score of the charging device based on the quantum prediction score, the current data, and the temperature data, the method further includes: Based on the comprehensive safety score, a display signal is sent to the user terminal so that the user terminal displays visual content corresponding to the charging device. Upon receiving a control command from the user terminal, the corresponding function is executed according to the control command.

7. A quantum model training method, characterized in that, The method includes: Obtain a training sample set; the training sample set includes multiple training samples, each of which includes historical operating data of the sample charging device and a corresponding fault category label; the historical operating data includes current data and temperature data of the sample charging device during operation; For each training sample, perform the following steps: Calculate the feature importance of each feature in the historical data to obtain the feature importance vector; Feature extraction is performed on the historical operation data to obtain the corresponding historical feature data; The historical feature data is input into the initial quantum model, and the historical feature data is mapped to the corresponding quantum state historical data; Based on the feature importance vector, the quantum gates in the quantum model are adjusted to extract the historical feature vector of the quantum state history data; Based on the historical feature vectors, the corresponding fault category prediction results are determined; Based on the fault category prediction results and the fault category labels, the loss function value of the initial quantum model is determined; Determine whether the loss function value of the initial quantum model meets the preset training stopping condition; If the conditions are not met, the model parameters of the initial quantum model are adjusted, and the adjusted initial quantum model is trained using the training sample set until the preset training stop condition is met, thus obtaining the trained quantum model.

8. The quantum model training method according to claim 7, characterized in that, The step of adjusting the model parameters of the initial quantum model when the conditions are not met includes: If the conditions are not met, calculate the quantum Fisher information matrix corresponding to the model parameters of the current initial quantum model; The model parameters of the initial quantum model are adjusted based on the quantum Fisher information matrix and the gradient of the preset loss function.

9. The quantum model training method according to claim 8, characterized in that, The step of calculating the quantum Fisher information matrix corresponding to the model parameters of the current initial quantum model when the conditions are not met includes: If the conditions are not met, determine the corresponding parameter quantum state of the model parameters of the current initial quantum model; Calculate the matrix elements corresponding to each of the quantum states of the parameters according to the parameter translation rule; By combining the elements of each matrix, the quantum Fisher information matrix is ​​obtained.

10. A fault prediction device for charging equipment, characterized in that, The device includes: The extraction module is used to extract features from the current and temperature data of the charging device to obtain the corresponding feature data. The mapping module is used to input the feature data into the quantum model and map the feature data into corresponding quantum state data; the quantum state data is encoded from the feature data and is represented by the superposition and entanglement states of multiple qubits. An adjustment module is used to adjust the quantum gates in the quantum model according to preset feature importance weights in order to extract the feature vectors of the quantum state data; the quantum gates are transformation rules for the quantum state data, which are used to adjust the probability amplitude distribution of the quantum state data. The determination module is used to determine the probability of occurrence of each fault type of the charging device based on the feature vector.

11. A fault prediction device for charging equipment, characterized in that, The device includes: a processor and a memory storing computer program instructions; the processor reads and executes the computer program instructions to implement the charging device fault prediction method as described in any one of claims 1-6.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions, which, when executed by a processor, implement the charging device fault prediction method as described in any one of claims 1-6.

13. A computer program product, characterized in that, When the instructions in the computer program product are executed by the processor of the electronic device, the electronic device performs the charging device fault prediction method as described in any one of claims 1-6.

Images