Battery operation and maintenance management platform compatible with multiple data forms
By integrating data acquisition, preprocessing, prediction and monitoring modules in the battery management platform, and using advanced algorithms to perform battery status analysis and early warning, the problems of untimely battery maintenance and unresponsiveness in the existing technology are solved, and the battery life is extended and reduce operating costs.
Patent Information
- Application Number
- CN202411058848.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-02
- Publication Date
- 2024-11-15
AI Technical Summary
The existing battery management platform lacks in-depth analysis and trend prediction capabilities for battery data, resulting in untimely battery maintenance and replacement, increasing operating costs and shortening battery life, and in the meantime response to abnormal battery status is not timely enough to intelligently adjust the strategy.
A battery operation and maintenance management platform compatible with multi-data forms is designed, including data acquisition, preprocessing, time series prediction, status estimation, dynamic threshold setting and real-time monitoring and early warning modules, using support vector machines, multi-layer perception machines and Algorithms such as attention mechanism perform data analysis and prediction, dynamically adjust thresholds and optimizer for adaptive adjustments.
It has achieved a comprehensive understanding of the battery status, timely maintenance and replacement, reduced operating costs, improved battery life, timely discovered abnormal status and notified operation and maintenance personnel, reduced failure risks, and optimized operation and maintenance resource allocation.
Smart Images

Figure CN118962458A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of battery management platforms, and in particular to a battery operation and maintenance management platform compatible with multiple data formats. Background Art
[0002] In modern energy systems, batteries are key components for storing and supplying energy. Their performance and reliability directly affect the efficiency and safety of the entire energy system. With the widespread use of electric vehicles, renewable energy, and portable electronic devices, the demand for efficient and intelligent battery management platforms is growing. These systems need to monitor key battery parameters such as voltage, current, and temperature in real time to ensure that the battery operates in a safe and optimal working state.
[0003] Existing battery management technology is already able to provide basic monitoring and partial prediction functions, but there are still some shortcomings: the existing battery management platform only focuses on basic data collection and monitoring, lacks the ability to conduct in-depth analysis and trend prediction of data, and limits the comprehensive understanding of the battery status. Due to the lack of prediction capabilities, battery maintenance and replacement plans are not timely enough, resulting in increased operating costs and shortened battery life. At the same time, when abnormal battery status is detected, the response of the existing system is often not timely enough, and the operation and maintenance personnel cannot be notified quickly to take measures, which can easily lead to further deterioration of the problem. In addition, the battery management platform lacks self-learning and adaptive adjustment functions and cannot intelligently adjust its strategy according to battery usage and environmental changes. Summary of the invention
[0004] In view of the deficiencies in the prior art, the present invention provides a battery operation and maintenance management platform that is compatible with multiple data formats, has the ability to conduct in-depth analysis and trend prediction of data, has a comprehensive understanding of the battery status, and can promptly maintain and replace the battery, thereby reducing operating costs and increasing the battery life. At the same time, when an abnormal battery status is detected, the operation and maintenance personnel can be quickly notified to take measures to avoid the problem from worsening. In addition, the battery management platform has self-learning and adaptive adjustment functions, and can intelligently adjust its strategy according to battery usage and environmental changes.
[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a battery operation and maintenance management platform compatible with multiple data formats, comprising:
[0006] Data acquisition module, used to collect the battery's current, voltage and temperature operating data in real time;
[0007] A data preprocessing module is used to preprocess the collected data;
[0008] Time series prediction module, which uses trend analysis model and sequence dependency model to predict the future trend of battery current, voltage and temperature indicators;
[0009] The state estimation module uses support vector machines and multi-layer perceptrons to estimate the health status of the battery;
[0010] Dynamic threshold setting module, dynamically adjusts the warning threshold according to historical data and forecast results;
[0011] The real-time monitoring and early warning module monitors the battery status in real time based on dynamic thresholds and triggers an early warning when an abnormality is detected.
[0012] Preferably, the data acquisition module includes:
[0013] Current sensor, used to collect real-time current data of the battery;
[0014] Voltage sensor, used to collect real-time voltage data of the battery;
[0015] Temperature sensor, used to collect real-time temperature data of the battery.
[0016] Preferably, the preprocessing step of the data preprocessing module includes:
[0017] Use wavelet transform to denoise the collected data;
[0018] Use Kalman filter to further remove data noise;
[0019] Linear interpolation was used to fill missing data;
[0020] The data is normalized so that it is distributed in the range of 0 to 1.
[0021] Preferably, the time series prediction module includes:
[0022] The trend analysis model automatically selects the optimal parameter set through the Bayesian optimization method. The specific formula is:
[0023] argmin (p,d,q)
[0024] Among them, L(θ) is the loss function, θ is the parameter set, and D is the historical data set;
[0025] The sequence dependency model uses a multi-layer sequence dependency model unit and an attention mechanism. The calculation formula of the attention mechanism is:
[0026]
[0027] Among them, h t is the hidden state of the sequence dependency model, s t-1 is the state of the previous step, W e ,Ue ,b e are learnable parameters.
[0028] Preferably, the sequence dependency model is further combined with adaptive learning rate adjustment and regularization techniques, including:
[0029] Use the dynamic adjustment optimizer to perform adaptive learning rate adjustment. The calculation formula of the dynamic adjustment optimizer optimization algorithm is:
[0030]
[0031] Among them, g t represents the gradient at time step t, m t and v t are the first-order and second-order momentum estimates for the t-th time step, β1 and β2 are the decay rates of the first-order and second-order momentum, β1 and β2 are set to 0.9 and 0.999, respectively, η is the learning rate, η is set to 0.001, and ∈ is a small constant used to prevent division by zero, ∈ is set to 10 -8 , T represents the total number of iterations;
[0032] Apply random dropout technology to prevent overfitting. The calculation formula of random dropout technology is:
[0033] h' i =h i Dropout (p)
[0034] Among them, h i is the neuron output of the i-th layer, p is the probability of the neuron being set to zero, and the value of p is 0.5;
[0035] Use standardized adjustment technology to enhance the robustness and prediction accuracy of the model. The calculation formula of standardized adjustment technology is:
[0036]
[0037] Among them, x (k) is the input of the kth neuron, μ B and are the mean and variance of the batch data, γ and β are learnable scaling and translation parameters, and ∈ is a small constant added for numerical stability, ∈ is set to 10 -5 .
[0038] Preferably, the state estimation module comprises:
[0039] Support vector machine performs nonlinear estimation, the specific formula is:
[0040]
[0041] Among them, K(xi ,x) is the kernel function, α i , b is the model parameter;
[0042] Multilayer Perceptron, using adaptive weight initialization and optimization algorithm, including:
[0043] h i =ReLU(W i ·h i-1 +b i )
[0044] Among them, W i ,b i are the weights and biases of the i-th layer.
[0045] Preferably, the optimization algorithm of the multilayer perceptron includes:
[0046] Use the Xavier initialization method to initialize weights. The specific formula is:
[0047] W i =Xavier(n inn° ,n out° )
[0048] Among them, n inn° ,n out° is the number of input and output neurons.
[0049] Use a dynamically adjusted optimizer for network training to improve training efficiency and prediction accuracy.
[0050] Preferably, the dynamic threshold setting module dynamically adjusts the warning threshold according to the mean and standard deviation of historical data. The specific formula is:
[0051] T dynamic =μ t +k·σ t
[0052] Among them, μ t is the historical mean, σ t is the standard deviation, and k is the adjustment coefficient.
[0053] Preferably, the monitoring and warning steps of the real-time monitoring and warning module include:
[0054] Compare real-time data with dynamic thresholds and trigger an alert when real-time data exceeds the dynamic threshold;
[0055] The wireless communication module notifies the operation and maintenance personnel to take measures.
[0056] A method for monitoring the state of a flow battery system comprises the following steps:
[0057] Collect real-time operating data of the battery, including current, voltage and temperature;
[0058] Preprocess the collected data, clean it and fill in missing data;
[0059] The time series forecasting module applies trend analysis models and sequence dependency models to predict future trends of key battery indicators;
[0060] Use support vector machines and multi-layer perceptrons to estimate battery health status;
[0061] Dynamically adjust warning thresholds based on historical data and forecast results;
[0062] Monitor the battery status in real time, trigger an alert when the data exceeds the dynamic threshold, and notify the operation and maintenance personnel.
[0063] The present invention provides a battery operation and maintenance management platform compatible with multiple data formats. It has the following beneficial effects:
[0064] 1. Through real-time data acquisition and advanced data processing, the present invention collects the current, voltage and temperature data of the battery in real time and compares them with dynamically adjusted thresholds. The platform can accurately predict the future trend of the battery and monitor the battery status in real time, thereby improving the ability of fault prevention and timely response, preventing possible failures or damage, thereby protecting battery assets and reducing maintenance costs.
[0065] 2. The present invention utilizes a composite model of support vector machines and multi-layer perceptrons. These two algorithms can effectively handle nonlinear problems and complex relationship patterns, thereby providing accurate estimates of battery health status, predicting and diagnosing potential battery problems, and supporting more refined maintenance decisions.
[0066] 3. The present invention uses dynamic threshold setting and dynamic adjustment optimizer to enable the platform to intelligently adjust operation and maintenance strategies according to real-time data and historical trends, reduce human intervention, and optimize operation and maintenance resource allocation.
[0067] 4. The present invention combines real-time monitoring and early warning modules, so that the platform can promptly detect any abnormal state that exceeds the dynamic threshold, and promptly notify the operation and maintenance personnel through wireless communication, thereby improving the efficiency and success rate of handling emergencies. BRIEF DESCRIPTION OF THE DRAWINGS
[0068] Figure 1 is a system structure diagram of the present invention;
[0069] Figure 2 The present invention is a flow chart of the method. DETAILED DESCRIPTION
[0070] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0071] Please refer to the attached Figure 1 The embodiment of the present invention provides a battery operation and maintenance management platform compatible with multiple data formats, including:
[0072] Data acquisition module, used to collect the battery's current, voltage and temperature operating data in real time. These data are the basis for the battery management platform's decision-making process, including but not limited to the battery's charge and discharge status, temperature changes and voltage levels, ensuring that the system can respond to any minor changes in battery performance in a timely manner, which is beneficial to improving battery efficiency and preventive maintenance, thereby extending battery life and reducing maintenance costs;
[0073] The data preprocessing module is used to preprocess the collected data, which can reduce data errors and improve the accuracy of subsequent model predictions. It is crucial to optimize battery performance and reduce false alarms;
[0074] Time series prediction module, which uses trend analysis model and sequence dependency model to predict the future trend of battery current, voltage and temperature indicators. The trend analysis model captures long-term change trends, while the sequence dependency model focuses on short-term fluctuations and cyclical changes. The application of this module helps to identify potential battery problems early, implement preventive maintenance strategies, and avoid expensive repairs and unexpected downtime;
[0075] The state estimation module uses support vector machines and multi-layer perceptrons to estimate the health status of the battery. These two models provide accurate battery status diagnosis. The state estimation module can timely detect battery performance degradation and potential failures, which helps to formulate more effective battery management strategies and improve the overall reliability and efficiency of the system;
[0076] Dynamic threshold setting module dynamically adjusts the warning threshold according to historical data and prediction results, so that the system can flexibly respond to changes in battery status. Dynamic threshold adjustment reduces insignificant alarms, ensures the accuracy and timeliness of alarms, and can respond to real problems more effectively;
[0077] The real-time monitoring and early warning module monitors the battery status in real time based on dynamic thresholds and triggers an early warning when an abnormality is detected. The real-time monitoring and early warning module enhances the ability to respond immediately to battery problems, reduces potential safety risks, and ensures the continuous operation of the energy system and the safe use of batteries.
[0078] Please see attached Figure 1 In a preferred embodiment of the present invention, the data acquisition module includes:
[0079] Current sensor, used to collect real-time current data of the battery. The current sensor measures the current passing through the battery or battery pack through electromagnetic induction. It can accurately capture the current changes of the battery under different working conditions, such as charging, discharging and idle state, monitor the current data in real time to understand the charging and discharging efficiency and state of the battery, and detect current anomalies in time to prevent the battery from overheating, thereby effectively extending the battery life;
[0080] Voltage sensor, used to collect real-time voltage data of the battery. The voltage sensor determines the voltage level by measuring the potential difference between the two ends of the battery. The sensor provides continuous voltage readings that can be used to assess the health status and charge level of the battery. By monitoring voltage changes, the platform can predict the capacity decay and life of the battery while avoiding low or high voltage situations;
[0081] Temperature sensor is used to collect real-time temperature data of the battery. The temperature sensor uses a thermistor to monitor the temperature of the battery. Real-time temperature monitoring can prevent damage to the battery due to overheating and ensure that the battery operates within a safe operating temperature range. In addition, temperature data can also help optimize the battery charging strategy and improve the thermal management system.
[0082] Please see attached Figure 1 In a preferred embodiment of the present invention, the preprocessing step of the data preprocessing module includes:
[0083] Use wavelet transform to denoise the collected data. In the denoising process, wavelet transform decomposes the signal into wavelet coefficients of different scales to isolate the noise and signal. By selectively retaining the wavelet coefficients representing the main characteristics of the signal, reconstructing the data, and removing the high-frequency noise component, this denoising method can significantly improve the signal-to-noise ratio of the data, so that the subsequent model analysis is not interfered by random noise, thereby improving the accuracy of prediction and diagnosis;
[0084] Use Kalman filter to further remove data noise. Kalman filter continuously updates the estimated value and uncertainty (covariance) of system state through prediction model and observation model, and adjusts the estimate in combination with new measurement data. It can effectively smooth transitional or time series data and is suitable for dynamically changing battery monitoring data, further reducing errors and improving data quality.
[0085] Linear interpolation is used to fill missing data. In a data set, there are missing values between two consecutive data points. Linear interpolation can use the values of these two points to estimate the value of the missing point and fill the data gap. A complete data series can provide more accurate trend and pattern analysis;
[0086] The data is normalized so that it is distributed in the range of 0 to 1. Normalization changes the scale of the data by adjusting the range of variables, so that data of different magnitudes and units can be compared and processed on the same scale, improving the subsequent calculation effect of data with the same scale.
[0087] Please refer to the attached Figure 1 In a preferred embodiment of the present invention, the time series prediction module includes:
[0088] The trend analysis model automatically selects the optimal parameter set through the Bayesian optimization method. The specific formula is:
[0089]
[0090] Among them, L(θ) is the loss function, θ is the parameter set, D is the historical data set, and Bayesian optimization is used to automatically select the parameters p, d, q of the trend analysis model to minimize the model's prediction loss L(θ). During the optimization process, the trend analysis model will be continuously updated according to the performance of historical parameter selection, gradually approaching the optimal parameter solution;
[0091] Formula explanation:
[0092]
[0093] Among them, L(θ) represents the prediction loss function of the model under the parameter set θ, θ represents the parameter set (p, d, q), and D is the historical data set used to train and verify the model. The optimal parameters are automatically selected through Bayesian optimization. The trend analysis model can more accurately capture and predict the operating status and trend of the battery. This method reduces the deviation and time consumption of manual parameter selection, while improving the adaptability of the trend analysis model and the accuracy of prediction, and can help managers better understand the use and degradation patterns of batteries, thereby optimizing the use of resources and energy, extending the service life of batteries, and reducing environmental impact;
[0094] The sequence dependency model uses a multi-layer sequence dependency model unit and an attention mechanism. The calculation formula of the attention mechanism is:
[0095]
[0096] Among them, h t is the hidden state of the sequence dependency model, s t-1 is the state of the previous step, W e ,U e ,b eFor learnable parameters, the sequence dependency model refers to the use of a multi-layer recurrent neural network that can process and memorize long-term dependencies in time series data, thereby effectively predicting the future state of the battery;
[0097] The formula of the attention mechanism is:
[0098]
[0099] Among them, h t is the hidden state at the current time step, s t-1 is the state of the previous time step, W e ,U e ,b e are learnable parameters in the model.
[0100] The attention mechanism enables the model to focus on the most relevant information in the sequence, improving the ability and accuracy of processing complex sequence data. The sequence dependency model that introduces the attention mechanism allows the management platform to accurately identify and utilize key information in the battery data, thereby greatly improving the accuracy of the prediction. The management platform can monitor the battery status in real time and respond quickly to possible changes or failures. The immediate response capability can reduce the risk of serious failures and ensure the stable operation of the entire system.
[0101] Please refer to the attached Figure 1 In a preferred embodiment of the present invention, the sequence dependency model is further combined with adaptive learning rate adjustment and regularization techniques, including:
[0102] Use the dynamic adjustment optimizer to perform adaptive learning rate adjustment. The calculation formula of the dynamic adjustment optimizer optimization algorithm is:
[0103]
[0104] Among them, g t represents the gradient at time step t, m t and v t are the first-order and second-order momentum estimates for the t-th time step, β1 and β2 are the decay rates of the first-order and second-order momentum, β1 and β2 are set to 0.9 and 0.999, respectively, η is the learning rate, η is set to 0.001, and ∈ is a small constant used to prevent division by zero, ∈ is set to 10 -8, T represents the total number of iterations. This optimization algorithm combines the momentum method of gradient descent and the adaptive learning rate technology, so that the update of each parameter depends not only on the current gradient, but also on the historical gradient information, thereby smoothing the learning process and accelerating the convergence speed. The dynamic adjustment of the learning rate can automatically adjust the update step size according to the different characteristics of the parameters, improving the optimization efficiency. Using the dynamic adjustment optimizer, the model can converge to the optimal solution faster, and automatically adjust the learning rate during the learning process, avoiding the complexity and limitations of manual adjustment, which helps the model to maintain efficiency and accuracy under various operating conditions, and can provide reliable data support for battery operation and maintenance, helping the operation and maintenance team to formulate more scientific maintenance plans and emergency response strategies. It not only improves the efficiency of battery use, but also reduces maintenance costs and extends the service life of batteries;
[0105] Apply random dropout technology to prevent overfitting. The calculation formula of random dropout technology is:
[0106] h' i =h i Dropout (p)
[0107] Among them, h i is the output of the neuron in the i-th layer, p is the probability that the neuron is set to zero, and the value of p is 0.5. When training a multi-layer perceptron, random dropout technology is applied in the hidden layer, which can reduce the model's dependence on a single training instance, increase the collaboration between different neurons, and make the model more robust when testing new data. By reducing the complex adaptive interactions between neurons during training, random dropout reduces the risk of overfitting of the model and improves the accuracy of overall predictions. Randomly shutting down neurons forces the network to learn more robust features, so that the model must learn to correctly interpret data through a variety of different neuronal pathways. This method increases the robustness of the model, making it more adaptable to changes in various operating conditions in practical applications;
[0108] Use standardized adjustment technology to enhance the robustness and prediction accuracy of the model. The calculation formula of standardized adjustment technology is:
[0109]
[0110] Among them, x (k) is the input of the kth neuron, μ B and are the mean and variance of the batch data, γ and β are learnable scaling and translation parameters, and ∈ is a small constant added for numerical stability, ∈ is set to 10 -5, batch normalization is usually applied after each hidden layer and before the activation function. This method not only standardizes the distribution of the input data, but also makes the layers of the network more independent, reducing the impact of the previous layer parameter update on the later layer, thereby allowing the use of higher learning rates and accelerating the convergence of the network. By reducing internal covariate shift, batch normalization allows the use of larger learning rates without causing instability during model training. This makes the training process faster and reduces the model training time. At the same time, by normalizing the feature distribution of the input layer, it reduces the model's dependence on the specific distribution of the input, thereby improving the model's performance on unknown data.
[0111] Please refer to the attached Figure 1 In a preferred embodiment of the present invention, the state estimation module includes:
[0112] Support vector machine performs nonlinear estimation, the specific formula is:
[0113]
[0114] Among them, K(x i ,x) is the kernel function, α i , b is the model parameter. Support vector machine is suitable for classification and regression tasks. In battery state estimation, support vector machine is used to identify and predict the health state of the battery. Support vector machine introduces kernel function K(x i ,x) to process nonlinear feature space, allowing the model to find the optimal segmentation plane in high-dimensional space;
[0115] The decision function expression is:
[0116]
[0117] where α i is the Lagrange multiplier, b is the bias term. By using appropriate kernel functions, the support vector machine can effectively identify complex nonlinear relationships in battery data and improve the accuracy of diagnosis. At the same time, the optimization of the support vector machine ensures the robustness of the model to outliers and noise, and can efficiently process the actual operation data of the battery.
[0118] Multilayer Perceptron, using adaptive weight initialization and optimization algorithm, including:
[0119] h i =ReLU(W i ·h i-1 +b i )
[0120] Among them, W i ,b iis the weight and bias of the i-th layer. The multilayer perceptron contains multiple hidden layers, each of which is composed of multiple neurons. The multilayer perceptron can capture complex patterns and relationships in the data and is suitable for processing the state estimation problem of multiple parameters of the battery. The multilayer perceptron can learn the high-level features of the input data through its multi-layer structure, which can improve the understanding of the complex behavior patterns of the battery. At the same time, by combining adaptive weight initialization and optimization algorithms, the multilayer perceptron can continuously adjust its parameters during the training process to adapt to the changes in the battery state and enhance the prediction ability and flexibility of the model.
[0121] Please see attached Figure 1 In a preferred embodiment of the present invention, the optimization algorithm of the multilayer perceptron includes:
[0122] Use the Xavier initialization method to initialize weights. The specific formula is:
[0123] W i =Xavier(n inn° ,n out° )
[0124] Among them, n inn° ,n out° Xavier initialization is a weight initialization method used to solve the problem of gradient vanishing or exploding when training deep neural networks. By ensuring that the activation distribution of each layer of the network remains relatively consistent, Xavier initialization helps accelerate convergence and reduce gradient problems in the early stages of training. It also reduces oscillations and instability during training through reasonable initial weight values, making the training process smoother and achieving better results.
[0125] Using a dynamic adjustment optimizer for network training improves training efficiency and prediction accuracy. The dynamic adjustment optimizer combines the advantages of the RMS propagation algorithm and the momentum method, and can automatically adjust the learning rate of each parameter. The dynamic adjustment optimizer can automatically adjust the update speed of each parameter according to its importance, reducing the need to manually adjust the learning rate, which can improve the convergence speed. By finely controlling the gradient descent process, the dynamic adjustment optimizer makes the model more stable during training, ultimately improving the accuracy of predictions.
[0126] Please see attached Figure 1 In a preferred embodiment of the present invention, the dynamic threshold setting module dynamically adjusts the warning threshold according to the mean and standard deviation of historical data. The specific formula is:
[0127] T dynamic =μ t +k·σ t
[0128] Among them, μ tis the historical mean, σ t is the standard deviation, k is the adjustment coefficient, μ t Represents the average performance of the battery over a period of time. In order to capture the overall trend of the battery status, σ t Indicates the variation or fluctuation of battery performance. A larger standard deviation means a higher instability of the battery state. By multiplying the standard deviation by the adjustment factor k and adding it to the mean, a threshold can be set. Exceeding this threshold means that the current state of the battery is significantly different from the normal range in the past. This can help the maintenance team detect problems early and prevent small problems from turning into serious failures. At the same time, it can reduce false alarms or missed alarms caused by improper fixed threshold settings and ensure the accuracy of early warnings. Dynamic thresholds make maintenance responses more accurate, help optimize maintenance plans and resource allocation, thereby reducing unnecessary maintenance costs and extending battery life.
[0129] Please refer to the attached Figure 1 In a preferred embodiment of the present invention, the monitoring and warning steps of the real-time monitoring and warning module include:
[0130] Compare real-time data with dynamic thresholds. When real-time data exceeds the dynamic threshold, trigger an early warning. Collect voltage, current and temperature data from the battery management platform regularly, and then compare these real-time data with the pre-calculated dynamic thresholds. If the real-time data exceeds these thresholds, the system will judge it as an abnormal state, thus further intervening;
[0131] The wireless communication module is used to notify the operation and maintenance personnel to take measures. The wireless communication module is a standard wireless network. Once the real-time data exceeds the dynamic threshold, the wireless communication module will immediately trigger an early warning signal to the operation and maintenance personnel. Through real-time monitoring and timely early warning, the platform can quickly identify and respond to potential faults such as battery performance degradation, reduce system downtime and potential maintenance costs, and significantly improve the stability and reliability of the platform through real-time data monitoring and dynamic threshold analysis, ensuring the continuity and efficiency of battery operation.
[0132] The method for monitoring the status of a liquid flow battery system described below can be referenced to the battery operation and maintenance management platform compatible with multiple data formats described above. Their principles and technical effects are similar and will not be repeated here.
[0133] Please refer to the attached Figure 2 The present invention also provides a method for monitoring the state of a flow battery system, comprising the following steps:
[0134] S1. Collect real-time operating data of the battery, including current, voltage and temperature;
[0135] S2, pre-process the collected data, clean and fill in missing data;
[0136] S3, using the trend analysis model and sequence dependency model to predict the future trend of key indicators of the battery through the time series prediction module;
[0137] S4, using support vector machine and multi-layer perceptron to estimate the battery health state;
[0138] S5. Dynamically adjust the warning threshold based on historical data and forecast results;
[0139] S6. Monitor the battery status in real time, trigger an alert when the data exceeds the dynamic threshold, and notify the operation and maintenance personnel.
[0140] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.
Claims
1. A battery operation and maintenance management platform compatible with multiple data formats, characterized in that: include: Data acquisition module, used to collect the battery's current, voltage and temperature operating data in real time; A data preprocessing module is used to preprocess the collected data; Time series prediction module, which uses trend analysis model and sequence dependency model to predict the future trend of battery current, voltage and temperature indicators; The state estimation module uses support vector machines and multi-layer perceptrons to estimate the health status of the battery; Dynamic threshold setting module, dynamically adjusts the warning threshold according to historical data and forecast results; The real-time monitoring and early warning module monitors the battery status in real time based on dynamic thresholds and triggers an early warning when an abnormality is detected.
2. A battery operation and maintenance management platform compatible with multiple data formats according to claim 1, characterized in that: The data acquisition module comprises: Current sensor, used to collect real-time current data of the battery; Voltage sensor, used to collect real-time voltage data of the battery; Temperature sensor, used to collect real-time temperature data of the battery.
3. A battery operation and maintenance management platform compatible with multiple data formats according to claim 1, characterized in that: The preprocessing steps of the data preprocessing module include: Use wavelet transform to denoise the collected data; Use Kalman filter to further remove data noise; Linear interpolation was used to fill missing data; The data is normalized so that it is distributed in the range of 0 to 1.
4. The battery operation and maintenance management platform compatible with multiple data formats according to claim 1, characterized in that: The time series prediction module includes: The trend analysis model automatically selects the optimal parameter set through the Bayesian optimization method. The specific formula is: Among them, L(θ) is the loss function, θ is the parameter set, and D is the historical data set; The sequence dependency model uses a multi-layer sequence dependency model unit and an attention mechanism. The calculation formula of the attention mechanism is: Among them, h t is the hidden state of the sequence dependency model, s t-1 is the state of the previous step, W e ,U e ,b e are learnable parameters.
5. A battery operation and maintenance management platform compatible with multiple data formats according to claim 4, characterized in that: The sequence dependency model further incorporates adaptive learning rate adjustment and regularization techniques, including: Use the dynamic adjustment optimizer to perform adaptive learning rate adjustment. The calculation formula of the dynamic adjustment optimizer optimization algorithm is: m t =β1m t-1 +(1-β1)g t , Among them, g t represents the gradient at time step t, m t and v t are the first-order and second-order momentum estimates for the t-th time step, β1 and β2 are the decay rates of the first-order and second-order momentum, β1 and β2 are set to 0.9 and 0.999, respectively, η is the learning rate, η is set to 0.001, and ∈ is a small constant used to prevent division by zero, ∈ is set to 10 -8 , T represents the total number of iterations; Apply random dropout technology to prevent overfitting. The calculation formula of random dropout technology is: h' i =h i ·Dropout(p) Among them, h i is the neuron output of the i-th layer, p is the probability of the neuron being set to zero, and the value of p is 0.5; Use standardized adjustment technology to enhance the robustness and prediction accuracy of the model. The calculation formula of standardized adjustment technology is: Among them, x (k) is the input of the kth neuron, μ B and are the mean and variance of the batch data, γ and β are learnable scaling and translation parameters, and ∈ is a small constant added for numerical stability, ∈ is set to 10 -5 .
6. The battery operation and maintenance management platform compatible with multiple data formats according to claim 1, characterized in that: The state estimation module comprises: Support vector machine performs nonlinear estimation, the specific formula is: Among them, K(x i ,x) is the kernel function, α i , b is the model parameter; Multilayer Perceptron, using adaptive weight initialization and optimization algorithm, including: h i =ReLU(W i h i-1 +b i ) Among them, W i ,b i are the weights and biases of the i-th layer.
7. A battery operation and maintenance management platform compatible with multiple data formats according to claim 6, characterized in that: The optimization algorithm of the multi-layer perceptron includes: Use the Xavier initialization method to initialize weights. The specific formula is: W i =Xavier(n inn° ,n out° ) Among them, n inn° ,n out° is the number of input and output neurons; Use a dynamically adjusted optimizer for network training to improve training efficiency and prediction accuracy.
8. The battery operation and maintenance management platform compatible with multiple data formats according to claim 1, characterized in that: The dynamic threshold setting module dynamically adjusts the warning threshold according to the mean and standard deviation of historical data. The specific formula is: T dynamic =μ t +k·s t Among them, μ t is the historical mean, σ t is the standard deviation, and k is the adjustment coefficient.
9. The battery operation and maintenance management platform compatible with multiple data formats according to claim 1, characterized in that: The monitoring and warning steps of the real-time monitoring and warning module include: Compare real-time data with dynamic thresholds and trigger an alert when real-time data exceeds the dynamic threshold; The wireless communication module notifies the operation and maintenance personnel to take measures.
10. A method for monitoring the state of a flow battery system, characterized in that: Using a battery operation and maintenance management platform compatible with multiple data formats as described in any one of claims 1 to 9 comprises the following steps: Collect real-time operating data of the battery, including current, voltage and temperature; Preprocess the collected data, clean it and fill in missing data; The time series forecasting module applies trend analysis models and sequence dependency models to predict future trends of key battery indicators; Use support vector machines and multi-layer perceptrons to estimate battery health status; Dynamically adjust warning thresholds based on historical data and forecast results; Monitor the battery status in real time, trigger an alert when the data exceeds the dynamic threshold, and notify the operation and maintenance personnel.
Citation Information
Cited By
High temperature resistance measuring system for lithium ion battery electrolyte
CN119199368A
Direct-current battery discharge timely power-off method and device
CN119382300A
Method and device for realizing accurate hoisting of steel plate by automatically adjusting magnetism
CN119660526A
Battery thermal runaway and safety threshold dynamic adjustment system and method
CN120327265A
A battery thermal runaway and safety threshold dynamic adjustment system and method
CN120327265B
