An automatic operation test method and system based on energy storage unit

Abstract

The present invention discloses an automated operation test method and system based on an energy storage unit, comprising the following steps: S1, constructing a complex virtual situation simulation environment, generating a virtual model in real time and synchronizing the physical state; S2, generating a multi-dimensional operation data stream; S3, analyzing the operation state of the energy storage unit through a self-learning algorithm, constructing a dynamic behavior model of the energy storage unit based on a neural network converter and a capsule network, and generating a behavior pattern mapping of the energy storage unit; S4, autonomously optimizing the test process using a hierarchical deep deterministic policy gradient algorithm, automatically generating a new test path and adjusting the test conditions; S5, based on the new test path, the system automatically applies behavior-inducing pressure to the energy storage unit; S6, combining Bayesian deep learning with variational inference algorithms to predict the performance of the energy storage unit in future working conditions; S7, establishing a virtual-real coupling system. The present invention combines multi-physical field simulation and behavior-inducing testing, etc., to achieve intelligent and precise testing of energy storage units.

Description

Technical Field

[0001] The present invention relates to the technical field of energy storage unit testing, and in particular to an automatic operation testing method and system based on an energy storage unit. Background Art

[0002] With the growth of modern energy demand, energy storage units play a vital role in applications such as power systems, electric vehicles, and renewable energy storage. The performance of energy storage units under actual working conditions is directly related to the safety, reliability, and economy of the system. Therefore, comprehensive, accurate, and efficient testing and verification of energy storage units has become one of the key links in research and development. The current energy storage unit testing technology is mainly carried out in the laboratory, and the testing methods are often relatively simple. Usually, the testing method under static conditions is adopted, such as charging and discharging cycles under constant temperature and constant current conditions to evaluate the capacity, internal resistance, efficiency, and life of the energy storage unit. However, this testing method often cannot fully reflect the true performance of the energy storage unit in actual application scenarios.

[0003] Existing energy storage unit testing technologies face challenges in many aspects. First, the performance of energy storage units in complex actual working conditions is the result of multi-dimensional and multi-physical field coupling, involving the interaction of many factors such as electromagnetic fields, thermal fields, mechanical stress fields, and electrochemical reactions. However, most existing technologies are unable to effectively simulate and integrate these multi-dimensional factors, and can usually only be tested under a single working condition, resulting in a large difference between the test results and the actual use conditions, making it difficult to truly reflect the performance and degradation trend of energy storage units under different working conditions.

[0004] Secondly, the existing test process generally relies on preset static test conditions and lacks the ability to provide real-time feedback on the operating status of the energy storage unit. These tests usually only measure and record pre-defined parameters, and then analyze the results after the test is completed. This static test method makes it impossible to adjust the test strategy in real time during the test process, and it is impossible to make adaptive test adjustments based on the real-time status of the energy storage unit. For example, the energy storage unit may show potential signs of failure under certain extreme conditions, such as voltage fluctuations, increased internal resistance, etc., but because the existing test methods cannot actively guide and accelerate the exposure of faults during the test process, the test efficiency is low, the cycle is long, and it is difficult to effectively discover potential problems with the energy storage unit.

[0005] Third, existing test systems generally lack intelligence and automation capabilities, and mainly rely on manual design of test plans and adjustment of test conditions, and are unable to optimize the test process through self-learning. Existing test systems often rely on experience or pre-set test processes, and are unable to dynamically capture the behavioral characteristics of energy storage units in complex operating environments, and it is also difficult to predict and detect failure modes early. In addition, traditional test data processing methods mainly focus on the collection and analysis of static data, lack the ability to comprehensively analyze multi-dimensional and multi-time series data, and it is difficult to extract the behavior patterns and degradation laws of energy storage units from large amounts of data. This has resulted in the inability of existing testing technologies to effectively extract deep information from the operating data of energy storage units, missing the opportunity for early warning of failures and performance degradation.

[0006] Based on this, the existing technology has significant deficiencies in the following aspects: First, it is impossible to effectively simulate the dynamic operating environment of the energy storage unit under complex multi-dimensional working conditions, and the test conditions are out of touch with the actual application scenarios, resulting in insufficient accuracy and comprehensiveness of the test results; second, the test process is rigid and lacks real-time feedback and adaptive adjustment capabilities, which cannot effectively improve the test efficiency and fault detection accuracy; third, the existing test system lacks intelligence and automation capabilities, and is unable to extract in-depth information from the multi-dimensional operating data of the energy storage unit, making it difficult to accurately predict and optimize performance.

[0007] Therefore, how to provide an automated operation test method and system based on an energy storage unit is a problem that technical personnel in this field urgently need to solve. Summary of the invention

[0008] One purpose of the present invention is to propose an automated operation test method and system based on energy storage units. The present invention constructs an intelligent and automated test system suitable for energy storage units through multi-physics field simulation technology, self-learning algorithm, behavior-induced stress test and big data analysis technology. The system can simulate the operating status of energy storage units under complex working conditions in real time, accurately capture their behavioral characteristics, and realize accelerated exposure and fault warning of potential failure modes of energy storage units by dynamically optimizing the test process. Compared with traditional testing methods, the present invention has high efficiency, intelligence and accuracy, can greatly improve the accuracy and efficiency of testing, and ensure the long-term safety and stability of energy storage units.

[0009] An automatic operation test method based on an energy storage unit according to an embodiment of the present invention comprises the following steps:

[0010] S1. Build a complex virtual scenario simulation environment through multi-physics field simulation technology and data twin technology, generate a virtual model of the energy storage unit in real time and synchronize the physical state of the energy storage unit, and simulate the operating environment under complex dynamic conditions, including multi-dimensional electromagnetic fields, thermal fields, mechanical stress fields and chemical reaction conditions;

[0011] S2, using a distributed sensor network to collect electrochemical parameters of energy storage units in real time, and transmit them to edge computing nodes through the Internet of Things technology for local processing to generate multi-dimensional operation data streams;

[0012] S3, based on the generated multi-dimensional operation data stream, analyze the operation status of the energy storage unit through a self-learning algorithm, build a dynamic behavior model of the energy storage unit based on a neural network converter and a capsule network, and generate a behavior pattern mapping of the energy storage unit;

[0013] S4. Combining the test results of the virtual scenario and the dynamic behavior model, the layered deep deterministic policy gradient algorithm is used to autonomously optimize the test process. After identifying the behavior pattern of the energy storage unit, a new test path is automatically generated and the test conditions are adjusted;

[0014] S5. Based on the new test path, the system automatically applies behavior-induced pressure to the energy storage unit, actively triggers potential failure or degradation behavior, and updates the behavior model according to the actual response of the energy storage unit under the induced pressure;

[0015] S6. Use big data analysis technology and time series analysis methods to analyze the multidimensional data generated during the induction test, extract the behavioral characteristics of the energy storage unit under different behavioral induction pressure conditions, and combine Bayesian deep learning and variational inference algorithms to predict the performance of the energy storage unit in future working conditions;

[0016] S7. Combine virtual scenario simulation with actual physical testing to establish a virtual-reality coupling system. Through two-way data interaction, the system applies the test results in the virtual scenario to the actual energy storage unit, adjusts the physical test conditions in real time, and uses physical test feedback to update the virtual model to form a closed-loop test process.

[0017] Optionally, the S1 specifically includes:

[0018] S11. Using multi-physics field simulation technology to establish a virtual situational simulation environment of the energy storage unit under complex dynamic conditions, wherein the multi-physics field includes electromagnetic field, thermal field, mechanical stress field and chemical reaction conditions;

[0019] S12, electromagnetic field simulation by solving Maxwell equations:

[0020]

[0021] in, represents the curl operator, represents the electric field strength, represents the magnetic induction intensity, represents the magnetic field strength, represents the current density, represents the electric displacement vector, represents the partial derivative with respect to time;

[0022] S13, thermal field simulation through heat conduction equation:

[0023]

[0024] Where ρ represents the material density, c p represents specific heat capacity, T represents temperature, represents the divergence operator, k represents the thermal conductivity, and Q represents the volume heat source;

[0025] S14. Mechanical stress field simulation is based on linear elastic mechanics equations:

[0026] σ ij =C ijkl ε kl ;

[0027] Among them, σ ij represents the stress tensor, C ijkl represents the elastic constant, ε kl represents the strain tensor, i, j, k and l represent the spatial coordinate index;

[0028] S15. Chemical reaction condition simulation uses Nernst-Planck equation:

[0029]

[0030] Among them, G i represents the concentration of component i, represents the flux of component i, R i Express the reaction rate of component i:

[0031]

[0032] Among them, A i represents the diffusion coefficient of component i, z i represents the charge number of component i, F represents the Faraday constant, φ represents the electric potential, represents the potential gradient, V represents the gas constant;

[0033] S16. Using data twin technology, a virtual model of the energy storage unit is generated in real time, and the physical state of the energy storage unit is synchronized through a sensor network;

[0034] S17. In a virtual scenario simulation environment, set up complex dynamic working conditions, including high temperature, low temperature, overcharge, over discharge, mechanical shock and vibration extreme conditions, to simulate the operating behavior of the energy storage unit in different environments.

[0035] Optionally, the S3 specifically includes:

[0036] S31, based on the multi-dimensional operation data stream generated in step S2, using a self-learning algorithm to analyze the operation status of the energy storage unit, wherein the self-learning algorithm uses a deep learning model that integrates a neural network converter and a capsule network to dynamically construct a behavior pattern mapping of the energy storage unit;

[0037] S32. In the neural network converter, a multi-head self-attention mechanism is used to model the multi-dimensional data of energy storage units from different sources:

[0038]

[0039] Among them, Attention represents the multi-head attention mechanism, Q chem Represents the chemical reaction parameter query matrix of the energy storage unit, K temp The bond matrix representing the temperature field data of the energy storage unit, V volt The value matrix representing the voltage fluctuation data of the energy storage unit, d key represents the dimension of the key vector, and softmax represents the probability distribution function;

[0040] S33, the sequence features output by the multi-head self-attention mechanism are passed to the capsule network, which further processes the multi-dimensional data through the dynamic routing algorithm to capture the spatial hierarchical relationship of the energy storage unit during operation:

[0041]

[0042]

[0043] Among them, s j represents the sum of the inputs of high-level capsule j, c ij represents the dynamic coupling coefficient from capsule i to capsule j, represents the prediction vector of the lower-level capsule i for the higher-level capsule j, represents the weight matrix describing the chemical behavior, Represents the voltage behavior vector output by the lower-level capsule i;

[0044] S34. Perform nonlinear activation on the output of the capsule through the Squash compression function:

[0045]

[0046] Among them, v j represents the output vector of high-level capsule j, and ∥·∥ represents the Euclidean norm;

[0047] S35. Update the dynamic coupling coefficient through a dynamic routing algorithm:

[0048]

[0049] Among them, b ij represents the coupling coefficient between capsule i and capsule j, c ij ′ represents the updated dynamic coupling coefficient from capsule i to capsule j, b ik represents the coupling coefficient between capsule i and capsule k, where k represents the index of the high-level capsule;

[0050] S36, generating a behavior pattern mapping of the energy storage unit, wherein the behavior pattern mapping is adaptively optimized through a feedback mechanism, and when an abnormal behavior pattern of the energy storage unit is detected, the weight matrix in the model and the coupling coefficient c ij Automatic adjustment.

[0051] Optionally, the S4 specifically includes:

[0052] S41, based on the dynamic behavior model of the energy storage unit generated in step S3, the test process is autonomously optimized using a hierarchical deep deterministic policy gradient algorithm, where the hierarchical policy network consists of a high-level global policy network and a low-level local policy network;

[0053] S42: In the high-level global policy network, define a global state S global and global action A global , the global state includes the multi-dimensional operating parameters of the energy storage unit, the global action includes the adjustment of the global test conditions, and the global strategy update formula is:

[0054]

[0055] Among them, π global (S global ) represents the global policy function, α global represents the learning rate of the global policy network, represents the gradient of the global policy network, J global represents the expected return of the global strategy;

[0056] S43, in the low-level local strategy network, define a local state S local and local action A local , the local state represents the operating state of the energy storage unit voltage change rate or temperature gradient, the local action includes instantaneous load or local temperature adjustment, and the local strategy update formula is:

[0057]

[0058] Among them, π local (S local ,A global ) represents the local strategy function, αlocal represents the local learning rate, represents the gradient of the local policy network, J local represents the expected return of the local strategy;

[0059] S44, the global strategy network is based on the overall operating status S of the energy storage unit global Generate global test conditions;

[0060] S45, global action A generated by the local strategy network according to the global strategy global , combined with the local state S local , generate specific local test parameters:

[0061]

[0062] Among them, A(t) represents the test action generated at time t, A global (S global ) represents the test action generated by the global strategy, A local (S local ,A global ) represents the test action generated by the local strategy, t threshold Indicates the time threshold for the collaboration between the global strategy and the local strategy;

[0063] S46. The system dynamically adjusts the test path and test conditions, and generates new test plans in real time.

[0064] Optionally, the S5 specifically includes:

[0065] S51, based on the test path and test conditions generated in step S4, the system automatically applies behavior-induced pressure to the energy storage unit, wherein the behavior-induced pressure includes adjusting the charge and discharge rate I charge , apply burst load P load And control the ambient temperature T env ;

[0066] S52, by adjusting the charge and discharge rate I charge , the system simulates the operation of the energy storage unit under extreme conditions:

[0067] I charge (t) = I base +ΔI·sin(ω1t);

[0068] Among them, I charge (t) represents the charging current at time t, I base represents the basic charging current, ΔI represents the amplitude of change of the charging current, and ω1 represents the frequency of fluctuation of the charging and discharging rate;

[0069] S53, applying a burst load Pload It is used to simulate the operation of energy storage units under instantaneous load changes. The load regulation is achieved through optimized piecewise functions:

[0070]

[0071] Among them, P load (t) Instantaneous load at time t, P base represents the base load, ΔP represents the increment of the burst load, λ represents the constant of the load change rate, t spike represents the start time of the load burst, and Δt represents the duration of the load burst;

[0072] S54, by controlling the ambient temperature T env , the system simulates the response of the energy storage unit under extreme temperature conditions:

[0073] T env (t) = T base +ΔT·(sin(ω2t)+cos(ω2t));

[0074] Among them, T env (t) represents the ambient temperature at time t, T base represents the basic ambient temperature, ΔT represents the temperature variation amplitude, and ω2 represents the temperature fluctuation frequency;

[0075] S55, real-time monitoring of the response of the energy storage unit under different induced pressures, the system collects the operating parameters of the energy storage unit through the sensor network, and dynamically adjusts the application timing and intensity of the induced pressure through a feedback mechanism;

[0076] S56. Based on the update of the real-time behavior model, the system automatically identifies the failure mode or degradation trend of the energy storage unit and adaptively adjusts the induction conditions.

[0077] Optionally, the S6 specifically includes:

[0078] S61, based on the behavior-induced pressure application process of step S5, the system collects the operating data of the energy storage unit in real time and performs preprocessing;

[0079] S62. Using big data analysis technology, the system conducts in-depth analysis on the pre-processed data, uses time series analysis methods to capture the time-dependent characteristics of energy storage units under different behavior-induced pressures, and uses long short-term memory networks to model the time series model;

[0080] S63. Based on the time correlation features extracted by the long short-term memory network model, the behavior characteristics of the energy storage unit under different induced conditions are systematically analyzed;

[0081] S64, the system uses Bayesian deep learning combined with variational inference to predict performance, and the Bayesian model quantifies uncertainty:

[0082] p(y|x,D)=∫p(y|x,θ)p(θ|D)dθ;

[0083] Where p(y|x,D) represents the predicted probability of output y given input x and data D, θ represents the model parameters, p(θ|D) represents the posterior distribution of the model parameters given data D, and p(y|x,θ) represents the model's prediction of output y under parameter θ.

[0084] S65. Through the Bayesian inference model, the system predicts the performance of the energy storage unit under future working conditions, including the prediction of the voltage change, internal resistance growth and temperature rise trend under specific induced conditions in the future;

[0085] S66. Based on the prediction results, the system adjusts the test process in real time.

[0086] An automatic operation test system based on an energy storage unit according to an embodiment of the present invention includes the following modules:

[0087] Virtual situation simulation module, used to build a virtual situation simulation environment through multi-physics field simulation technology and data twin technology;

[0088] A data acquisition module, used to collect electrochemical parameters of energy storage units in real time through a distributed sensor network and generate multi-dimensional operation data streams;

[0089] Behavior modeling module, used to analyze the operating status of the energy storage unit through a self-learning algorithm and generate a dynamic behavior model;

[0090] The test optimization module is used to optimize the test process using a hierarchical deep deterministic policy gradient algorithm and generate new test paths and test conditions;

[0091] A behavior induction module, used to apply behavior induction pressure and update the behavior model according to the actual response of the energy storage unit;

[0092] Performance prediction module, which combines big data analysis and Bayesian deep learning algorithms to predict the future performance of energy storage units;

[0093] The virtual-real coupling module is used to synchronize the test results with the actual working conditions through virtual-real coupling.

[0094] The beneficial effects of the present invention are:

[0095] First, the present invention constructs a complex virtual scenario simulation environment through multi-physics field simulation technology and data twin technology, so that the energy storage unit can experience multi-dimensional dynamic conditions that are closer to actual applications in the simulation environment, including electromagnetic fields, thermal fields, mechanical stress fields, and chemical reaction conditions. The construction of this virtual environment makes the test conditions of the energy storage unit no longer limited to the single static conditions of the laboratory, but can cover more complex scenarios, so as to more comprehensively and accurately evaluate the performance and potential failure points of the energy storage unit under extreme conditions, and improve the authenticity and coverage of the test results.

[0096] Secondly, the present invention collects the electrochemical parameters of the energy storage unit in real time through a distributed sensor network, and combines the Internet of Things technology for data transmission and edge computing, thereby realizing accurate monitoring of the real-time status of the energy storage unit and data stream generation. This function enables dynamic analysis of the operating status of the energy storage unit based on real-time data during the test process, provides high-quality input data for subsequent self-learning algorithms, ensures that the test process can be adjusted at any time according to the current status, and avoids the problem of information lag in traditional testing processes.

[0097] In addition, the present invention constructs a dynamic behavior model of the energy storage unit by using a neural network converter and a capsule network, and generates a behavior pattern mapping, which can conduct an in-depth analysis of the behavior characteristics of the energy storage unit under different operating conditions. This self-learning behavior mapping model can not only identify the normal operating characteristics of the energy storage unit, but also capture its subtle changes and potential fault characteristics under complex working conditions, realizing the intelligent transformation of testing from data acquisition to behavior analysis. By generating and optimizing the test path in real time, the system can automatically adjust the test parameters, such as charge and discharge rate, temperature, etc., to ensure that each step of the test is adaptively adjusted according to the current state of the energy storage unit, thereby significantly improving the flexibility and accuracy of the test.

[0098] Furthermore, the behavior induction module provides strong support for the test efficiency and fault detection accuracy of the present invention. By actively applying induced pressure, the system can accelerate the exposure of potential failure modes of the energy storage unit. This behavior-induced test mechanism makes the test no longer rely on the traditional passive observation of faults, but can actively guide the system into a fault state, thereby greatly shortening the test cycle and effectively improving the test efficiency. Especially under extreme conditions such as high voltage, high temperature or sudden changes in charge and discharge rates, the system can more quickly identify the degradation trend and failure mode of the energy storage unit, and predict the limit behavior of the energy storage unit in advance.

[0099] Finally, through big data analysis technology and time series analysis methods, the system can extract the multi-dimensional characteristics of energy storage units under different behavior-induced stress conditions, combine Bayesian deep learning and variational inference algorithms to predict performance, and further optimize the test plan. This prediction function enables the system to dynamically adjust the test strategy based on current data, identify possible future failure risks of energy storage units in advance, and optimize the test path, greatly improving the pertinence and preventiveness of the test. In particular, through the virtual-real coupling system combined with actual physical testing, the system can adjust the actual physical test conditions in real time based on the virtual scenario test results, and update the virtual model based on physical feedback, forming a closed-loop optimization process to ensure the accuracy and real-time performance of the test. BRIEF DESCRIPTION OF THE DRAWINGS

[0100] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0101] Figure 1 A flowchart of an automated operation test method based on an energy storage unit proposed by the present invention;

[0102] Figure 2 This is a structural schematic diagram of an automatic operation test system based on an energy storage unit proposed by the present invention. DETAILED DESCRIPTION

[0103] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0104] refer to Figure 1 , an automated operation test method based on an energy storage unit, comprising the following steps:

[0105] S1. Build a complex virtual scenario simulation environment through multi-physics field simulation technology and data twin technology, generate a virtual model of the energy storage unit in real time and synchronize the physical state of the energy storage unit, and simulate the operating environment under complex dynamic conditions, including multi-dimensional electromagnetic fields, thermal fields, mechanical stress fields and chemical reaction conditions;

[0106] S2, using a distributed sensor network to collect electrochemical parameters of energy storage units in real time, and transmit them to edge computing nodes through the Internet of Things technology for local processing to generate multi-dimensional operation data streams;

[0107] S3, based on the generated multi-dimensional operation data stream, analyze the operation status of the energy storage unit through a self-learning algorithm, build a dynamic behavior model of the energy storage unit based on a neural network converter and a capsule network, and generate a behavior pattern mapping of the energy storage unit;

[0108] S4. Combining the test results of the virtual scenario and the dynamic behavior model, the layered deep deterministic policy gradient algorithm is used to autonomously optimize the test process. After identifying the behavior pattern of the energy storage unit, a new test path is automatically generated and the test conditions are adjusted;

[0109] S5. Based on the new test path, the system automatically applies behavior-induced pressure to the energy storage unit, actively triggers potential failure or degradation behavior, and updates the behavior model according to the actual response of the energy storage unit under the induced pressure;

[0110] S6. Use big data analysis technology and time series analysis methods to analyze the multidimensional data generated during the induction test, extract the behavioral characteristics of the energy storage unit under different behavioral induction pressure conditions, and combine Bayesian deep learning and variational inference algorithms to predict the performance of the energy storage unit in future working conditions;

[0111] S7. Combine virtual scenario simulation with actual physical testing to establish a virtual-reality coupling system. Through two-way data interaction, the system applies the test results in the virtual scenario to the actual energy storage unit, adjusts the physical test conditions in real time, and uses physical test feedback to update the virtual model to form a closed-loop test process.

[0112] In this implementation, S1 specifically includes:

[0113] S11. Using multi-physics field simulation technology to establish a virtual situational simulation environment of the energy storage unit under complex dynamic conditions, wherein the multi-physics field includes electromagnetic field, thermal field, mechanical stress field and chemical reaction conditions;

[0114] S12, electromagnetic field simulation by solving Maxwell equations:

[0115]

[0116] in, represents the curl operator, represents the electric field strength, represents the magnetic induction intensity, represents the magnetic field strength, represents the current density, represents the electric displacement vector, represents the partial derivative with respect to time;

[0117] S13, thermal field simulation through heat conduction equation:

[0118]

[0119] Where ρ represents the material density, c p represents specific heat capacity, T represents temperature, represents the divergence operator, k represents the thermal conductivity, and Q represents the volume heat source;

[0120] S14. Mechanical stress field simulation is based on linear elastic mechanics equations:

[0121] σ ij =C ijkl ε kl ;

[0122] Among them, σ ij represents the stress tensor, C ijkl represents the elastic constant, ε kl represents the strain tensor, i, j, k and l represent the spatial coordinate index;

[0123] S15. Chemical reaction condition simulation uses Nernst-Planck equation:

[0124]

[0125] Among them, G i represents the concentration of component i, represents the flux of component i, R i Express the reaction rate of component i:

[0126]

[0127] Among them, A i represents the diffusion coefficient of component i, z i represents the charge number of component i, F represents the Faraday constant, φ represents the electric potential, represents the potential gradient, V represents the gas constant;

[0128] S16. Using data twin technology, a virtual model of the energy storage unit is generated in real time, and the physical state of the energy storage unit is synchronized through a sensor network;

[0129] S17. In a virtual scenario simulation environment, set up complex dynamic working conditions, including high temperature, low temperature, overcharge, over discharge, mechanical shock and vibration extreme conditions, to simulate the operating behavior of the energy storage unit in different environments.

[0130] In this implementation, S3 specifically includes:

[0131] S31, based on the multi-dimensional operation data stream generated in step S2, using a self-learning algorithm to analyze the operation status of the energy storage unit, wherein the self-learning algorithm uses a deep learning model that integrates a neural network converter and a capsule network to dynamically construct a behavior pattern mapping of the energy storage unit;

[0132] S32. In the neural network converter, a multi-head self-attention mechanism is used to model the multi-dimensional data of energy storage units from different sources:

[0133]

[0134] Among them, Attention represents the multi-head attention mechanism, Q chem Represents the chemical reaction parameter query matrix of the energy storage unit, K temp The bond matrix representing the temperature field data of the energy storage unit, V volt The value matrix representing the voltage fluctuation data of the energy storage unit, d key represents the dimension of the key vector, and softmax represents the probability distribution function;

[0135] S33, the sequence features output by the multi-head self-attention mechanism are passed to the capsule network, which further processes the multi-dimensional data through the dynamic routing algorithm to capture the spatial hierarchical relationship of the energy storage unit during operation:

[0136]

[0137] Among them, s j represents the sum of the inputs of high-level capsule j, c ij represents the dynamic coupling coefficient from capsule i to capsule j, represents the prediction vector of the lower-level capsule i for the higher-level capsule j, represents the weight matrix describing the chemical behavior, Represents the voltage behavior vector output by the lower-level capsule i;

[0138] S34. Perform nonlinear activation on the output of the capsule through the Squash compression function:

[0139]

[0140] Among them, v j represents the output vector of high-level capsule j, and ∥·∥ represents the Euclidean norm;

[0141] S35. Update the dynamic coupling coefficient through a dynamic routing algorithm:

[0142]

[0143] Among them, b ij represents the coupling coefficient between capsule i and capsule j, c ij ′ represents the updated dynamic coupling coefficient from capsule i to capsule j, b ik represents the coupling coefficient between capsule i and capsule k, where k represents the index of the high-level capsule;

[0144] S36, generating a behavior pattern mapping of the energy storage unit, wherein the behavior pattern mapping is adaptively optimized through a feedback mechanism, and when an abnormal behavior pattern of the energy storage unit is detected, the weight matrix in the model and the coupling coefficient c ij Automatic adjustment.

[0145] In this implementation manner, the S4 specifically includes:

[0146] S41, based on the dynamic behavior model of the energy storage unit generated in step S3, the test process is autonomously optimized using a hierarchical deep deterministic policy gradient algorithm, where the hierarchical policy network consists of a high-level global policy network and a low-level local policy network;

[0147] S42: In the high-level global policy network, define a global state S global and global action A global , the global state includes the multi-dimensional operating parameters of the energy storage unit, the global action includes the adjustment of the global test conditions, and the global strategy update formula is:

[0148]

[0149] Among them, π global (S global ) represents the global policy function, α global represents the learning rate of the global policy network, represents the gradient of the global policy network, J global represents the expected return of the global strategy;

[0150] S43, in the low-level local strategy network, define a local state S local and local action A local , the local state represents the operating state of the energy storage unit voltage change rate or temperature gradient, the local action includes instantaneous load or local temperature adjustment, and the local strategy update formula is:

[0151]

[0152] Among them, π local (S local ,A global ) represents the local strategy function, α local represents the local learning rate, represents the gradient of the local policy network, J local represents the expected return of the local strategy;

[0153] S44, the global strategy network is based on the overall operating status S of the energy storage unit global Generate global test conditions;

[0154] S45, global action A generated by the local strategy network according to the global strategy global , combined with the local state S local , generate specific local test parameters:

[0155]

[0156] Among them, A(t) represents the test action generated at time t, A global (S global ) represents the test action generated by the global strategy, A local (S local ,A global ) represents the test action generated by the local strategy, t threshold Indicates the time threshold for the collaboration between the global strategy and the local strategy;

[0157] S46. The system dynamically adjusts the test path and test conditions, and generates new test plans in real time.

[0158] In this implementation manner, S5 specifically includes:

[0159] S51, based on the test path and test conditions generated in step S4, the system automatically applies behavior-induced pressure to the energy storage unit, wherein the behavior-induced pressure includes adjusting the charge and discharge rate I charge , apply burst load P load And control the ambient temperature T env ;

[0160] S52, by adjusting the charge and discharge rate I charge , the system simulates the operation of the energy storage unit under extreme conditions:

[0161] I charge (t) = I base +ΔI·sin(ω1t);

[0162] Among them, I charge (t) represents the charging current at time t, I base represents the basic charging current, ΔI represents the amplitude of change of the charging current, and ω1 represents the frequency of fluctuation of the charging and discharging rate;

[0163] S53, applying a burst load P load It is used to simulate the operation of energy storage units under instantaneous load changes. The load regulation is achieved through optimized piecewise functions:

[0164]

[0165] Among them, P load (t) Instantaneous load at time t, P baserepresents the base load, ΔP represents the increment of the burst load, λ represents the constant of the load change rate, t spike represents the start time of the load burst, and Δt represents the duration of the load burst;

[0166] S54, by controlling the ambient temperature T env , the system simulates the response of the energy storage unit under extreme temperature conditions:

[0167] T env (t) = T base +ΔT·(sin(ω2t)+cos(ω2t));

[0168] Among them, T env (t) represents the ambient temperature at time t, T base represents the basic ambient temperature, ΔT represents the temperature variation amplitude, and ω2 represents the temperature fluctuation frequency;

[0169] S55, real-time monitoring of the response of the energy storage unit under different induced pressures, the system collects the operating parameters of the energy storage unit through the sensor network, and dynamically adjusts the application timing and intensity of the induced pressure through a feedback mechanism;

[0170] S56. Based on the update of the real-time behavior model, the system automatically identifies the failure mode or degradation trend of the energy storage unit and adaptively adjusts the induction conditions.

[0171] In this implementation manner, S6 specifically includes:

[0172] S61, based on the behavior-induced pressure application process of step S5, the system collects the operating data of the energy storage unit in real time and performs preprocessing;

[0173] S62. Using big data analysis technology, the system conducts in-depth analysis on the pre-processed data, uses time series analysis methods to capture the time-dependent characteristics of energy storage units under different behavior-induced pressures, and uses long short-term memory networks to model the time series model;

[0174] S63. Based on the time correlation features extracted by the long short-term memory network model, the behavior characteristics of the energy storage unit under different induced conditions are systematically analyzed;

[0175] S64, the system uses Bayesian deep learning combined with variational inference to predict performance, and the Bayesian model quantifies uncertainty:

[0176] p(y|x,D)=∫p(y|x,θ)p(θ|D)dθ;

[0177] Where p(y|x,D) represents the predicted probability of output y given input x and data D, θ represents the model parameters, p(θ|D) represents the posterior distribution of the model parameters given data D, and p(y|x,θ) represents the model's prediction of output y under parameter θ.

[0178] S65. Through the Bayesian inference model, the system predicts the performance of the energy storage unit under future working conditions, including the prediction of the voltage change, internal resistance growth and temperature rise trend under specific induced conditions in the future;

[0179] S66. Based on the prediction results, the system adjusts the test process in real time.

[0180] refer to Figure 2 , an automatic operation test system based on energy storage unit, including the following modules:

[0181] Virtual situation simulation module, used to build a virtual situation simulation environment through multi-physics field simulation technology and data twin technology;

[0182] A data acquisition module, used to collect electrochemical parameters of energy storage units in real time through a distributed sensor network and generate multi-dimensional operation data streams;

[0183] Behavior modeling module, used to analyze the operating status of the energy storage unit through a self-learning algorithm and generate a dynamic behavior model;

[0184] The test optimization module is used to optimize the test process using a hierarchical deep deterministic policy gradient algorithm and generate new test paths and test conditions;

[0185] A behavior induction module, used to apply behavior induction pressure and update the behavior model according to the actual response of the energy storage unit;

[0186] Performance prediction module, which combines big data analysis and Bayesian deep learning algorithms to predict the future performance of energy storage units;

[0187] The virtual-real coupling module is used to synchronize the test results with the actual working conditions through virtual-real coupling.

[0188] Embodiment 1:

[0189] In order to verify the feasibility of the present invention in implementation, the present invention was applied to the test laboratory of an electric vehicle manufacturer. In order to evaluate the performance of the new energy storage battery under different extreme working conditions, it was decided to use the automated operation test system based on the energy storage unit provided by the present invention to conduct a comprehensive test on this batch of batteries. The energy storage battery is mainly used as a power source for electric vehicles. It is necessary to maintain stable performance under various complex operating conditions and ensure safety and durability. Traditional testing methods usually rely on a single static test in the laboratory, such as charge and discharge cycles under constant temperature and constant current conditions, which makes it difficult to fully evaluate the performance of the battery in actual application scenarios. In order to make up for this shortcoming, the manufacturer decided to adopt the automated operation test system of the present invention, which aims to simulate complex real-world conditions, quickly reveal the potential failure modes of energy storage units, and make accurate predictions of their long-term performance.

[0190] The test scenario simulates the complex situations that electric vehicles may encounter during actual operation, such as high-frequency charging and discharging at high speeds, extreme temperature changes in different seasons and climate conditions (such as hot desert environments and cold mountain climates), and frequent starts and stops in urban traffic. Through the virtual scenario simulation module, the system first constructs a complex operating environment based on multi-physics field simulation. The environment includes multi-dimensional working conditions such as electromagnetic fields, thermal fields, mechanical stress fields, and chemical reactions, and synchronizes the physical state of the energy storage battery. The electrochemical parameters monitored by real-time sensors, such as battery voltage, temperature, internal resistance, and current, are transmitted to the data acquisition module and processed locally, providing a basis for subsequent data analysis.

[0191] In this test scenario, the system's self-learning algorithm analyzed the real-time transmitted battery data through the neural network converter and capsule network, and established a behavior model of the battery under different operating conditions. In this process, the system identified abnormal voltage fluctuations in the battery during high-speed driving, and a significant increase in internal resistance under extreme temperature changes. If these problems are not detected in time during actual use, they may cause battery failure or even safety accidents.

[0192] To solve this problem, the system automatically generates an optimized test path, combines the strategy of the test optimization module, and applies behavior-induced pressure. The system proactively adjusts the battery's charge and discharge rates at critical moments, and accelerates the degradation process through burst load simulation. For example, during high-frequency charging, the system increases the charging rate from 2C to 4C. Under multiple load bursts, the battery's internal resistance increases by 8% and the voltage fluctuation increases by 6%. Under extreme temperatures, the system adjusts the ambient temperature from normal 25°C to 70°C and -30°C, respectively. In high-temperature environments, the battery's thermal field responds slowly and the temperature rises by 12%, while in low-temperature environments, the charge and discharge efficiency decreases by 15%. These data show that behavior-induced testing can effectively accelerate the exposure of potential failure modes, making the test more targeted.

[0193] Through the system's performance prediction module, combined with time series analysis and Bayesian deep learning algorithms, testers are able to extract battery degradation trends from the data. For example, after 60 rapid charge and discharge cycles at high temperatures, the system predicts that the battery's charging capacity will drop by 15% under the same conditions for the next 200 times, and obvious voltage anomalies will occur after 480 charge and discharge cycles. This prediction enables testers to make adjustments in actual applications to ensure that the battery can maintain stable performance within a predetermined period of time in electric vehicle applications.

[0194] In this test, through the virtual-reality coupling module, the system was able to dynamically adjust the conditions of the actual physical test. During each test, the data generated by the virtual scenario simulation was bidirectionally synchronized with the actual test data. The system updated the virtual model based on physical feedback and adjusted the physical test conditions in real time. Specifically, after the 30th load burst test, the physical test data showed that the internal resistance growth trend of the battery exceeded the predicted value. The system immediately adjusted the subsequent load test parameters and reduced the burst load frequency from 10Hz to 7Hz to ensure that the battery would not fail prematurely under high load conditions. After 150 charge and discharge cycles, the thermal field and voltage fluctuation data of the battery showed that the data error between the adjusted test path and the virtual scenario prediction was less than 3%. Table 1 shows the experimental data of the performance changes of the energy storage unit under complex working conditions.

[0195] Table 1 Experimental data of performance changes of energy storage units under complex working conditions

[0196]

[0197] Through this test, the electric car manufacturer not only solved the problem of static test limitations in traditional tests, but also successfully revealed the potential failure modes of batteries under complex working conditions. Specific data show that the system can effectively accelerate the battery degradation process and shorten the fault exposure time to 60% of traditional tests by optimizing tests for different loads, temperatures, and charge and discharge rates. The system's performance prediction accuracy is above 98%, and the error between the prediction results and the actual test results is extremely small. This not only greatly shortens the test time, but also provides important data support for the long-term safe operation of the battery.

[0198] Through this application, it can be proved that the energy storage unit automated testing system of the present invention can quickly and accurately identify the performance changes of the battery under complex dynamic conditions, and effectively improve the efficiency and accuracy of the test through self-learning algorithms and big data analysis technology. The system's behavioral induction function also accelerates the exposure of potential battery problems, making the test results more realistic and instructive. In the end, the test cycle was shortened from the traditional 8 months to 4 months, which greatly reduced the testing cost and ensured the reliability and safety of the battery in practical applications.

[0199] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. An automated operation test method based on an energy storage unit, characterized in that: The steps include: S1. Build a complex virtual scenario simulation environment through multi-physics field simulation technology and data twin technology, generate a virtual model of the energy storage unit in real time and synchronize the physical state of the energy storage unit, and simulate the operating environment under complex dynamic conditions, including multi-dimensional electromagnetic fields, thermal fields, mechanical stress fields and chemical reaction conditions; S2, using a distributed sensor network to collect electrochemical parameters of energy storage units in real time, and transmit them to edge computing nodes through the Internet of Things technology for local processing to generate multi-dimensional operation data streams; S3, based on the generated multi-dimensional operation data stream, analyze the operation status of the energy storage unit through a self-learning algorithm, build a dynamic behavior model of the energy storage unit based on a neural network converter and a capsule network, and generate a behavior pattern mapping of the energy storage unit; S4. Combining the test results of the virtual scenario and the dynamic behavior model, the layered deep deterministic policy gradient algorithm is used to autonomously optimize the test process. After identifying the behavior pattern of the energy storage unit, a new test path is automatically generated and the test conditions are adjusted; S5. Based on the new test path, the system automatically applies behavior-induced pressure to the energy storage unit, actively triggers potential failure or degradation behavior, and updates the behavior model according to the actual response of the energy storage unit under the induced pressure; S6. Use big data analysis technology and time series analysis methods to analyze the multidimensional data generated during the induction test, extract the behavioral characteristics of the energy storage unit under different behavioral induction pressure conditions, and combine Bayesian deep learning and variational inference algorithms to predict the performance of the energy storage unit in future working conditions; S7. Combine virtual scenario simulation with actual physical testing to establish a virtual-real coupling system. Through two-way data interaction, the system applies the test results in the virtual scenario to the actual energy storage unit, adjusts the physical test conditions in real time, and uses physical test feedback to update the virtual model to form a closed-loop test process. The S3 specifically includes: S31, based on the multi-dimensional operation data stream generated in step S2, using a self-learning algorithm to analyze the operation status of the energy storage unit, wherein the self-learning algorithm uses a deep learning model that integrates a neural network converter and a capsule network to dynamically construct a behavior pattern mapping of the energy storage unit; S32. In the neural network converter, a multi-head self-attention mechanism is used to model the multi-dimensional data of energy storage units from different sources: Among them, Attention represents the multi-head attention mechanism, Q chem Represents the chemical reaction parameter query matrix of the energy storage unit, K temp The bond matrix representing the temperature field data of the energy storage unit, V volt The value matrix representing the voltage fluctuation data of the energy storage unit, d key represents the dimension of the key vector, and softmax represents the probability distribution function; S33, the sequence features output by the multi-head self-attention mechanism are passed to the capsule network, which further processes the multi-dimensional data through the dynamic routing algorithm to capture the spatial hierarchical relationship of the energy storage unit during operation: Among them, s j represents the sum of the inputs of high-level capsule j, c ij represents the dynamic coupling coefficient from capsule i to capsule j, represents the prediction vector of the lower-level capsule i for the higher-level capsule j, represents the weight matrix describing the chemical behavior, Represents the voltage behavior vector output by the lower-level capsule i; S34. Perform nonlinear activation on the output of the capsule through the Squash compression function: Among them, v j represents the output vector of high-level capsule j, and ∥·∥ represents the Euclidean norm; S35. Update the dynamic coupling coefficient through a dynamic routing algorithm: Among them, b ij represents the coupling coefficient between capsule i and capsule j, c ij ' represents the updated dynamic coupling coefficient from capsule i to capsule j, b ik represents the coupling coefficient between capsule i and capsule k, where k represents the index of the high-level capsule; S36, generating a behavior pattern mapping of the energy storage unit, wherein the behavior pattern mapping is adaptively optimized through a feedback mechanism, and when an abnormal behavior pattern of the energy storage unit is detected, the weight matrix in the model and the coupling coefficient c ij Automatic adjustment; The S4 specifically includes: S41, based on the dynamic behavior model of the energy storage unit generated in step S3, the test process is autonomously optimized using a hierarchical deep deterministic policy gradient algorithm, where the hierarchical policy network consists of a high-level global policy network and a low-level local policy network; S42: In the high-level global policy network, define a global state S global and global action A global , the global state includes the multi-dimensional operating parameters of the energy storage unit, the global action includes the adjustment of the global test conditions, and the global strategy update formula is: Among them, π global (S global ) represents the global policy function, α global represents the learning rate of the global policy network, represents the gradient of the global policy network, J global represents the expected return of the global strategy; S43, in the low-level local strategy network, define a local state S local and local action A local , the local state represents the operating state of the energy storage unit voltage change rate or temperature gradient, the local action includes instantaneous load or local temperature adjustment, and the local strategy update formula is: Among them, π local (S local ,A global ) represents the local strategy function, α local represents the local learning rate, represents the gradient of the local policy network, J local represents the expected return of the local strategy; S44, the global strategy network is based on the overall operating status S of the energy storage unit global Generate global test conditions; S45, global action A generated by the local strategy network according to the global strategy global , combined with the local state S local , generate specific local test parameters: Among them, A(t) represents the test action generated at time t, A global (S global ) represents the test action generated by the global strategy, A local (S local ,A global ) represents the test action generated by the local strategy, t threshold Indicates the time threshold for the collaboration between the global strategy and the local strategy; S46. The system dynamically adjusts the test path and test conditions, and generates new test plans in real time.

2. The automatic operation test method based on the energy storage unit according to claim 1, characterized in that: The S1 specifically includes: S11. Using multi-physics field simulation technology to establish a virtual situational simulation environment of the energy storage unit under complex dynamic conditions, wherein the multi-physics field includes electromagnetic field, thermal field, mechanical stress field and chemical reaction conditions; S12, electromagnetic field simulation by solving Maxwell equations: in, represents the curl operator, represents the electric field strength, represents the magnetic induction intensity, represents the magnetic field strength, represents the current density, represents the electric displacement vector, represents the partial derivative with respect to time; S13, thermal field simulation through heat conduction equation: Where ρ represents the material density, c p represents specific heat capacity, T represents temperature, represents the divergence operator, k represents the thermal conductivity, and Q represents the volume heat source; S14. Mechanical stress field simulation is based on linear elastic mechanics equations: s ij =C ijkl e kl ; Among them, σ ij represents the stress tensor, C ijkl represents the elastic constant, ε kl represents the strain tensor, i, j, k and l represent the spatial coordinate index; S15. Chemical reaction condition simulation uses Nernst-Planck equation: Among them, G i represents the concentration of component i, represents the flux of component i, R i Express the reaction rate of component i: Among them, A i represents the diffusion coefficient of component i, z i represents the charge number of component i, F represents the Faraday constant, φ represents the electric potential, represents the potential gradient, V represents the gas constant; S16. Using data twin technology, a virtual model of the energy storage unit is generated in real time, and the physical state of the energy storage unit is synchronized through a sensor network; S17. In a virtual scenario simulation environment, set up complex dynamic working conditions, including high temperature, low temperature, overcharge, over discharge, mechanical shock and vibration extreme conditions, to simulate the operating behavior of the energy storage unit in different environments.

3. The automatic operation test method based on the energy storage unit according to claim 1 is characterized in that: The S5 specifically includes: S51, based on the test path and test conditions generated in step S4, the system automatically applies behavior-induced pressure to the energy storage unit, wherein the behavior-induced pressure includes adjusting the charge and discharge rate I charge , apply burst load P load And control the ambient temperature T env ; S52, by adjusting the charge and discharge rate I charge , the system simulates the operation of the energy storage unit under extreme conditions: I charge (t)=I base +ΔI·sin(ω1t); Among them, I charge (t) represents the charging current at time t, I base represents the basic charging current, ΔI represents the amplitude of change of the charging current, and ω1 represents the frequency of fluctuation of the charging and discharging rate; S53, applying a burst load P load It is used to simulate the operation of energy storage units under instantaneous load changes. The load regulation is achieved through optimized piecewise functions: Among them, P load (t) Instantaneous load at time t, P base represents the base load, ΔP represents the increment of the burst load, λ represents the constant of the load change rate, t spike represents the start time of the load burst, and Δt represents the duration of the load burst; S54, by controlling the ambient temperature T env , the system simulates the response of the energy storage unit under extreme temperature conditions: T env (t)=T base +ΔT·(sin(ω2t)+cos(ω2t)); Among them, T env (t) represents the ambient temperature at time t, T base represents the basic ambient temperature, ΔT represents the temperature variation amplitude, and ω2 represents the temperature fluctuation frequency; S55, real-time monitoring of the response of the energy storage unit under different induced pressures, the system collects the operating parameters of the energy storage unit through the sensor network, and dynamically adjusts the application timing and intensity of the induced pressure through a feedback mechanism; S56. Based on the update of the real-time behavior model, the system automatically identifies the failure mode or degradation trend of the energy storage unit and adaptively adjusts the induction conditions.

4. The automatic operation test method based on the energy storage unit according to claim 1, characterized in that: The S6 specifically includes: S61, based on the behavior-induced pressure application process of step S5, the system collects the operating data of the energy storage unit in real time and performs preprocessing; S62. Using big data analysis technology, the system conducts in-depth analysis on the pre-processed data, uses time series analysis methods to capture the time-dependent characteristics of energy storage units under different behavior-induced pressures, and uses long short-term memory networks to model the time series model; S63. Based on the time correlation features extracted by the long short-term memory network model, the behavior characteristics of the energy storage unit under different induced conditions are systematically analyzed; S64, the system uses Bayesian deep learning combined with variational inference to predict performance, and the Bayesian model quantifies uncertainty: p(y|x,D)=∫p(y|x,θ)p(θ|D)dθ; Where p(y|x,D) represents the predicted probability of output y given input x and data D, θ represents the model parameters, p(θ|D) represents the posterior distribution of the model parameters given data D, and p(y|x,θ) represents the model's prediction of output y under parameter θ. S65. Through the Bayesian inference model, the system predicts the performance of the energy storage unit under future working conditions, including the prediction of the voltage change, internal resistance growth and temperature rise trend under specific induced conditions in the future; S66. Based on the prediction results, the system adjusts the test process in real time.

5. An automatic operation test system based on an energy storage unit, executing the automatic operation test method based on an energy storage unit according to any one of claims 1 to 4, characterized in that: Includes the following modules: Virtual situation simulation module, used to build a virtual situation simulation environment through multi-physics field simulation technology and data twin technology; A data acquisition module, used to collect electrochemical parameters of energy storage units in real time through a distributed sensor network and generate multi-dimensional operation data streams; Behavior modeling module, used to analyze the operating status of the energy storage unit through a self-learning algorithm and generate a dynamic behavior model; The test optimization module is used to optimize the test process using a hierarchical deep deterministic policy gradient algorithm and generate new test paths and test conditions; A behavior induction module, used to apply behavior induction pressure and update the behavior model according to the actual response of the energy storage unit; Performance prediction module, which combines big data analysis and Bayesian deep learning algorithms to predict the future performance of energy storage units; The virtual-real coupling module is used to synchronize the test results with the actual working conditions through virtual-real coupling.

Images