Short-circuit protection method and system for energy storage system, medium and equipment
By using LSTM and interpretable neural networks to predict future load power and environmental data and adjust the short-circuit protection threshold of the energy storage system, the problems of slow response and misjudgment in existing technologies are solved, and faster and more accurate short-circuit protection is achieved.
Patent Information
- Application Number
- CN202510810774.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-17
- Publication Date
- 2025-09-16
AI Technical Summary
Existing short-circuit protection methods for energy storage systems have a slow reaction speed and cannot respond to instantaneous short-circuit currents in a timely manner. They also cannot consider misjudgments or delays caused by changes in load power and environmental parameters.
An LSTM network and an interpretable neural network are used to predict future load power and environmental data, adjust the short-circuit protection thresholds of electronic switches at each layer of the energy storage system, and achieve adaptive short-circuit protection.
It achieves faster response speed and higher protection accuracy, adapts to load and environmental changes, reduces misjudgment, and improves the safety and reliability of the energy storage system.
Smart Images
Figure CN120657674A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of energy storage systems, and in particular to a short-circuit protection method, system, medium and equipment for an energy storage system. Background Art
[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.
[0003] With the widespread adoption of renewable energy sources (such as solar and wind power), energy storage systems are playing an increasingly important role in power systems. Energy storage systems can balance grid loads, address the intermittent and volatile nature of renewable energy, and support load regulation during peak power demand. However, energy storage systems can experience short-circuit failures during operation, resulting in battery damage, equipment burnout, and even fires, explosions, and other safety incidents. Therefore, research on short-circuit protection technology is crucial to ensuring the safety of energy storage systems.
[0004] Currently, there are mainly the following types of short-circuit protection methods for energy storage systems:
[0005] 1) Fuse-based short-circuit protection for energy storage systems: The fuse type and specifications are selected based on the energy storage system's rated operating current and short-circuit current. When a short circuit occurs, the fuse generates a large amount of heat and melts, isolating the fault. However, this method suffers from slow response, inability to respond promptly to transient short-circuit currents, and the inability to reuse fuses.
[0006] 2) Energy storage system short-circuit protection based on electronic switches: High-precision current sensors monitor the energy storage system's current in real time. If the detected current exceeds a set threshold, the electronic switch is quickly disconnected. Electronic switches offer fast response times, effectively reducing delays caused by slow fuses. However, the threshold setting fails to account for load power and environmental parameter variations, making it prone to misjudgment or slow response.
[0007] Based on the defects of the above methods, how to design a short-circuit protection method that has a fast response speed, is reusable, and takes into account the working status of the energy storage system has become an urgent problem to be solved in the existing technology. Summary of the Invention
[0008] In response to the shortcomings of the existing technology, the purpose of the present invention is to provide a short-circuit protection method, system, medium and equipment for energy storage systems, which can adaptively adjust the short-circuit protection threshold according to changes in environmental parameters and load, achieve more accurate short-circuit protection, and realize safe and reliable operation of the energy storage system.
[0009] In order to achieve the above object, the present invention is implemented through the following technical solutions:
[0010] A first aspect of the present invention provides a short-circuit protection method for an energy storage system, comprising the following steps:
[0011] Obtain current energy storage system operating status data and environmental data;
[0012] Using a current prediction model to predict the current of each layer of the energy storage system based on the current energy storage system operating status data and environmental data to obtain a current prediction result, wherein the current prediction model includes a first prediction model and a second prediction model. The first prediction model is used to capture the time relationship of the energy storage system operating status data to obtain a predicted future load power. The second prediction model is used to process the environmental data and the predicted future load power to obtain a current prediction result;
[0013] The short-circuit protection thresholds of the electronic switches at each layer of the energy storage system are adjusted based on the current prediction results.
[0014] Furthermore, the energy storage system operating status data includes load power and current data of each layer of the energy storage system, and the environmental data includes temperature and humidity.
[0015] Furthermore, the training steps of the current prediction model are:
[0016] Collect historical operating status data and environmental data of the energy storage system and perform preprocessing;
[0017] Using the collected load power data from the historical operating status data of the energy storage system to train a first prediction model based on a gradient descent method;
[0018] The second prediction model is trained by fitting the corresponding relationship between the load power, environmental data and currents at different levels in the historical operating status data of the energy storage system.
[0019] Furthermore, the first prediction model adopts LSTM network, and the second prediction model adopts interpretable neural network.
[0020] Furthermore, the preprocessing steps include:
[0021] Normalize the historical operating status data and environmental data of the energy storage system.
[0022] Furthermore, the historical operating status data and environmental data of the energy storage system are continuously collected to update the training set, and the current prediction model is regularly optimized based on the updated training set.
[0023] Furthermore, the specific steps for adjusting the short-circuit protection thresholds of the electronic switches at each layer of the energy storage system based on the current prediction results are as follows:
[0024] The fault correlation coefficient is introduced and the protection threshold of the electronic switch is calculated based on the current prediction result and the fault correlation coefficient;
[0025] The electronic switches at each layer of the energy storage system are protected according to the electronic switch protection threshold.
[0026] A second aspect of the present invention provides an energy storage system short-circuit protection system, comprising:
[0027] A data acquisition module is configured to acquire current energy storage system operating status data and environmental data;
[0028] a current prediction module configured to use a current prediction model to predict the current of each layer of the energy storage system based on the current energy storage system operating status data and environmental data to obtain a current prediction result, wherein the current prediction model includes a first prediction model and a second prediction model, the first prediction model is used to capture the time relationship of the energy storage system operating status data to obtain a predicted future load power, and the second prediction model is used to process the environmental data and the predicted future load power to obtain the current prediction result;
[0029] The threshold adjustment module is configured to adjust the short-circuit protection threshold of the electronic switches at each layer of the energy storage system based on the current prediction result.
[0030] A third aspect of the present invention provides a medium having a program stored thereon, which, when executed by a processor, implements the steps of the energy storage system short-circuit protection method as described in the first aspect of the present invention.
[0031] The fourth aspect of the present invention provides a device comprising a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps in the energy storage system short-circuit protection method as described in the first aspect of the present invention are implemented.
[0032] One or more of the above technical solutions have the following beneficial effects:
[0033] The present invention discloses a short-circuit protection method, system, medium and equipment for an energy storage system. The method uses collected load power data to train an LSTM model to predict future load power data. The method uses collected load power, environmental parameters and current data at each level to construct an interpretable neural network to predict future currents at each level of the system. The method then adaptively controls the short-circuit protection thresholds at each level of the system based on the predicted current, thereby achieving faster response speed and higher protection accuracy.
[0034] The present invention can adaptively adjust the short-circuit protection threshold according to changes in environmental parameters and load, thereby achieving more accurate short-circuit protection and having a faster response speed.
[0035] The present invention first uses a long short-term memory neural network to predict future load changes, and then predicts the future current of each level of the system based on environmental parameters and the predicted future load power. The adjustment based on the predicted current has a faster diagnostic speed and higher diagnostic accuracy.
[0036] The present invention can analyze the importance and influence of various parameters on current estimation, help better understand model decision-making, and provide a useful reference for engineers to understand the impact of various factors on the operation of energy storage systems.
[0037] The method of the present invention has data-driven characteristics and can be conveniently used for short-circuit protection of different types of DC systems without the need to construct an energy storage system model to calculate the rated current and short-circuit current and the relatively complex selection of fuses.
[0038] Advantages of additional aspects of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.
[0040] Figure 1 This is a flow chart of a short-circuit protection method for an energy storage system in Embodiment 1 of the present invention;
[0041] Figure 2 This is a flow chart of the current prediction model operation in Example 1 of the present invention. DETAILED DESCRIPTION
[0042] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.
[0043] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or their combinations;
[0044] Example 1:
[0045] The first embodiment of the present invention provides a short-circuit protection method for an energy storage system, comprising the following steps: obtaining current energy storage system operating status data and environmental data; using a current prediction model to predict the current of each layer of the energy storage system based on the current energy storage system operating status data and environmental data, and obtaining a current prediction result. Figure 2 As shown, the current prediction model includes a first prediction model composed of a long short-term memory neural network (LSTM) model and a second prediction model composed of an interpretable neural network. The first prediction model is used to capture the time relationship of the load power data in the energy storage system operation status data to obtain the predicted future load power. The second prediction model is used to analyze and process the humidity, temperature and predicted future load power in the environmental data based on the interpretable neural network GANDE. According to the analysis results, the parameters are removed and retrained to obtain the future current prediction results of each level; based on the current prediction results, the short-circuit protection threshold of the electronic switch of each layer of the energy storage system is adjusted.
[0046] like Figure 1 The specific steps are as follows:
[0047] S1: Collect historical operating status data and environmental data of the energy storage system.
[0048] S2: Use the collected load power data to train a long short-term memory neural network (LSTM) to predict future load power.
[0049] S3: Use an interpretable neural network to fit the corresponding relationship between environmental parameters, load power, and current at each level of the energy storage system, and analyze the impact of each parameter on the current. Based on the analysis results, remove parameters with little impact on the estimated current and retrain the interpretable neural network.
[0050] S4: Based on the latest collected data, the trained LSTM and interpretable neural network model is used to predict the future currents at each level of the system. The short-circuit protection threshold of the electronic switch is adjusted based on the predicted module-level currents.
[0051] S5: Continuously collect historical operating status data and environmental data of the energy storage system to update the training set, and regularly optimize the model based on the updated training set to ensure the effectiveness of the threshold setting.
[0052] In S1, the energy storage system operating status data includes but is not limited to load power and current and voltage data of each layer of the energy storage system. The environmental data includes but is not limited to temperature and humidity, which can be adjusted according to actual influencing parameters. The collection interval d is set according to actual needs.
[0053] In S2 and S3, only LSTM is used to predict load power because environmental parameters such as temperature and humidity change slowly and the change at the next collection point is small.
[0054] The training steps of the current prediction model are:
[0055] Historical operating status and environmental data of the energy storage system are collected and preprocessed. The preprocessing step includes normalizing the historical operating status and environmental data. A first prediction model is trained using the gradient descent method using the load power data from the collected historical operating status data. Different second prediction models are constructed to fit the corresponding relationships between the load power and environmental data from the historical operating status data and the current at different levels. The first prediction model uses an LSTM network, while the second prediction model uses an interpretable neural network.
[0056] The specific steps include:
[0057] S21: For the load data in the historical operating status data of the energy storage system, use a denoising method including but not limited to the median absolute deviation (MAD) denoising method to remove gross errors in the data, and normalize the load power data.
[0058] S22: Use the sliding window to process the load power data to obtain the load power data sequence P t :P t ={P(t),P(t+d),P(t+2d)…,P(t+kd)}.
[0059] Where: k is the length of the sequence extracted by the sliding window, P(t) represents the load power at time t. t and P t+d The samples used to train the model are obtained as input and labels of the LSTM model respectively.
[0060] S23: Using the obtained samples, a gradient descent method is used to train an LSTM model as the first prediction model, so that it can use the existing load power data to predict future load power data.
[0061] S3: Train the second prediction model by fitting the corresponding relationship between the load power, environmental data and different levels of current in the historical operating status data of the energy storage system.
[0062] Compared to conventional neural networks, interpretable neural networks can clearly explain how the neural network reaches a conclusion or prediction result, and which features or inputs play what role in the final output. In this embodiment, the interpretable neural network can help understand the impact of each parameter on the current estimation, improve the transparency of the neural network, and increase the trust in the prediction results when making decisions. More specifically, the interpretable neural network can not only accurately estimate the current at each level of the energy storage system, but also analyze the impact of each parameter on the current estimation, such as importance, and the influence relationship between each parameter and the current (positive correlation, negative correlation, etc.).
[0063] The interpretable neural network used is the generalized additive neural decision ensemble (GANDE), which is composed of the generalized additive model (GAM) and the neural independent decision ensemble (NODE).
[0064] GAM is a classic and effective statistical model, and its formula is as follows:
[0065]
[0066] In the formula, y represents the output of the model, x i represents the i-th input, f i is a smooth function, μ is a learnable intercept term, and g represents an unspecified monotonic link function (such as the identity function or the logarithmic function). The unique structure of GAM makes it possible to enhance the understanding of the model and its decision-making process by visualizing the impact of each feature.
[0067] NODE is an efficient deep learning architecture designed to handle nonlinear regression problems. NODE consists of N differentiable oblivious decision trees (ODTs) of L layers and depth d. The operation mode of ODT is similar to the traditional decision tree model, but in ODT, all nodes at the same depth level share features and thresholds. Specifically, ODT selects d features and thresholds, splits the data at each layer of the tree, and generates 2 d possible responses. Following a similar design to DenseNet, the output ODT(x) of the previous layer is used as the input of the subsequent layer. This means that the input of layer l can be expressed as:
[0068] x l =[x 1 ,ODT 1 (x 1 ),ODT 2 (x 2 ),…,ODT l-1 (x l-1 )].
[0069] Where, ODT lis the output of all ODTs in layer l. In summary, for a NODE with L layers, each layer consists of N ODT trees, and the output of the NODE is the average of the outputs of all trees in the model, as shown below:
[0070]
[0071] Where, ODT ln Indicates the nth ODT in the lth layer of the NODE.
[0072] GAM, due to its unique structure, offers good interpretability, while NODE offers superior performance in relational mapping. Combining these two features allows for current prediction while maintaining good interpretability. Specifically, to achieve this, the original NODE structure needs to be modified so that each tree is partitioned based on a single feature (such as load power). Only trees with the same feature can be connected between different layers. Finally, the weighted sum of the outputs of all trees is taken to obtain the final output value: the estimated current.
[0073] The training data is:
[0074] S31: In the training data, the input is various environmental parameters and load power:
[0075] Input=[T,W,U,P].
[0076] Where Input is the model input, T, W, and U are temperature, humidity, and voltage, respectively. Labels are the corresponding temperatures at each level.
[0077] S32: After training, visualization results of the importance and influence relationship of each parameter on current estimation can be obtained.
[0078] S33: Based on the analysis results of GANDE, we can conclude the importance of each parameter to current estimation. Removing the least important parameters and retraining GANDE can effectively reduce the complexity of the model.
[0079] This embodiment uses a compensable interpretable neural network. Since it does not have the ability to predict time series specifically, directly using it to predict future parameters is insufficient. Therefore, combined with LSTM, it can achieve the prediction of future current while analyzing the impact of various characteristic parameters on the current.
[0080] In S4, S41: using the trained first prediction model to use the newly collected load power data to predict future load power.
[0081] S42: Use the trained second prediction model to predict the future system current I at each level using the newly collected temperature and humidity data and the future load power predicted by the first prediction model.
[0082] S43: Introduce a fault correlation coefficient and calculate the electronic switch protection threshold according to the current prediction result and the fault correlation coefficient.
[0083] Among them, the electronic switch protection threshold J = αI, (α>1), α is the fault correlation coefficient, and the selection needs to balance the short-circuit fault response speed and fault misjudgment rate.
[0084] S44: Protect the electronic switches at each layer of the energy storage system according to the electronic switch protection threshold.
[0085] Example 2:
[0086] A second embodiment of the present invention provides an energy storage system short-circuit protection system, including:
[0087] A data acquisition module is configured to acquire current energy storage system operating status data and environmental data;
[0088] a current prediction module configured to use a current prediction model to predict the current of each layer of the energy storage system based on the current energy storage system operating status data and environmental data to obtain a current prediction result, wherein the current prediction model includes a first prediction model and a second prediction model, the first prediction model is used to capture the time relationship of the energy storage system operating status data to obtain a predicted future load power, and the second prediction model is used to process the environmental data and the predicted future load power to obtain the current prediction result;
[0089] The threshold adjustment module is configured to adjust the short-circuit protection threshold of the electronic switches at each layer of the energy storage system based on the current prediction result.
[0090] Example 3:
[0091] A third embodiment of the present invention provides a medium having a program stored thereon. When the program is executed by a processor, the steps of the energy storage system short-circuit protection method described in the first embodiment of the present invention are implemented.
[0092] Example 4:
[0093] A fourth embodiment of the present invention provides a device comprising a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps of the energy storage system short-circuit protection method as described in the first embodiment of the present invention are implemented.
[0094] The steps involved in the above embodiments 2, 3 and 4 correspond to those in the method embodiment 1. For the specific implementation methods, please refer to the relevant description part of the embodiment 1.
[0095] Those skilled in the art will appreciate that the modules or steps of the present invention described above can be implemented using a general-purpose computer device. Alternatively, they can be implemented using program code executable by a computing device, which can then be stored in a storage device and executed by the computing device. Alternatively, they can be fabricated into separate integrated circuit modules, or multiple modules or steps can be fabricated into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.
[0096] Although the above describes the specific embodiments of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without any creative work are still within the scope of protection of the present invention.
Claims
1. A short-circuit protection method for an energy storage system, characterized in that: The following steps are involved: Obtain current energy storage system operating status data and environmental data; Using a current prediction model to predict the current of each layer of the energy storage system based on the current energy storage system operating status data and environmental data to obtain a current prediction result, wherein the current prediction model includes a first prediction model and a second prediction model. The first prediction model is used to capture the time relationship of the energy storage system operating status data to obtain a predicted future load power. The second prediction model is used to process the environmental data and the predicted future load power to obtain a current prediction result; The short-circuit protection thresholds of the electronic switches at each layer of the energy storage system are adjusted based on the current prediction results.
2. The energy storage system short-circuit protection method according to claim 1, characterized in that: The energy storage system operating status data includes load power and current data of each layer of the energy storage system, and the environmental data includes temperature and humidity.
3. The short-circuit protection method for an energy storage system according to claim 1, wherein: The training steps of the current prediction model are: Collect historical operating status data and environmental data of the energy storage system and perform preprocessing; Using the collected load power data from the historical operating status data of the energy storage system to train a first prediction model based on a gradient descent method; The second prediction model is trained by fitting the corresponding relationship between the load power, environmental data and currents at different levels in the historical operating status data of the energy storage system.
4. The energy storage system short-circuit protection method according to claim 1, characterized in that: The first prediction model uses LSTM network, and the second prediction model uses interpretable neural network.
5. The short-circuit protection method for an energy storage system according to claim 3, wherein: The preprocessing steps include: Normalize the historical operating status data and environmental data of the energy storage system.
6. The energy storage system short-circuit protection method according to claim 1, characterized in that: Continuously collect historical operating status data and environmental data of the energy storage system to update the training set, and regularly optimize the current prediction model based on the updated training set.
7. The short-circuit protection method for an energy storage system according to claim 1, wherein: The specific steps for adjusting the short-circuit protection thresholds of the electronic switches at each layer of the energy storage system based on the current prediction results are as follows: The fault correlation coefficient is introduced and the protection threshold of the electronic switch is calculated based on the current prediction result and the fault correlation coefficient; The electronic switches at each layer of the energy storage system are protected according to the electronic switch protection threshold.
8. A short-circuit protection system for an energy storage system, characterized in that: include: A data acquisition module is configured to acquire current energy storage system operating status data and environmental data; a current prediction module configured to use a current prediction model to predict the current of each layer of the energy storage system based on the current energy storage system operating status data and environmental data to obtain a current prediction result, wherein the current prediction model includes a first prediction model and a second prediction model, the first prediction model is used to capture the time relationship of the energy storage system operating status data to obtain a predicted future load power, and the second prediction model is used to process the environmental data and the predicted future load power to obtain the current prediction result; The threshold adjustment module is configured to adjust the short-circuit protection threshold of the electronic switches at each layer of the energy storage system based on the current prediction result.
9. A computer-readable storage medium, characterized in that A plurality of instructions are stored therein, and the instructions are suitable for being loaded by a processor of a terminal device and executed by the energy storage system short-circuit protection method according to any one of claims 1 to 7.
10. A terminal device, characterized in that: The method comprises a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, wherein the instructions are suitable for being loaded by the processor and executing the energy storage system short-circuit protection method according to any one of claims 1 to 7.