AI-based mobile energy storage vehicle operation state monitoring and analysis system

Abstract

The invention relates to the technical field of mobile energy storage vehicles, and discloses an AI-based mobile energy storage vehicle running state monitoring and analysis system which comprises a main controller, an AI monitoring coprocessor and a state analysis coprocessor. The main controller calls a monitoring analysis instruction and sends the monitoring analysis instruction to the state analysis coprocessor; an AI monitoring coprocessor dispatches a state analysis instruction and processes multi-modal data; and the state analysis coprocessor analyzes the parameters to generate a reference monitoring signal, cooperatively monitors the battery temperature, the charging and discharging efficiency and the load fluctuation in real time, and generates an abnormal correction signal and a state control instruction in combination with an analysis optimization mode. The system realizes accurate analysis of multi-source data and dynamic control of the energy storage unit through data preprocessing, feature dimension reduction and deep neural network prediction, improves monitoring accuracy, dynamic adaptability and intelligent level, and ensures safe and efficient operation of the mobile energy storage vehicle.

Description

Technical Field

[0001] The present invention relates to the technical field of mobile energy storage vehicles, and particularly to an operation status monitoring and analysis system for mobile energy storage vehicles based on AI. Background Art

[0002] With the rapid development of new energy technologies, mobile energy storage vehicles, as a flexible energy storage and supply device, have been widely used in fields such as emergency power supply, electric vehicle charging, and renewable energy consumption. However, during operation, mobile energy storage vehicles face complex working environments and changing load demands, and the operating status of their core energy storage units (such as battery packs) directly affects the safety, stability, and service life of the entire system. Currently, the following significant problems exist in traditional mobile energy storage vehicle operation status monitoring and analysis systems:

[0003] During the operation of mobile energy storage vehicles, it is necessary to collect multi-type sensor signals such as voltage, current, and temperature. Traditional systems often adopt a single data processing architecture, making it difficult to effectively process the fusion and analysis of multi-modal data. For example, the sampling frequencies, data formats, and noise characteristics of different sensors vary, resulting in data time series offset and information distortion, and unable to accurately reflect the true operating status of the energy storage vehicle. At the same time, for unstructured data such as thermal imaging signals, traditional methods lack efficient feature extraction and analysis means, making it difficult to achieve precise control of battery thermal management.

[0004] During the charging and discharging process of mobile energy storage vehicles, parameters such as battery temperature, charging and discharging efficiency, and load fluctuations exhibit dynamic change characteristics. Traditional systems usually adopt a monitoring mode with fixed thresholds and are unable to dynamically adjust monitoring strategies and optimize parameters according to real-time operating conditions. For example, when the energy storage vehicle faces extreme conditions such as high load impacts or transient charging and discharging, traditional systems are difficult to respond quickly and perform parameter iteration, easily leading to overcharging, over-discharging, or excessive temperature of the battery, thereby causing safety hazards or shortening the battery life. In addition, for different operating scenarios (such as emergency power supply for urban power grids, power supply for field operations, etc.), traditional systems lack flexible mode adaptation capabilities and cannot achieve the optimal monitoring and analysis effects.

[0005] When traditional systems detect abnormal operating status of energy storage vehicles, they often can only issue simple warning signals, lacking in-depth analysis of the causes of abnormalities and effective correction strategies. For example, when the battery temperature exceeds the safe range, traditional systems may only trigger the cooling fan to start, without being able to perform dynamic compensation and parameter fine-tuning according to the differences in temperature distribution and load fluctuations. At the same time, for complex abnormalities under the coupling action of multiple parameters (such as voltage fluctuations and temperature increases occurring simultaneously), traditional systems are difficult to conduct collaborative analysis and comprehensive processing, resulting in poor abnormal correction effects and affecting the continuous and stable operation of the energy storage vehicle.

[0006] Most existing mobile energy storage vehicle monitoring systems rely on manual parameter setting and empirical judgment, lacking the ability of autonomous learning and prediction based on artificial intelligence. For example, it is impossible to train a model through historical data to predict the state evolution trend of the energy storage unit, and it is difficult to detect potential fault hazards in advance and perform preventive maintenance. In addition, in terms of multi-module collaborative work, the control logic of traditional systems is relatively rigid, and it is impossible to achieve intelligent scheduling and efficient collaboration among the main controller, monitoring coprocessor, and analysis coprocessor, which limits the improvement of the overall system performance. Summary of the Invention

[0007] The purpose of the present invention is to provide a monitoring and analysis system for the operating state of a mobile energy storage vehicle based on AI to solve the problems raised in the above background technology.

[0008] To achieve the above purpose, the present invention provides the following technical solution: A monitoring and analysis system for the operating state of a mobile energy storage vehicle based on AI, the system includes:

[0009] A main controller, an AI monitoring coprocessor, and a state analysis coprocessor;

[0010] The main controller is used to retrieve the corresponding monitoring and analysis instructions according to the operating state data of the energy storage vehicle and send the monitoring and analysis instructions to the state analysis coprocessor;

[0011] The AI monitoring coprocessor is used to schedule the state analysis instructions of the state analysis coprocessor during the execution of the multi-modal data parsing instructions;

[0012] The state analysis coprocessor is used to parse the dynamic monitoring parameters in the monitoring and analysis instructions to generate a reference monitoring signal; perform real-time collaborative monitoring on the battery temperature, charge and discharge efficiency, and load fluctuation, and select the corresponding analysis and optimization mode for dynamic adaptation according to the monitoring results to generate an abnormal correction signal, and combine the reference monitoring signal to generate a state control instruction for the energy storage unit.

[0013] Preferably, the AI monitoring coprocessor includes:

[0014] A data preprocessing module, which is used to filter the noise of multi-source sensor signals and convert the processed sensor signals into standard state parameters as the reference monitoring signal; among them, the sensor signals include voltage signals, current signals, and thermal imaging signals;

[0015] A feature extraction module, which is respectively deployed in the dynamic adaptation module and the abnormal correction module, and is used to perform feature dimensionality reduction on the dynamic adaptation module and the abnormal correction module respectively to extract a first feature set corresponding to the dynamic adaptation module and a second feature set corresponding to the abnormal correction module;

[0016] A prediction model module, which is used to generate a time series prediction signal through a deep neural network according to the first feature set and the second feature set, and generate a state control instruction for the energy storage unit in combination with the reference monitoring signal.

[0017] Preferably, the main controller is further configured to, according to a first target monitoring task, retrieve a corresponding preset monitoring threshold as a first target analysis instruction through the parameter storage unit, and send the first target analysis instruction to the state analysis co-processor.

[0018] Preferably, the main controller is further configured to, according to a second target monitoring task, use a dynamic optimization algorithm based on time series correlation analysis as a second target analysis instruction, and send the second target analysis instruction to the state analysis co-processor.

[0019] Preferably, the analysis and optimization mode includes: bypassing the data verification function, performing global parameter verification, and performing local parameter verification.

[0020] Preferably, the prediction model module includes:

[0021] A dynamic adjustment module, which is used to select an optimization and correction strategy corresponding to the dynamic adaptation module and the anomaly correction module respectively according to the first feature set and the second feature set, perform parameter iteration, and generate the anomaly correction signal; wherein, the optimization and correction strategy includes a load balancing strategy, a temperature suppression strategy, and a charge and discharge smoothing strategy;

[0022] A signal fusion module, which is used to perform spatio-temporal alignment operation on the reference monitoring signal and the anomaly correction signal to generate a final control instruction;

[0023] An instruction issuing module, which is used to transmit the final control instruction to the control terminal of the energy storage unit.

[0024] Preferably, the dynamic adjustment module includes:

[0025] A first adjustment module, which is used to superimpose a temperature compensation signal in the current monitoring period when it is detected that the battery temperature corresponding to the reference monitoring signal exceeds the safe range, and generate a first target adjustment signal;

[0026] A second adjustment module, which is used to perform trend prediction and correction according to the charge and discharge efficiency decay curve, and generate a second target adjustment signal.

[0027] Preferably, the first adjustment module includes:

[0028] A steady-state monitoring module, which is used to determine whether the deviation between the current load fluctuation and the expected load exceeds the tolerance threshold when it is detected that the energy storage unit is in steady-state operation. If it exceeds, perform a parameter fine-tuning on the load balancing strategy. If it does not exceed, maintain the original strategy;

[0029] A transient response module, which is used to judge whether the difference between the thermal imaging signal and the preset temperature distribution exceeds the critical value before the charge and discharge action starts when it is monitored that the energy storage unit is in transient charge and discharge. If it exceeds, a dynamic compensation is performed on the temperature suppression strategy. If it does not exceed, no compensation is performed;

[0030] An abnormal handling module, which is used to dynamically adapt the charge and discharge smoothing strategy in combination with voltage fluctuation data when it is monitored that the energy storage unit is under high load impact, and adjust the smoothing parameters once when the voltage fluctuation exceeds the threshold;

[0031] A single-parameter adjustment module, which is used to locally adapt the correction strategy corresponding to the signal within the monitoring period when it is monitored that a single sensor signal is abnormal, and the remaining signals maintain the original correction strategy.

[0032] Preferably, the second adjustment module includes:

[0033] A prediction configuration module, which is used to activate the trend prediction function for specific working conditions; wherein, the specific working conditions include high load cycle, abnormal temperature gradient and current mutation interval, and each of the working conditions is configured with a prediction parameter register and a prediction action period register;

[0034] A log recording module, which is used to continuously record the parameter evolution data of each prediction node under the specific working conditions after activating the prediction function;

[0035] A non-linear prediction module, which is used to perform a non-linear compensation once according to the evolution data when the prediction target is reached within the specified period of the specific working condition.

[0036] Preferably, the data preprocessing module further includes a timestamp synchronization unit, which is used to perform clock calibration on the sampling frequencies of different sensors to eliminate the timing offset of multi-source data.

[0037] Compared with the prior art, the beneficial effects of the present invention are:

[0038] Through the data preprocessing module of the AI monitoring coprocessor in the system, the multi-source sensor signals are filtered for noise and clock calibrated, eliminating the timing offset of multi-source data, and converting the processed sensor signals into standard state parameters as the reference monitoring signals, ensuring the accuracy and reliability of the data, and providing a solid foundation for subsequent monitoring and analysis. The feature extraction modules are respectively deployed in the dynamic adaptation module and the abnormal correction module, perform feature dimensionality reduction on them, extract the corresponding feature sets, and combined with the deep neural network of the prediction model module, can generate time series prediction signals, realizing the forward-looking prediction of the state of the energy storage unit, discovering potential problems in advance and intervening, effectively improving the safety and stability of the system.

[0039] According to different monitoring tasks, the main controller flexibly retrieves preset monitoring thresholds or dynamic optimization algorithms as analysis instructions and sends them to the state analysis coprocessor, enabling the system to dynamically adjust the monitoring strategy according to different operating conditions and monitoring requirements, improving the adaptability and flexibility of the system. The state analysis coprocessor parses the dynamic monitoring parameters to generate a reference monitoring signal, performs real-time collaborative monitoring on the battery temperature, charge-discharge efficiency, and load fluctuation, and selects the corresponding analysis and optimization modes (including data verification function bypass, global parameter verification, and local parameter verification) according to the monitoring results for dynamic adaptation, generates an abnormal correction signal, and combines it with the reference monitoring signal to generate a state control instruction, achieving precise control and dynamic optimization of the operating state of the energy storage unit, effectively improving the charge-discharge efficiency and service life of the energy storage unit.

[0040] The dynamic adjustment module of the prediction model module performs parameter iteration according to the feature set to select optimization and correction strategies (load balancing strategy, temperature suppression strategy, and charge-discharge smoothing strategy) and generates an abnormal correction signal. Among them, the first adjustment module takes corresponding adjustment measures for different operating conditions such as the battery temperature exceeding the safe range, the energy storage unit being in steady-state operation, transient charge-discharge, high-load impact, and single-sensor signal abnormality, achieving fine processing and precise control of various complex operating conditions; the second adjustment module activates the trend prediction function for specific operating conditions (high-load cycle, abnormal temperature gradient, and current mutation interval), records the parameter evolution data and performs non-linear compensation, further improving the system's prediction and adaptation capabilities for dynamic operating conditions. The signal fusion module performs spatio-temporal alignment operation on the reference monitoring signal and the abnormal correction signal to generate a final control instruction, and the instruction issuing module transmits it to the control terminal of the energy storage unit, ensuring the accuracy and timeliness of the control instruction and achieving efficient control of the energy storage unit.

[0041] Through the application of artificial intelligence technology, the entire system realizes the intelligent parsing, feature extraction, and prediction analysis of multi-modal data, as well as the collaborative work and intelligent scheduling among various modules, significantly improving the intelligent level of the operation state monitoring and analysis of the mobile energy storage vehicle, providing a strong guarantee for the safe, stable, and efficient operation of the mobile energy storage vehicle, and having significant economic and social benefits. Brief Description of the Drawings

[0042] Figure 1 It is the working principle diagram of the AI-based mobile energy storage vehicle operation state monitoring and analysis system described in the present invention;

[0043] Figure 2 It is the design diagram of the AI monitoring coprocessor module;

[0044] Figure 3 It is the working flowchart of the state analysis coprocessor;

[0045] Figure 4 Design diagram for dynamically adjusting module logic;

[0046] Figure 5 Design diagram for adjusting strategy of prediction model. Specific implementation manners

[0047] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0048] Please refer to Figures 1-5 , the AI-based mobile energy storage vehicle operation status monitoring and analysis system involved in the present invention includes a main controller, an AI monitoring coprocessor, and a status analysis coprocessor. The specific steps are as follows:

[0049] The main controller is configured to retrieve corresponding monitoring and analysis instructions according to the operation status data of the energy storage vehicle, and send the monitoring and analysis instructions to the status analysis coprocessor. In this process, as the core control hub of the system, the main controller real-time obtains various status data during the operation of the energy storage vehicle, such as basic information such as the voltage, current, temperature, charge and discharge efficiency, and load change of the battery, and based on preset logic rules or algorithms, determines the type of monitoring and analysis tasks to be executed currently, and then retrieves the corresponding monitoring and analysis instructions from the internal storage unit or external interaction module. These instructions may include content such as monitoring parameter settings, analysis algorithm selection, and control strategy invocation for different operation scenarios. The main controller accurately transmits the instructions to the status analysis coprocessor through a standardized communication interface and protocol to trigger the subsequent analysis and processing process.

[0050] The AI monitoring coprocessor is used to schedule the status analysis instructions of the status analysis coprocessor during the execution of the multi-modal data parsing instructions. This coprocessor has the ability to parse and process multi-source heterogeneous data, and can receive data of different modalities such as voltage signals, current signals, and thermal imaging signals from multiple sensors, and perform preliminary preprocessing and parsing on these data. During the parsing process, the AI monitoring coprocessor identifies the valid information in the data according to the requirements of the multi-modal data parsing instructions, and at the same time determines whether it is necessary to invoke the function of the status analysis coprocessor. For example, when abnormal features or trends that require in-depth analysis are detected in the data, the AI monitoring coprocessor will actively schedule the status analysis instructions of the status analysis coprocessor, and transfer the preprocessed key data or characteristic parameters to the status analysis coprocessor to achieve more in-depth analysis and processing.

[0051] The status analysis coprocessor is used to parse the dynamic monitoring parameters in the monitoring and analysis instructions, generate a reference monitoring signal; perform real-time collaborative monitoring on the battery temperature, charge and discharge efficiency, and load fluctuation, and select the corresponding analysis and optimization mode for dynamic adaptation according to the monitoring results, generate an anomaly correction signal, and combine it with the reference monitoring signal to generate a status control instruction for the energy storage unit. Specifically, the status analysis coprocessor first parses the monitoring and analysis instructions sent by the main controller, extracts the dynamic monitoring parameters contained therein, such as temperature thresholds, charge and discharge efficiency standards, allowable range of load fluctuations, etc., and generates a reference monitoring signal for subsequent monitoring and control based on these parameters. At the same time, by collecting the data of devices such as battery temperature sensors, charge and discharge efficiency monitoring modules, and load sensors in real time, the three key indicators are synchronously monitored. During the monitoring process, the system continuously compares the difference between the actual monitoring data and the reference monitoring signal. When an abnormal situation is detected (such as the temperature exceeding the safety threshold, a significant decrease in charge and discharge efficiency, or severe load fluctuation), according to the preset rules or AI algorithms, a suitable mode (such as data verification function bypass, global parameter verification, local parameter verification) is selected from the set of analysis and optimization modes to dynamically adjust the monitoring data or control strategy, and an anomaly correction signal is generated. Finally, the reference monitoring signal and the anomaly correction signal are fused to form a status control instruction for the energy storage unit (such as battery packs, charge and discharge modules, etc.), and the operating status of the energy storage vehicle is adjusted in real time through the actuator.

[0052] Embodiment 1:

[0053] In this embodiment, the AI monitoring coprocessor includes a data preprocessing module, a feature extraction module, and a prediction model module. The data preprocessing module filters the noise of multi-source sensor signals (including voltage signals, current signals, and thermal imaging signals). Specifically, digital filtering algorithms (such as Kalman filtering and Butterworth filtering) are used to remove high-frequency noise and low-frequency drift in the signals to ensure the accuracy and stability of the sensor signals. Subsequently, the processed sensor signals are converted into standard state parameters (such as standardized voltage values, current values, temperature values, etc.) as the reference monitoring signal.

[0054] The data preprocessing module first receives the original signals from different types of sensors. The voltage signal is usually collected by a high-precision voltage sensor, which reflects the change in the potential difference of the battery pack or other electrical components of the energy storage vehicle. The current signal is collected by a Hall effect sensor or a shunt resistor to monitor the current intensity in the circuit. The thermal imaging signal is obtained by an infrared thermal imager, which can provide an image of the temperature distribution of the key components of the energy storage vehicle. Due to factors such as the electrical characteristics of the sensors themselves, environmental interference, and electromagnetic interference during signal transmission, the original signals often contain various noise components.

[0055] For voltage signals and current signals, the data preprocessing module uses the Kalman filtering algorithm for processing. Kalman filtering is a recursive optimal estimation algorithm. Through two steps of prediction and update, it uses the prior knowledge of the system state and the current measurement value to perform optimal estimation of the system state. When processing voltage and current signals, the Kalman filter regards the signal as the output of a dynamic system. By establishing a state space model, it estimates and predicts the true value of the signal. Specifically, the filter first predicts the state value at the current moment based on the state estimation value at the previous moment and the system model; then it compares the measurement value at the current moment with the predicted value to calculate the error; finally, according to the magnitude of the error, it adjusts the predicted value to obtain the optimal estimation value at the current moment. By continuously repeating this process, the Kalman filter can effectively suppress the random noise in the signal and improve the signal quality.

[0056] For thermal imaging signals, due to their spatial distribution characteristics, the data preprocessing module uses the Butterworth filtering algorithm for processing. The Butterworth filter is a low-pass filter with the maximum flat amplitude response. It can provide the flattest frequency response within the passband and achieve rapid attenuation within the stopband. When processing thermal imaging signals, the Butterworth filter first performs a Fourier transform on the image to convert the image from the spatial domain to the frequency domain; then it designs a suitable low-pass filter in the frequency domain to filter out the high-frequency noise components; finally, through the inverse Fourier transform, it converts the filtered image back to the spatial domain. In this way, the Butterworth filter can effectively remove the high-frequency noise in the thermal imaging image while retaining the detail information of the image.

[0057] After completing the noise filtering, the data preprocessing module converts the processed sensor signals into standard state parameters. For voltage signals and current signals, they are converted into standardized voltage values and current values, that is, the actual measurement values are mapped to a standard numerical range for subsequent analysis and processing. For thermal imaging signals, they are converted into standardized temperature values, that is, by analyzing and processing the thermal imaging image, the temperature information of each pixel point is extracted and converted into the actual temperature value. These standardized state parameters will be used as the reference monitoring signals for subsequent feature extraction and state analysis.

[0058] The feature extraction module is respectively deployed in the dynamic adaptation module and the anomaly correction module. Through dimensionality reduction algorithms such as principal component analysis (PCA) and linear discriminant analysis (LDA), it performs feature dimensionality reduction processing on the data input to the dynamic adaptation module and the anomaly correction module to extract the first feature set corresponding to the dynamic adaptation module and the second feature set corresponding to the anomaly correction module.

[0059] In the dynamic adaptation module, the feature extraction module receives the preprocessed reference monitoring signals, including normalized voltage values, current values, temperature values, etc. To extract the key features from this data, the feature extraction module uses the principal component analysis (PCA) algorithm for feature dimensionality reduction. PCA is an unsupervised learning algorithm that projects the original data into a new low-dimensional space through linear transformation, so that the projected data has the maximum variance. When processing the reference monitoring signals, PCA first calculates the covariance matrix of the data, and then solves the eigenvalues and eigenvectors of the covariance matrix; then it selects several eigenvectors with the largest eigenvalues to form the projection matrix; finally, it projects the original data onto this projection matrix to obtain the data after dimensionality reduction. Through the PCA algorithm, the feature extraction module can extract the principal components that best represent the data change trend from the reference monitoring signals, and these principal components constitute the first feature set corresponding to the dynamic adaptation module. For example, the first feature set may include the principal components reflecting the load change trend, the principal components characterizing the temperature distribution features, etc. These features can help the system better understand the operating state of the energy storage vehicle, so as to achieve dynamic adaptation.

[0060] In the anomaly correction module, the feature extraction module uses the linear discriminant analysis (LDA) algorithm to perform feature dimensionality reduction on the input data. LDA is a supervised learning algorithm whose goal is to find a projection direction that maximally separates data of different classes in the projected space and brings data of the same class as close as possible in the projected space. When processing the reference monitoring signals, LDA first calculates the between-class scatter matrix and the within-class scatter matrix based on the known anomaly samples and normal samples; then it solves the generalized eigenvalue problem to obtain the optimal projection direction; finally, it projects the original data onto this projection direction to obtain the data after dimensionality reduction. Through the LDA algorithm, the feature extraction module can extract the features that best distinguish the anomaly state from the normal state from the reference monitoring signals, and these features constitute the second feature set corresponding to the anomaly correction module. For example, the second feature set may include features characterizing the charge-discharge efficiency decay mode, features reflecting abnormal voltage fluctuations, etc. These features can help the system timely detect anomalies during the operation of the energy storage vehicle and make corresponding corrections.

[0061] The dynamic adaptation module realizes the dynamic adaptation function by receiving the first feature set output by the feature extraction module. This module has a built-in policy selection matrix, in which adaptation policy templates under different working conditions are pre-stored, including steady-state operation policies, transient response policies, and high-load coping policies. When the first feature set is input, a real-time matching algorithm calculates the similarity between the feature vector and the working condition labels of the policy matrix, and selects the policy template with the highest matching degree as the basic policy. The parameter fine-tuning unit dynamically adjusts the basic policy parameters according to the feature values: when the load fluctuation deviation exceeds the threshold during steady-state operation, the load balancing policy response coefficient is increased by 0.1 to 0.3; when the temperature is abnormal during transient charge and discharge, the temperature suppression compensation intensity is increased by a preset ratio. Finally, the adjusted policy parameters are encapsulated as adaptation policy instructions and output. After receiving the second feature set, the anomaly correction module uses a decision tree model built based on the historical fault database by its anomaly pattern recognizer to identify the anomaly type. The policy generator calls the correction policy library according to the anomaly type: when a temperature mutation anomaly occurs, the temperature compensation policy is extracted and the compensation amount is calculated; when a sudden drop in efficiency is detected, the charge and discharge smoothing policy is superimposed to dynamically adjust the slope of the charge and discharge curve. All correction policies are encapsulated as instructions and output. The two modules adopt a real-time pipeline architecture and are updated and processed every 100 milliseconds. Intermediate data is exchanged through shared memory to avoid repeated calculations.

[0062] Based on the first feature set and the second feature set, the prediction model module uses deep neural networks (such as LSTM and Transformer models) to learn and predict time series data, generating a time series prediction signal that reflects the future change trend of the operating state of the energy storage vehicle. The prediction model module further combines the reference monitoring signal and generates a state control instruction for the energy storage unit through a signal fusion algorithm (such as weighted average fusion and Kalman filter fusion) to achieve forward-looking regulation of the operating state of the energy storage vehicle.

[0063] The prediction model module first receives the first feature set and the second feature set from the feature extraction module. Since these feature sets usually contain time series data, that is, data sequences arranged in chronological order, the prediction model module uses a long short-term memory network (LSTM) to learn and predict these time series data. LSTM is a special type of recurrent neural network (RNN) that can effectively handle long-term dependencies in long sequence data. By introducing gating mechanisms, including input gates, forget gates, and output gates, LSTM can selectively remember and forget information, thus better capturing the dynamic changes in time series data. When processing the first feature set and the second feature set, LSTM first inputs the feature sequence into the network, extracts the feature patterns in the sequence through the processing of the hidden layer; then, based on these feature patterns, predicts the future state and generates a time series prediction signal. For example, LSTM can predict the load demand and temperature changes of the energy storage vehicle in the future period according to the historical load change trend and temperature distribution characteristics, providing a basis for the forward-looking regulation of the system.

[0064] In some cases, the prediction model module can also use the Transformer model to process time series data. The Transformer model is a deep learning model based on the attention mechanism. It can process sequence data in parallel and capture long-range dependencies in the sequence. Compared with the LSTM, the Transformer model has higher efficiency and better performance when processing long sequence data. When processing the first feature set and the second feature set, the Transformer model first embeds the feature sequence, mapping each feature into a low-dimensional vector space; then through the multi-head attention mechanism, it calculates the degree of association between each position in the sequence; finally, through the feed-forward neural network, it processes the features weighted by attention to generate a time series prediction signal.

[0065] After the prediction model module generates the time series prediction signal, it fuses it with the reference monitoring signal. Specifically, the prediction model module uses the Kalman filter fusion algorithm to fuse the time series prediction signal and the reference monitoring signal. The Kalman filter fusion algorithm is an optimal estimation fusion algorithm. It can perform weighted fusion on multiple signals according to the statistical characteristics of the signals to obtain the optimal estimation result. During the fusion process, the Kalman filter first calculates the respective weights according to the variances of the time series prediction signal and the reference monitoring signal; then according to these weights, it performs a weighted average on the two signals to obtain the fused signal; finally, through two steps of prediction and update, it continuously adjusts the fusion result to make it more accurately reflect the operating state of the energy storage vehicle. In this way, the prediction model module can combine the advantages of the time series prediction signal and the reference monitoring signal to generate more accurate and reliable state control instructions.

[0066] The generated state control instructions will be sent to the actuators of the energy storage unit, such as the battery management system (BMS), charge and discharge controller, etc., to achieve forward-looking regulation of the operating state of the energy storage vehicle. For example, if the prediction model module predicts that the load demand of the energy storage vehicle will increase in the next period of time, it can adjust the charge and discharge strategy of the battery in advance and increase the output power of the battery to meet the load demand; if it predicts that the battery temperature will rise, it can start the cooling system in advance to lower the battery temperature and ensure the safe operation of the battery. Through this forward-looking regulation, the system can manage the operating state of the energy storage vehicle more efficiently and improve the reliability and performance of the energy storage vehicle.

[0067] In addition, the data preprocessing module further includes a timestamp synchronization unit. Through a hardware clock synchronization circuit (such as the PTP Precision Time Protocol, GPS clock synchronization) or a software timestamp calibration algorithm, it calibrates the sampling frequencies of different sensors to ensure the consistency of multi-source data on the time axis, eliminates the timing offset problem caused by asynchronous sampling times, and provides a reliable data basis for subsequent data analysis and condition monitoring. The timestamp synchronization unit first obtains the sampling timestamps of each sensor, and then unifies these timestamps to the same time reference through a hardware clock synchronization circuit or a software algorithm. For example, when using the PTP Precision Time Protocol, the timestamp synchronization unit synchronizes the clocks of each sensor with a high-precision master clock to ensure that the sampling timestamps of all sensors have the same time reference. When using a software timestamp calibration algorithm, the timestamp synchronization unit corrects the sampling timestamps according to the communication delay and processing delay between sensors to eliminate the time deviation caused by the delay. Through the processing of the timestamp synchronization unit, the data collected by different sensors are accurately aligned on the time axis, providing a reliable data basis for subsequent feature extraction and condition analysis.

[0068] Embodiment 2:

[0069] In this embodiment, the main controller also has the function of retrieving a preset monitoring threshold according to the first target monitoring task. Specifically, when the system receives the first target monitoring task (such as safety threshold monitoring under normal operating conditions), the main controller retrieves the corresponding preset monitoring threshold through a parameter storage unit (such as storage devices like EEPROM, flash memory), and this threshold includes the upper safety limit of battery temperature, the minimum standard of charge-discharge efficiency, the allowable range of load fluctuation, etc., as the first target analysis instruction. The main controller sends the first target analysis instruction to the condition analysis coprocessor through a serial communication interface (such as UART, CAN bus). The condition analysis coprocessor compares and analyzes the preset monitoring threshold with the real-time monitored battery temperature, charge-discharge efficiency, and load fluctuation data. When the actual data exceeds the preset threshold, it triggers corresponding alarm or control mechanisms, such as starting the temperature cooling system, adjusting the charge-discharge power, restricting the load access, etc., to ensure that the energy storage vehicle operates within a safe and reliable range.

[0070] The main controller, as the core control unit of the entire system, undertakes the important responsibilities of receiving, processing, and distributing various tasks. When facing the first target monitoring task, the main controller first needs to accurately identify the task type. This identification process relies on the task classification mechanism preset in the system, which may be designed based on various factors such as task priority, time characteristics, and functional requirements. For example, the safety threshold monitoring task under normal operating conditions may be marked as a high-priority task, and the system will allocate dedicated processing procedures and resources for it. When the main controller receives an external input or an internally triggered first target monitoring task, it will conduct a detailed analysis of the task through the task parsing module, extract the key features and parameters of the task to determine the specific monitoring requirements.

[0071] After determining the task type, the main controller will access the parameter storage unit to retrieve the preset monitoring thresholds corresponding to the first target monitoring task. The parameter storage unit is a module in the system specifically used to store various preset parameters and configuration information, and it can be implemented using various storage technologies such as EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory, etc. These storage devices have the characteristic of non-volatility and can maintain the integrity of data even after the system power-off. The preset monitoring thresholds are a series of reference values preset in the system design and debugging stages based on the performance parameters, safety standards, and operating requirements of the energy storage vehicle, including the upper safety limit of battery temperature, the minimum standard of charge and discharge efficiency, the allowable range of load fluctuations, etc. These thresholds are important bases for ensuring the safe and stable operation of the energy storage vehicle.

[0072] The main controller communicates with the parameter storage unit through the internal data bus, and accurately locates and reads the corresponding preset monitoring thresholds according to the task type and requirements. To ensure the accuracy and reliability of the data, the main controller may adopt data verification mechanisms such as CRC verification and parity verification during the reading process to verify the read data. If it is found that the data is incorrect or inconsistent, the main controller will take corresponding measures for processing, such as re-reading the data, triggering an error alarm, etc.

[0073] Once the preset monitoring thresholds are successfully obtained, the main controller will convert these thresholds into a unified data format and protocol within the system to form the first target analysis instruction. This conversion process needs to consider the interface requirements and data format specifications of the subsequent processing modules to ensure that the instruction can be correctly understood and executed by the status analysis coprocessor. In addition to the preset monitoring thresholds, the first target analysis instruction may also include auxiliary data such as task identifiers, timestamps, and priority information, so that the status analysis coprocessor can accurately identify and process the instruction.

[0074] The main controller sends the first target analysis instruction to the state analysis coprocessor through a serial communication interface. The serial communication interface is a commonly used data transmission interface, which has the advantages of long transmission distance, strong anti-interference ability, low cost, etc. In this system, serial communication technologies such as UART (Universal Asynchronous Receiver-Transmitter) or CAN bus (Controller Area Network) may be adopted. UART is a full-duplex asynchronous communication interface, which realizes two-way data transmission through the transmit line (TX) and receive line (RX), and is suitable for point-to-point communication scenarios. The CAN bus is a multi-master and slave serial communication bus, which uses differential signal transmission, has high reliability and real-time performance, and is suitable for communication between multiple nodes. When the main controller sends an instruction, it will package and encode the instruction according to the protocol specifications of the selected communication interface, and add control information such as start bits, stop bits, and parity bits to ensure the correct transmission of data.

[0075] After receiving the first target analysis instruction, the state analysis coprocessor first parses and decodes the instruction to extract the preset monitoring threshold and other relevant information. Then, it obtains real-time monitoring data from various sensors, including battery temperature, charge and discharge efficiency, and load fluctuation data, etc. These sensors are distributed in various key parts of the energy storage vehicle, such as the battery pack, charge and discharge circuit, load interface, etc., and they can sense and collect the operating state parameters of the energy storage vehicle in real time.

[0076] The state analysis coprocessor compares and analyzes the real-time monitoring data with the preset monitoring threshold. This analysis process is a multi-dimensional and multi-level comparison process. For the battery temperature, the state analysis coprocessor will compare the real-time measured temperature value with the preset temperature safety upper limit. If the real-time temperature exceeds the safety upper limit, it indicates that the battery may have an overheating risk and measures need to be taken in time for cooling. For the charge and discharge efficiency, the state analysis coprocessor will calculate the efficiency value of the current charge and discharge process and compare it with the preset minimum standard. If the efficiency value is lower than the minimum standard, it may mean that there is a fault in the charge and discharge circuit or excessive energy loss, and inspection and adjustment are required. For the load fluctuation, the state analysis coprocessor will analyze the change of the load current or power to determine whether it is within the preset allowable range. If the load fluctuation exceeds the allowable range, it may cause an impact on the power system of the energy storage vehicle and affect the stability and reliability of the system.

[0077] During the comparative analysis process, the status analysis coprocessor may adopt various analysis methods and techniques. For example, it can perform statistical analysis on real-time monitoring data, calculate statistical features such as the mean, standard deviation, maximum value, minimum value, etc. of the data, so as to comprehensively understand the distribution and change trend of the data. It can also use time series analysis methods to model and predict the historical change situation of the data, and judge whether the current abnormal situation is a temporary fluctuation or a potential fault symptom. In addition, the status analysis coprocessor can also combine machine learning algorithms to perform pattern recognition and classification on the monitoring data, so as to more accurately identify abnormal states and fault types.

[0078] When the status analysis coprocessor discovers that the real-time monitoring data exceeds the preset monitoring threshold, it will trigger the corresponding alarm or control mechanism. The alarm mechanism can be implemented in various ways, such as audible and visual alarms, SMS alarms, email alarms, etc. The audible and visual alarm system will emit obvious audible and visual signals on the control panel of the energy storage vehicle or in the monitoring center to remind the operator to pay attention to the abnormal situation. SMS alarms and email alarms will send the abnormal information to the mobile phones or email boxes of relevant personnel in a timely manner to ensure that they can understand the situation in time and take measures.

[0079] The control mechanism will take corresponding control measures according to the type and severity of the abnormal situation. If the battery temperature exceeds the safety upper limit, the status analysis coprocessor will send a control instruction to the temperature cooling system to start the cooling fan or liquid cooling system to accelerate the heat dissipation speed of the battery and reduce the battery temperature. If the charge-discharge efficiency is lower than the minimum standard, the status analysis coprocessor will adjust the charge-discharge power, optimize the charge-discharge strategy, and improve the charge-discharge efficiency. If the load fluctuation exceeds the allowable range, the status analysis coprocessor will limit the load access to avoid the impact of excessive load fluctuations on the system. In some cases, in order to ensure the safety of the system, the status analysis coprocessor may take emergency shutdown measures to cut off the power supply of the energy storage vehicle to prevent the fault from further expanding.

[0080] To ensure the reliability and stability of the entire monitoring and control process, the status analysis coprocessor will monitor and feedback the execution situation of the control instruction in real time. It will obtain the status information of the actuator through sensors to judge whether the control instruction is correctly executed. If it is found that there are abnormal situations during the execution process, the status analysis coprocessor will promptly adjust the control strategy and take remedial measures to ensure that the system can resume normal operation as soon as possible.

[0081] In practical applications, the preset monitoring threshold is not fixed, but needs to be dynamically adjusted according to factors such as the usage environment, operating status, and performance changes of the energy storage vehicle. The main controller can receive threshold update instructions from the cloud server or maintenance personnel through the remote communication interface to update and maintain the preset monitoring threshold in the parameter storage unit. In addition, the system can also automatically optimize the preset monitoring threshold according to long-term operation data and experience through a self-learning mechanism, improving the adaptability and accuracy of the system.

[0082] Embodiment 3:

[0083] In this embodiment, the main controller can call a dynamic optimization algorithm according to the second target monitoring task. When the system needs to execute the second target monitoring task (such as optimizing the operating status under complex working conditions), the main controller uses a dynamic optimization algorithm based on time series correlation analysis (such as particle swarm optimization algorithm, genetic algorithm) as the second target analysis instruction and transmits it to the state analysis co-processor through the data bus. After receiving the instruction, the state analysis co-processor performs time series correlation analysis on the operation data of the energy storage vehicle, identifies the time series dependence relationship between different parameters (such as the lag correlation between battery temperature change and charge-discharge current, the causal relationship between load fluctuation and voltage response), and uses the dynamic optimization algorithm to globally optimize the monitoring parameters and control strategies.

[0084] As the task scheduling core of the system, the main controller first distinguishes the first target monitoring task and the second target monitoring task through the task identification module. The second target monitoring task usually corresponds to more complex working conditions, such as high-load cycles, abnormal temperature gradients, or current mutations. Such tasks need to break through the fixed limit of the preset threshold and instead achieve adaptive optimization of parameters through dynamic algorithms. After receiving such task trigger signals (such as abnormal working condition marks from sensors, manually set optimization instructions), the main controller retrieves the dynamic optimization algorithm based on time series correlation analysis from the algorithm storage unit. The algorithm storage unit can store multiple optimization algorithms in the form of a firmware library, including particle swarm optimization (PSO), genetic algorithm (GA), simulated annealing algorithm, etc. Each algorithm corresponds to different parameter optimization scenarios.

[0085] The main controller packages the selected dynamic optimization algorithm into a second target analysis instruction. The instruction structure includes algorithm type identification, optimization target parameters (such as maximizing charge-discharge efficiency, minimizing temperature fluctuation), constraint conditions (such as voltage safety range, load power upper limit), and time series analysis window parameters (such as the number of historical data sampling periods). The instruction is transmitted to the state analysis co-processor through the system data bus (such as AMBA AHB bus). The data bus supports high-speed parallel transmission to ensure that the algorithm instruction and the accompanying historical data block can reach the target module in real time.

[0086] After receiving the instruction, the status analysis coprocessor first starts the timing correlation analysis module. This module analyzes the dependence relationship between parameters by using Granger Causality Test or Vector Auto-Regression (VAR) for the timing characteristics in the operation data of the energy storage vehicle. Taking the lag correlation between the battery temperature ( ), and the charge and discharge current ( ) as an example, the following lag regression model is constructed:

[0087]

[0088] Where, represents the battery temperature at the current moment , represents the charge and discharge current lagging time units, represents the battery temperature lagging time units, is the constant term, and are the regression coefficients, is the random error term, represents the lag order of the charge and discharge current, represents the lag order of the battery temperature. By estimating the coefficients using the least squares method and conducting significance tests, the lag effect order of the current on the temperature can be determined, thereby establishing a timing correlation model between parameters.

[0089] The output result of the timing correlation analysis is an association matrix containing parameter causal relationships and lag durations, which serves as one of the input conditions for the dynamic optimization algorithm. Taking the particle swarm optimization algorithm as an example, the status analysis coprocessor defines the optimization variables as the adjustment factors of the monitoring parameters (such as the charge and discharge power correction coefficient , the temperature threshold offset ) and the control strategy parameters (such as the load balancing weight , the starting threshold of the cooling system ), and constructs a multi-dimensional search space. Each particle in the particle swarm represents a set of parameter combinations, and its position vector corresponds to different control strategy configurations.

[0090] The design of the optimization objective function needs to comprehensively consider multi-dimensional indicators. For example, under high-load working conditions, the objective function can be defined as:

[0091]

[0092] Where, is the function of the charge and discharge efficiency with respect to the power correction coefficient, is the standard deviation of the battery temperature, is the maximum value of the voltage deviation from the rated value, is the weight coefficient of each index, reflecting the optimization focus direction (for example, when giving priority to ensuring voltage stability, takes a higher value).

[0093] The particle swarm optimization algorithm searches for the optimal solution that minimizes (or maximizes) the objective function by iteratively updating the positions and velocities of particles. In each iteration, the particle updates its velocity according to its own historical best position and the global best position as follows:

[0094]

[0095]

[0096] where, is the inertia weight, which is used to balance the global search and local search capabilities; are the acceleration constants, which control the movement trends of the particles towards their own best and the global best positions; is a random number between [0, 1], which introduces randomness to avoid local optima.

[0097] In each iteration, the state analysis coprocessor substitutes the parameter combination corresponding to the particle into the time-series correlation model, predicts the optimized operating state of the energy storage vehicle, and calculates the objective function value. After a preset number of iterations or when the convergence condition is met (such as no significant change in the optimal solution for several consecutive generations), the algorithm terminates and outputs the optimal parameter combination. For example, in the optimization of the charge and discharge smoothing strategy, the power correction coefficient determined by the particle swarm optimization can reduce the voltage fluctuation amplitude to a certain proportion of the initial value, while ensuring that the charge and discharge efficiency is maintained within a reasonable range.

[0098] The output result of the dynamic optimization algorithm is transmitted to the control strategy generation module through a standardized interface. This module generates specific control instructions according to the optimal parameter combination, such as adjusting the charge and discharge current limits of the battery management system (BMS), updating the fan speed control curve of the thermal management system, and correcting the power scheduling logic of the load distributor. The instructions are sent to the actuators of the energy storage unit through a fieldbus (such as CANopen) to achieve the global optimization of the operating state of the energy storage vehicle.

[0099] To ensure the real-time performance of the dynamic optimization process, the state analysis coprocessor uses a hardware acceleration unit (such as a parallel computing module implemented by FPGA) to accelerate the algorithm iteration process and shorten the time consumed for a single iteration. At the same time, the system sets an optimization cycle timer, which can dynamically adjust the optimization frequency according to the complexity of the working conditions. For example, the optimization cycle is extended during steady-state operation to reduce the computational load, and the cycle is shortened under transient working conditions to quickly respond to parameter changes.

[0100] Embodiment 4:

[0101] In this embodiment, the analysis and optimization mode includes three modes: data verification function bypass, global parameter verification, and local parameter verification. The three modes are respectively applicable to different working condition scenarios, and the automatic switching and execution of the modes are realized through the logic judgment module of the state analysis coprocessor.

[0102] The triggering scenarios for the data verification function bypass mode are usually emergency working conditions or abnormal data transmission states. For example, when a mobile energy storage vehicle is suddenly caught in a heavy rain during outdoor operation, resulting in temporary data transmission interruption of some sensors due to moisture, or a sudden load short-circuit fault occurs during the charging and discharging process, the system needs to respond quickly to avoid safety accidents. At this time, the fault detection unit of the state analysis coprocessor identifies data anomalies through communication link status monitoring (such as CAN bus error frame counting) or the absence of sensor heartbeat signals. The logic judgment module immediately triggers the data verification function bypass mode, temporarily skipping the verification process of abnormal sensor data, and instead calling the most recently stored valid data (such as historical data 100 ms before the fault) to generate control instructions. Taking the charge and discharge control of the battery pack as an example, if a current sensor fails instantaneously during a short circuit, the system will directly send an instruction to cut off the output to the charge and discharge controller based on the current value before the fault and the preset safe current threshold, avoiding response delays caused by waiting for data verification. In this mode, the system only maintains the most basic safety control function, and at the same time sends a data anomaly warning to the operation and maintenance personnel through indicator light flashing or a remote communication interface, prompting them to perform sensor maintenance.

[0103] The global parameter verification mode is suitable for scenarios where the system is operating normally and the status of the energy storage vehicle needs to be fully evaluated, such as daily routine inspections, pre-inspections before starting long-term charging and discharging tasks, or status confirmation when switching working conditions (such as switching from static energy storage mode to mobile driving mode). After receiving the global verification trigger signal (such as scheduled task scheduling, manual verification instructions initiated through the touch screen), the state analysis coprocessor starts the multi-threaded parallel processing mechanism to perform a systematic verification of all monitoring parameters. The verification scope covers core indicators such as battery temperature (temperature sensor data of all battery modules in the vehicle), charging and discharging efficiency (input and output power calculation of each charging and discharging circuit), load fluctuation (comparison of real-time load current and historical load curve), and also includes consistency verification of multi-modal data such as voltage, current, and thermal imaging.

[0104] Taking battery temperature verification as an example, the system first compares the real-time temperature data of each module with the temperature distribution of the same working condition in the historical operation data. If it is found that the temperature of a module deviates significantly from the average level (such as the temperature difference exceeds 5°C), the local thermal runaway warning of the module is triggered, and the thermal imaging data is combined to determine whether there is poor contact or abnormal battery cell. In the charge and discharge efficiency verification, the state analysis coprocessor will calculate the difference between the input power and the output power according to the principle of energy conservation. If the difference exceeds the preset loss threshold (such as higher than 5%), the impedance detection of the charge and discharge circuit is started to check whether there is a problem of line aging or increased contact resistance. During the global verification process, the system will generate a detailed verification report, recording the deviation value, verification timestamp and abnormal point location of each parameter for operation and maintenance personnel to review and analyze. After the verification is completed, if all parameters meet the preset standards, the system automatically switches back to normal operation mode; if multiple abnormalities are found, the deep diagnosis process is triggered, and more complex analysis algorithms (such as fault tree analysis) are called to locate the root cause of the problem.

[0105] The local parameter verification mode responds precisely to abnormal alarms of specific monitoring parameters or subsystems to avoid interference of global verification on system operation. For example, when the energy storage vehicle is running under low-load conditions, the temperature sensor of a battery module suddenly sends an over-threshold signal. The single parameter anomaly detection unit of the state analysis coprocessor first determines whether the signal is a valid anomaly through the signal jump recognition algorithm (such as exceeding the threshold for 3 consecutive sampling cycles and the temperature difference with the adjacent module is greater than 3°C). If the anomaly is confirmed, the logic judgment module immediately activates the local parameter verification mode and performs in-depth verification only on the module and its associated subsystems (such as the corresponding BMS sub-controller and cooling fan branch).

[0106] During the specific implementation process, the system first performs self-calibration on the temperature sensor of the module. By switching to the redundant backup sensor or invoking the historical temperature-voltage correlation model (the mapping relationship between temperature and open-circuit voltage established based on the historical data of the same module), the accuracy of the original data is verified. If the temperature is still abnormal after calibration, the system further checks whether the charge and discharge current of the module exceeds the rated value, whether the cooling fan is operating normally, and whether the thermal conductive silicone between the modules is aged and peeled off. In the local verification scenario with abnormal load fluctuations, if the current mutation of a certain load interface exceeds the allowable range, the system will isolate the load branch, test its impedance characteristics separately, and compare them with the nominal parameters of the load device to determine whether it is caused by a load device failure or a surge current at the moment of connection.

[0107] The execution of the local parameter verification mode is highly targeted and time-sensitive. At the hardware level, the state analysis coprocessor addresses and collects data from the target subsystem separately through an independent sub-module controller (such as the parallel processing unit implemented by FPGA), avoiding affecting other normally operating subsystems; at the software level, an incremental verification algorithm is adopted to deeply analyze only the data stream related to abnormal parameters, rather than repeatedly processing all data. For example, when dealing with an abnormal single temperature sensor, the system only retrieves the historical data of the module where the sensor is located, the temperature field distribution data of adjacent modules, and the corresponding heat dissipation control logic, without reloading the configuration information of all sensors in the whole vehicle, thus controlling the verification time consumption within milliseconds.

[0108] The three analysis and optimization modes work together through the mode management module of the state analysis coprocessor. The mode management module maintains a priority queue, in which the data verification function bypass mode has the highest priority (corresponding to emergency working conditions), the global parameter verification mode is the second (corresponding to regular maintenance requirements), and the local parameter verification mode is the normal response mode. When the system receives multiple mode trigger signals simultaneously, the mode management module arbitrates according to the preset conflict resolution rules (such as "emergency first", "local prior to global") to ensure the stable operation of the system under any working conditions.

[0109] In practical applications, the switching logic of the three modes can be hardware-accelerated through a field-programmable gate array (FPGA) to ensure that the mode switching delay is less than 10 ms. For example, when the system detects an abnormality in a certain subsystem from the global parameter verification mode, it can immediately interrupt the global process and jump to the local parameter verification mode. After the local problem is processed, the global verification can be resumed from the breakpoint. This flexible mode switching mechanism enables the system to complete a comprehensive health check under normal working conditions and quickly respond under abnormal working conditions, realizing the refined management of the operating state of the energy storage vehicle.

[0110] Through the organic combination of the above three analysis and optimization modes, the system can dynamically adjust the verification strategy in different scenarios to balance the requirements of security, reliability, and operation efficiency. The data verification function bypass mode ensures the system's survivability in emergency situations. The global parameter verification mode supports the establishment of a preventive maintenance system. The local parameter verification mode realizes the accurate positioning and rapid repair of abnormal events, jointly constituting the intelligent decision-making core of the mobile energy storage vehicle operation status monitoring and analysis system.

[0111] Embodiment 5:

[0112] In this embodiment, the prediction model module includes a dynamic adjustment module, a signal fusion module, and an instruction issuance module. Each module realizes the precise control of the energy storage unit through data interaction and logical coordination. The following details its implementation method in combination with specific scenarios:

[0113] I. Implementation of the dynamic adjustment module

[0114] The dynamic adjustment module selects corresponding strategies from the optimization and correction strategy set for parameter iteration according to the first feature set (such as load distribution feature) and the second feature set (such as temperature anomaly feature) output by the feature extraction module. For example, when the energy storage vehicle is connected to the temporary power load at a construction site, the simultaneous startup of multiple power tools may cause load imbalance. The first feature set identifies through principal component analysis that the load power deviation of each energy storage unit exceeds 20%. The dynamic adjustment module triggers the load balancing strategy. At this time, the module calculates the optimal power distribution coefficient by traversing parameters such as the remaining capacity and charge-discharge efficiency of each battery pack. For example, it transfers part of the power of the high-load unit to the low-load unit and realizes load balancing by adjusting the current limit value of each charge-discharge controller.

[0115] If the second feature set shows that the temperature of a certain battery module rises by 15°C within 5 minutes, exceeding the preset temperature rise rate threshold, the dynamic adjustment module calls the temperature suppression strategy. The specific process is as follows: First, confirm whether the corresponding cooling fan of this module has been running at full speed. If it has not been started, send an instruction to increase the fan speed to 100%. If the fan is already at the highest speed but the temperature still continues to rise, further trigger the opening of the solenoid valve of the liquid cooling system to increase the coolant flow. During this process, the module forms a closed-loop control by monitoring the temperature change rate in real time and updating the cooling parameters every 10 seconds.

[0116] When a voltage fluctuation exceeding ±5% of the nominal value is detected during the charge and discharge process (e.g., a sudden drop from 48V to 45.6V), the dynamic adjustment module activates the charge and discharge smoothing strategy. Taking the charging condition as an example, the module calculates the adjustment step of the charging current (e.g., adjusting 5A each time) according to the frequency and amplitude of the voltage fluctuation, and suppresses the voltage overshoot or drop by gradually reducing the charging current. If the fluctuation is caused by a sudden increase in the load, the module preferentially cuts off the power supply to non-critical loads (such as the lighting system) to ensure the voltage stability of the energy storage unit.

[0117] II. Implementation of Sub-modules of the Dynamic Adjustment Module

[0118] The First Adjustment Module (Temperature and Load-related Adjustment)

[0119] When the energy storage unit is in steady-state operation (such as continuously powering a communication base station), the steady-state monitoring module compares the current load fluctuation with the expected load curve (a 24-hour load template generated based on historical data) in real time. If the actual load in a certain period exceeds 15% of the expected value (e.g., the expected load is 5kW and the actual reaches 5.75kW), the power distribution coefficient of the load balancing strategy is slightly adjusted (e.g., increasing the power transfer amount by 10%); if the deviation is within the tolerance range (±10%), the original strategy is maintained.

[0120] In transient charge and discharge scenarios (such as the start and stop phases of rapid charging of electric vehicles), the transient response module captures the thermal imaging signal at the moment when the charging gun is inserted and compares it with the preset temperature distribution model (the temperature uniformity deviation of the battery pack during normal charging ≤2℃). If the temperature difference in a certain area exceeds 3℃, the temperature suppression strategy is immediately dynamically compensated, for example, increasing the coolant flow rate by 5% in the liquid cooling pipeline corresponding to this area; if the difference does not exceed the critical value, no additional operation is performed.

[0121] When the energy storage vehicle encounters a high-load impact (such as connecting to a large motor startup), the abnormal handling module dynamically adapts the charge and discharge smoothing strategy by combining the voltage fluctuation data (such as the voltage dropping to 85% of the rated value) and the current peak value (exceeding 120% of the rated current). Specific measures include: temporarily increasing the current limit threshold to 110% of the rated value (for a duration not exceeding 5 seconds) to allow short-time large current to pass, and at the same time starting the pre-discharge mechanism of the battery pack to relieve the voltage sudden drop by releasing part of the reserve energy.

[0122] If a signal jump occurs in a single sensor (such as the voltage sensor of a certain battery module) (e.g., suddenly changing from 3.6V to 4.2V within 1 second), the single-parameter adjustment module locally adapts the correction strategy corresponding to this signal within the monitoring period (such as 1 minute). For example, temporarily switching to the redundant sensor data for control and performing a moving average filter on the historical data of the faulty sensor. If the signal is still abnormal for 5 consecutive periods, mark this sensor as a faulty state and trigger an operation and maintenance alarm.

[0123] Second adjustment module (trend prediction and compensation)

[0124] The prediction configuration module activates the trend prediction function for specific working conditions such as high-load cycles, abnormal temperature gradients, and sudden current change intervals. For example, when the energy storage vehicle enters an industrial park to supply power to multiple welding machines, the system recognizes that the load current fluctuates periodically between 50 - 150 A (high-load cycle condition), and automatically activates the prediction parameter register (storing the training cycle number, sliding window size, etc. of the prediction model) and the prediction action period register (setting to predict in the first 10 seconds of each load cycle).

[0125] After the prediction function is activated, the log recording module continuously records the parameter evolution data under specific working conditions. Taking the abnormal temperature gradient condition as an example, when the temperature of a certain battery cluster shows a temperature difference of 10℃ from the inlet to the outlet (exceeding the normal condition standard of 5℃), the module records in real-time the temperature of each battery cell, the corresponding charge and discharge current, the rotation speed of the cooling fan, etc. data to form a time-series log. These data are used to train the non-linear prediction model and serve as historical references in subsequent similar working conditions.

[0126] When the non-linear prediction module reaches the prediction target (such as predicting that the battery temperature will exceed the safety threshold) within the specified time period (such as 30 minutes before the end of charging) under specific working conditions, it performs compensation based on the evolution data. For example, when it detects that the voltage rise rate slows down at the end of charging (indicating that it is about to be fully charged), it gradually reduces the charging current 5 minutes in advance to avoid a sudden increase in temperature due to overcharging. The compensation process uses the exponential smoothing algorithm to dynamically adjust the prediction step size according to historical errors to ensure that the lead of the control instruction matches the actual working condition.

[0127] III. Implementation of the signal fusion module and the instruction issuing module

[0128] The signal fusion module performs spatio-temporal alignment operations on the reference monitoring signals (such as standardized temperature and voltage data) and the abnormal correction signals (such as the power adjustment value after load balancing). Taking the multi-battery pack parallel connection scenario as an example, the voltage signals of each battery pack may have a slight time sequence offset (such as a maximum deviation of 5 ms) due to the sampling clock difference. The fusion module unifies the signals to the global clock reference of the main controller through the timestamp interpolation algorithm (such as linear interpolation) to ensure that parameters such as voltage, current, and temperature are aligned at the same moment. Spatial alignment is for the thermal imaging signal and the temperature sensor data, and through the coordinate mapping algorithm (such as corresponding the thermal imaging pixel points to the sensors at the physical positions one by one), the spatial error caused by the installation position difference is eliminated.

[0129] The fused signal undergoes logical judgment (such as checking whether the voltage is within the safe range and whether the temperature correction value exceeds the upper limit of the heat dissipation system's capacity) to generate the final control instruction. The instruction distribution module transmits the instruction to the control terminal of the energy storage unit via the CAN bus. For example:

[0130] Send a charge and discharge current limit update instruction (such as adjusting from 100A to 80A) to the battery management system (BMS);

[0131] Send a coolant flow rate adjustment instruction (such as increasing from 5L / min to 8L / min) to the liquid cooling system controller;

[0132] Send a power switching instruction (such as switching non-critical loads from the main battery pack to the backup battery pack) to the load distributor.

[0133] The instruction transmission process adopts a priority mechanism. Emergency instructions (such as a shutdown instruction when the temperature exceeds the safety threshold) have the highest transmission priority to ensure delivery to the control terminal within 1ms; regular adjustment instructions (such as load balancing parameter updates) are sent periodically (such as updated every 100ms) to reduce the bus load.

[0134] IV. Function of the Timestamp Synchronization Unit

[0135] The timestamp synchronization unit of the data preprocessing module calibrates the sampling frequencies of sensors such as voltage, current, and thermal imaging through a hardware clock synchronization circuit (such as the PTP Precision Time Protocol). For example, the voltage sensor uses a sampling rate of 100Hz, and the thermal imager uses a sampling rate of 25Hz. The synchronization unit makes the sampling time error between the two less than 10μs through the PTP protocol. For sensors that do not support hardware synchronization (such as old models of current sensors), the timestamps are calibrated post hoc through software algorithms: based on the signal transmission delay (the average delay measured through ping-pong testing is 2.3ms) and the processor processing time (fixed at 0.5ms), the collected timestamps are compensated to ensure the consistency of multi-source data on the time axis.

[0136] Through the coordinated operation of the above-mentioned modules, the prediction model module realizes the dynamic prediction and precise control of the operating state of the energy storage unit. From the real-time adjustment of load balancing to the trend compensation of temperature anomalies, and then to the fusion and alignment of multi-modal signals, the entire process combines logical judgment and data-driven methods to ensure the stable operation of the mobile energy storage vehicle under complex working conditions, and at the same time provides a solid data foundation for fault warning and optimization strategy iteration.

[0137] It should be noted that in this text, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device.

[0138] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An AI-based operating status monitoring and analysis system for mobile energy storage vehicles, characterized in that, Including: A main controller, an AI monitoring co-processor, and a status analysis co-processor; The main controller is used to retrieve corresponding monitoring and analysis instructions according to the operation status data of the energy storage vehicle, and send the monitoring and analysis instructions to the status analysis co-processor; The AI monitoring co-processor is used to schedule the status analysis instructions of the status analysis co-processor during the execution of multi-modal data parsing instructions; The status analysis co-processor is used to parse the dynamic monitoring parameters in the monitoring and analysis instructions to generate a reference monitoring signal; perform real-time collaborative monitoring on battery temperature, charge and discharge efficiency, and load fluctuation, and select corresponding analysis and optimization modes according to the monitoring results for dynamic adaptation to generate an anomaly correction signal, and combine the reference monitoring signal to generate a status control instruction for the energy storage unit.

2. The AI-based mobile energy storage vehicle operation status monitoring and analysis system according to claim 1, wherein The AI monitoring co-processor includes: A data preprocessing module, which is used to filter the noise of multi-source sensor signals and convert the processed sensor signals into standard status parameters as the reference monitoring signal; among them, the sensor signals include voltage signals, current signals, and thermal imaging signals; A feature extraction module, which is respectively deployed in the dynamic adaptation module and the anomaly correction module, and is used to perform feature dimensionality reduction on the dynamic adaptation module and the anomaly correction module respectively to extract a first feature set corresponding to the dynamic adaptation module and a second feature set corresponding to the anomaly correction module; A dynamic adaptation module, which is used to perform dynamic adaptation on battery temperature, charge and discharge efficiency, and load fluctuation according to the first feature set to generate an adaptation strategy instruction; An anomaly correction module, which is used to correct the anomalies of battery temperature, charge and discharge efficiency, and load fluctuation according to the second feature set to generate a correction strategy instruction; A prediction model module, which is used to generate a time series prediction signal through a deep neural network according to the adaptation strategy instruction and the correction strategy instruction, and combine the reference monitoring signal to generate a status control instruction for the energy storage unit.

3. The AI-based mobile energy storage vehicle operation status monitoring and analysis system according to claim 1, characterized in that The main controller is also used to retrieve a corresponding preset monitoring threshold as a first target analysis instruction through the parameter storage unit according to the first target monitoring task, and send the first target analysis instruction to the status analysis co-processor.

4. The AI-based mobile energy storage vehicle operation status monitoring and analysis system according to claim 1, characterized in that, The main controller is also used to use the dynamic optimization algorithm based on time series correlation analysis as a second target analysis instruction according to the second target monitoring task, and send the second target analysis instruction to the status analysis co-processor.

5. The AI-based mobile energy storage vehicle operation status monitoring and analysis system according to claim 1, wherein The analysis and optimization modes include: bypassing the data verification function, performing global parameter verification, and performing local parameter verification.

6. The AI-based mobile energy storage vehicle operation status monitoring and analysis system according to claim 2, wherein The prediction model module includes: A dynamic adjustment module, which is used to select optimization and correction strategies corresponding to the dynamic adaptation module and the anomaly correction module respectively according to the adaptation strategy instruction and the correction strategy instruction for parameter iteration to generate the anomaly correction signal; among them, the optimization and correction strategies include load balancing strategies, temperature suppression strategies, and charge and discharge smoothing strategies; A signal fusion module, which is used to perform spatio-temporal alignment operation on the reference monitoring signal and the anomaly correction signal to generate a final control instruction; An instruction issuing module, configured to transmit the final control instruction to a control terminal of an energy storage unit.

7. The AI-based mobile energy storage vehicle operation status monitoring and analysis system according to claim 6, characterized in that, The dynamic adjustment module includes: A first adjustment module, configured to, when it is monitored that the battery temperature corresponding to the reference monitoring signal exceeds a safe range, superimpose a temperature compensation signal within the current monitoring period to generate a first target adjustment signal; A second adjustment module, configured to perform trend prediction correction according to a charge-discharge efficiency decay curve to generate a second target adjustment signal.

8. The AI-based mobile energy storage vehicle operation status monitoring and analysis system according to claim 7, wherein The first adjustment module includes: A steady-state monitoring module, configured to, when it is monitored that the energy storage unit is in steady-state operation, determine whether the deviation between the current load fluctuation and the expected load exceeds a tolerance threshold. If it exceeds, perform a parameter fine-tuning on the load balancing strategy once; if not, maintain the original strategy; A transient response module, configured to, when it is monitored that the energy storage unit is in transient charge-discharge, before the charge-discharge action starts, determine whether the difference between the thermal imaging signal and a preset temperature distribution exceeds a critical value. If it exceeds, perform a dynamic compensation on the temperature suppression strategy once; if not, no compensation is performed; An exception handling module, configured to, when it is monitored that the energy storage unit is under high load impact, perform dynamic adaptation on the charge-discharge smoothing strategy in combination with voltage fluctuation data, and adjust the smoothing parameters once when the voltage fluctuation exceeds a threshold; A single-parameter adjustment module, configured to, when it is monitored that a single sensor signal is abnormal, perform local adaptation on the correction strategy corresponding to the signal within the monitoring period, and maintain the original correction strategies for the remaining signals.

9. The AI-based mobile energy storage vehicle operation status monitoring and analysis system according to claim 7, characterized in that, The second adjustment module includes: A prediction configuration module, configured to activate a trend prediction function for a specific working condition; wherein, the specific working condition includes high load cycle, abnormal temperature gradient, and current mutation interval, and each of the working conditions is configured with a prediction parameter register and a prediction action period register; A log recording module, configured to continuously record the parameter evolution data of each prediction node under the specific working condition after the prediction function is activated; A non-linear prediction module, configured to perform a non-linear compensation once according to the evolution data when the prediction target is reached within a specified period of the specific working condition.

10. The AI-based mobile energy storage vehicle operation status monitoring and analysis system according to claim 2, wherein The data preprocessing module further includes a timestamp synchronization unit, configured to perform clock calibration on the sampling frequencies of different sensors to eliminate the timing offset of multi-source data.

Images