State detection method of electric vehicle battery energy storage system based on digital twinning

Abstract

The invention relates to an electric vehicle battery energy storage system state detection method based on digital twinning, comprising the following steps: step 1, constructing a BESS digital twinning and entity model structure to realize more efficient state monitoring and parameter optimization; 2, building a digital model based on the physical structure and the physical property according to the physical structure and the physical property of the BESS, wherein the digital model comprises digital models of a battery unit and a battery pack; 3, building a cloud platform based on a machine learning algorithm, collecting and storing a large amount of BESS operation data, performing parameter optimization on the model by using the machine learning algorithm, and feeding back an optimization result to the digital model; step 4, realizing SOC and SOH detection of the BESS system by using a digital twinborn model; the method has the advantages that the digital twinborn model is constructed, optimization is performed through a machine learning algorithm, the accuracy is improved, and accurate monitoring is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of electric vehicle energy storage systems, and specifically relates to a state detection method for an electric vehicle battery energy storage system based on digital twins. Background Art

[0002] With the transformation of the global energy structure and the rapid development of renewable energy, the power system is undergoing unprecedented changes. Renewable energy such as wind and solar energy has gradually become an important part of the power system due to its clean and renewable characteristics. However, the intermittent and unstable nature of these energy sources has brought challenges to the stable operation of the power grid. Battery energy storage system (BESS) is considered to be one of the key technologies to solve this problem because it can provide multiple functions such as frequency regulation, peak regulation, and emergency backup. Therefore, many emerging energy storage technologies are promising, such as lithium-ion batteries, vanadium liquid flow batteries, compressed air energy storage, flywheel energy storage, and sodium-ion batteries. Among them, lithium batteries are technically and economically mature and can be applied to energy storage from electric vehicles to grid-scale. First of all, BESS is mainly composed of a large number of small battery cells and a battery management system (BMS). BMS is essential for providing key information to the power conversion system (PCS) and energy management system (EMS), including state of charge (SOC), state of health (SOH), and remaining useful life (RUL). Existing research has the following problems: For example, data-driven BMS, although it improves the accuracy of state monitoring to a certain extent, Another example is the state monitoring method of BESS based on machine learning. Although it has high accuracy in theory, it still faces the challenges of data collection and processing in practical applications. At the same time, the existing BMS itself has certain limitations. First, the complex electrochemical process of lithium batteries makes it difficult to accurately describe their behavior in charging, discharging or idle states through traditional measurement and calculation methods. Second, although the combination of data-driven models and artificial intelligence (AI) provides new possibilities for BESS state monitoring, the amount of data from a single BESS is often insufficient to support the iterative refinement of the algorithm, and the local controller has limited ability to process large amounts of data and calculations. Finally, existing models often cannot accurately reflect the impact of battery aging, environmental changes and operating conditions. Therefore, it is very necessary to provide a state detection method for electric vehicle battery energy storage systems based on digital twins that builds a digital twin model, optimizes through machine learning algorithms, improves accuracy, and achieves precise monitoring. Summary of the invention

[0003] 1. Technical issues

[0004] In view of the above-mentioned existing technical status, this application mainly addresses the following technical problems:

[0005] 1. Existing technologies are limited by the amount of data and processing capabilities, making it difficult to achieve large-scale applications;

[0006] 2. Existing technologies fail to fully consider the impact of battery aging and environmental factors;

[0007] 3. Although existing technologies have high accuracy in theory, they still face challenges in data collection and processing in practical applications;

[0008] 4. The complexity and dynamic changes of battery performance make it difficult for traditional condition monitoring methods to meet the increasing accuracy requirements;

[0009] 5. The amount of data from a single BESS is often insufficient to support the iterative refinement of the algorithm, and the local controller has limited capabilities in processing large amounts of data and computing;

[0010] 6. Existing models often fail to accurately reflect the impact of battery aging, environmental changes, and operating conditions.

[0011] (II) Technical solution

[0012] The purpose of the present invention is to overcome the shortcomings of the prior art and to provide a state detection method for an electric vehicle battery energy storage system based on digital twins, which constructs a digital twin model, optimizes through a machine learning algorithm, improves accuracy, and achieves precise monitoring.

[0013] The object of the present invention is achieved by: a state detection method of an electric vehicle battery energy storage system based on digital twin, the method comprising the following steps:

[0014] Step 1: Build the BESS digital twin and physical model structure to achieve more efficient condition monitoring and parameter optimization;

[0015] Step 2: Build a digital model based on the physical structure and properties of the BESS, including digital models of battery cells and battery packs;

[0016] Step 3: Build a cloud platform based on machine learning algorithms to collect and store a large amount of BESS operation data, optimize the model parameters using machine learning algorithms, and feed the optimization results back to the digital model;

[0017] Step 4: Use the digital twin model to detect the SOC and SOH of the BESS system.

[0018] Furthermore, the BESS digital twin and physical model constructed in step 1 consists of two parts: a physical BESS and a cloud-based digital model.

[0019] Furthermore, the physical BESS is composed of a large number of battery cells and a battery management system.

[0020] Furthermore, the physical BESS also includes a local controller, which is responsible for performing data acquisition, processing, analysis, state estimation and protection; the local controller works in conjunction with the BMS, power conversion system PCS, energy management system and auxiliary equipment to monitor the state and environmental conditions of the physical BESS in real time.

[0021] Furthermore, the cloud-based digital model identifies and processes key parameters of battery performance through machine learning algorithms, receives data from the local controller in real time and performs parameter optimization. The optimized parameters and models are sent back to the local controller to adjust the operating status of the BESS, thereby realizing data-based battery system modeling and state estimation.

[0022] Furthermore, the digital model based on physical structure and properties constructed in step 2 is an equivalent circuit model of the battery module based on the Thevenin model. The digital twin model of the battery pack can be obtained by connecting the equivalent circuit models of each battery module in series.

[0023] Furthermore, the step 3 comprises the following steps:

[0024] Step 3.1: Selection and principles of machine learning algorithms;

[0025] Step 3.2: Building the cloud platform framework.

[0026] Furthermore, the selection and principle of the machine learning algorithm in step 3.1 are as follows: a long short-term memory network LSTM consisting of three main parts: an input gate, a forget gate, and an output gate is used. The input gate is responsible for filtering and memorizing the input data at the current moment, and the input data needs to be appropriately scaled and a memory control sequence needs to be set; the forget gate is responsible for filtering and forgetting the output data at the previous moment, and a forget control sequence needs to be set; the output gate is responsible for determining the output content at the current moment, and an output control sequence needs to be set; the relevant function expressions are as follows: Among them, z n is the input data after scaling; z i Select memory control sequence for information; f Select forget control sequence for information; o Select the output control sequence for the information; W n , W i , W f , W o is the weight matrix; b n 、b i 、b f 、bo is the bias matrix; x t is the input data at time t; h t-1 is the short-term memory output of the LSTM unit at time t-1; c t ′ is the long-term memory data information after the input data at time t is processed; c t 、c t-1 are the long-term memory outputs of the LSTM unit at time t and t-1 respectively; h t is the short-term memory output of the LSTM unit at time t;

[0027] The input memory gate, forget gate, and output gate can be coupled and reorganized to obtain a new control gate, and an improved LSTM unit structure can be obtained. That is, the optimized LSTM structure contains only one coupled control gate, and the function expression of the coupled control gate is as follows: Among them, z i ′ is the input data for information scaling processing.

[0028] Furthermore, the construction of the cloud platform framework in step 3.2 is specifically as follows: the cloud platform collects and stores a large amount of BESS operation data, and uses machine learning algorithms to optimize model parameters, and feeds the optimization results back to the digital model. The cloud platform can receive data and information from decentralized local BESS stations, including the moving average of battery charging / discharging power, DC side current or voltage ripple, and then aggregates and compares the data and information with the data from the simultaneously running digital twins, and judges them by their error covariance or other indicators to evaluate their differences: |X k -Y k |<ε, where X k is the operating data of the BESS station; Y k is the operating data of the digital twin model; ε is the error threshold; if the difference between the physical system and the digital twin exceeds the threshold, the parameters of the digital twin are optimized through a machine learning program, and the optimization objective function is to minimize the difference between the operating data of the BESS station and the operating data of the digital twin model, that is, minX k -Y k |.

[0029] Furthermore, the detection of the SOC and SOH of the BESS system in step 4 is specifically as follows: SOC is defined as the current state of charge and the nominal battery capacity C N Ratio of: Since SOC cannot be measured directly, the digital twin model of BESS can be used to identify the parameters of the equivalent circuit of the battery to estimate the SOC of the battery. oc The principle of linear relationship between U and SOC is to detect ocTo realize the indirect detection of SOC, according to the equivalent circuit, U oc Satisfies the following relationship: t =U oc -U1-U2-IR0;

[0030] Similarly, the SOH of a battery is defined as its aging level, which can be expressed as capacity decay and internal resistance increase: Among them C N and C current is the nominal capacity of the battery and the actual capacity at the current time t; R 0,0 and R 0,t are the internal resistance of the battery when it leaves the factory and at the current time t, respectively; therefore, the current battery capacity and internal resistance can be used to estimate the battery SOH.

[0031] (III) Beneficial effects

[0032] 1. First, the present invention designs and constructs a digital twin model of the battery energy storage system, which can accurately simulate the physical structure and chemical properties of the battery. By combining with the local controller of the physical system, the digital twin model not only realizes real-time data collection and processing, but also provides key functions such as state estimation and system protection;

[0033] 2. Furthermore, the present invention uses a cloud platform to collect and store a large amount of battery operation data, and optimizes the parameters of the digital model through a machine learning algorithm to improve the accuracy of state estimation. The optimized model can be fed back to the digital twin model in real time to achieve accurate monitoring of the battery energy storage system state, including detection of charge and health status;

[0034] 3. The effectiveness of the proposed method was verified through simulation experiments, proving the necessity and superiority of digital twin technology in the state detection of electric vehicle battery energy storage systems, and providing strong technical support for the safe and reliable operation of battery energy storage systems;

[0035] 4. The digital twin model of the present invention can collect BESS data distributed at different sites, aggregate all necessary data and information in the cloud platform, and identify the main parameters of different models through artificial intelligence algorithms;

[0036] 5. In addition, the model of the cloud platform can utilize the real-time data and information from the local BESS for timely optimization and calibration, thus having higher adaptability and accuracy in parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is the framework diagram of the BESS digital twin model of the present invention.

[0038] Figure 2This is an equivalent circuit model diagram based on the Thevenin model of the present invention.

[0039] Figure 3 Schematic diagram of the traditional LSTM unit structure of the present invention.

[0040] Figure 4 The figure is a schematic diagram of the principle structure of the improved LSTM unit of the present invention.

[0041] Figure 5 It is a detection flow chart of SOC and SOH of the BESS system of the present invention.

[0042] Figure 6 It is a schematic diagram of the SOC estimation situation of the present invention.

[0043] Figure 7 Schematic diagram of SOC estimation error of the present invention. DETAILED DESCRIPTION

[0044] In the field of electric vehicles, efficient management and condition monitoring of battery energy storage systems are key to ensuring vehicle performance and safety. However, the complexity and dynamic changes of battery performance make it difficult for traditional condition monitoring methods to meet the growing accuracy requirements. To address this challenge, this paper proposes an innovative electric vehicle battery energy storage system condition detection method based on digital twin technology; the method of the present invention utilizes the data storage and computing capabilities of the cloud platform to aggregate and analyze real-time data from different BESSs in the system for modeling and parameter identification, making the constructed model more flexible and accurate; in addition, based on the machine learning algorithm, a battery model of the series cluster and its SOC and SOH estimation method are proposed, which greatly improves the reliability and safety of the BESS.

[0045] The present invention is further described below in conjunction with embodiments and / or drawings.

[0046] Example 1

[0047] like Figure 1-7 As shown, a state detection method of an electric vehicle battery energy storage system based on digital twins comprises the following steps:

[0048] Step 1: Build the BESS digital twin and physical model structure to achieve more efficient condition monitoring and parameter optimization;

[0049] In the present invention, in modern power systems, battery energy storage systems play a vital role, especially in integrating renewable energy and improving grid stability. In order to achieve accurate monitoring and optimization of BESS performance, digital twin technology provides an innovative solution. This step of the present invention proposes a system design that combines a physical BESS and a digital twin model to achieve more efficient state monitoring and parameter optimization.

[0050] like Figure 1 As shown in the figure, the BESS digital twin model framework consists of two parts: the physical BESS and the cloud-based digital model; the physical BESS consists of a large number of battery cells and a battery management system, while the cloud-based digital model is responsible for data-based battery system modeling and state estimation.

[0051] Among them, ① the core of the physical BESS system is the local controller, which is responsible for performing key tasks such as data acquisition, processing, analysis, state estimation and protection; the local controller works together with the BMS, power conversion system (PCS), energy management system and auxiliary equipment to monitor the state and environmental conditions of the physical BESS in real time.

[0052] The local controller collects data from the BESS through high-precision sensors, including key parameters such as voltage, current, and temperature. These data are preprocessed to eliminate noise and outliers to ensure the accuracy and reliability of the data. In addition, in order to reduce the data transmission and calculation burden, the local controller will also filter out the data with the largest amount of information and ignore those ordinary data that have little impact on state estimation. Based on this, using the collected data, the local controller executes state estimation algorithms, such as SOC and SOH monitoring, to evaluate the current state and health level of the battery (the common methods used for local controllers are: Kalman Filter method (Ka lman Filter): including Extended Kalman Filter (Extended Ka lman Filter, EKF), Unscented Ka lman Filter (Unscented Ka lman Filter, UKF), Square Root Unscented Ka lman Filter (SquareRoot Unscented Ka lman Filter, SR-UKF) and Adaptive Extended Kalman Filter (AdaptiveExtended Ka lman Filter) lter, AEKF), etc. These methods can be used in the local controller and estimate the SOC and SOH of the battery); this information is crucial for predicting the remaining service life of the battery and formulating maintenance strategies; at the same time, the local controller is also responsible for executing protection strategies to prevent overcharging, over-discharging and other operations that may cause battery damage.

[0053] ② The cloud-based digital model is another key component of the BESS digital twin. It uses the data sent from the local controller for more in-depth analysis and modeling. By applying machine learning algorithms (generally representing the neural network method (Neura l Network): by simulating the human brain and its neurons to process nonlinear systems, a large number of samples are extracted from the target battery for training to achieve real-time acquisition of the battery SOC value; in the present invention, the long short-term memory network (Long Short Term Memory, LSTM) is used. It is an improved version of the traditional recurrent neural network (Recurrent Neura l Networks, RNN), which is discussed in detail below), the cloud-based model can identify the key parameters of battery performance and optimize the parameters to improve the accuracy of state estimation; the cloud-based digital model can not only process historical operating data, but also receive data from the local controller in real time. This real-time data stream enables the model to be updated and optimized in time to adapt to changes in battery performance; the optimized parameters and models are sent back to the local controller to adjust the operating status of the BESS to achieve more accurate state monitoring and performance optimization.

[0054] Based on this digital twin structure design that combines the physical BESS and the cloud digital model, real-time monitoring and optimization of BESS performance can be achieved. This design not only improves the accuracy of state estimation, but also enhances the reliability and safety of the system, providing strong support for the stable operation of the power system.

[0055] Step 2: Build a digital model based on the physical structure and properties of the BESS, including digital models of battery cells and battery packs;

[0056] In the present invention, further, a digital model of BESS can be constructed according to the physical structure and physical properties of BESS, including digital models of battery cells and battery packs. The digital model is an equivalent circuit model of the battery module based on the Thevenin model, such as Figure 2 As shown in FIG. 1 ; the more the number of resistor-capacitor (RC) pairs in the model, the higher the accuracy of the replicated battery; in the present invention, two RC pairs are used to model a single battery, namely C n1 / R n1 and C n2 / R n2 ; U noc Represents open circuit voltage; R no Indicates ohmic resistance; U nt Represents the terminal voltage; the batteries in BESS are usually connected in series to form a module with a certain terminal voltage. Several modules are connected in series to form a battery pack, and several battery packs form a battery pack. The digital twin model of the battery pack can be obtained by connecting the equivalent circuit models of each battery module in series.

[0057] Step 3: Build a cloud platform based on machine learning algorithms to collect and store a large amount of BESS operation data, optimize the model parameters using machine learning algorithms, and feed the optimization results back to the digital model;

[0058] Step 4: Use the digital twin model to detect the SOC and SOH of the BESS system.

[0059] The present invention is a state detection method for an electric vehicle battery energy storage system based on digital twins. During use, the present invention establishes a digital twin model of an electric vehicle BESS, processes the operating data collected on a cloud platform through a machine learning algorithm, identifies the model parameters of the battery module equivalent circuit model, and then realizes the state monitoring of the BESS; compared with the prior art, the present invention has the following advantages and beneficial effects: ① The model parameters can be calibrated in real time using the operating data, eliminating the influence of parameter degradation and environmental factors on the model parameters during operation; ② The digital twin model data is collected on the cloud platform, which solves the problem that the local controller cannot process a large amount of data, and also improves the accuracy of state monitoring; the present invention has the advantages of constructing a digital twin model, optimizing through a machine learning algorithm, improving accuracy, and realizing precise monitoring.

[0060] Example 2

[0061] like Figure 1-7 As shown, a state detection method of an electric vehicle battery energy storage system based on digital twins comprises the following steps:

[0062] Step 1: Build the BESS digital twin and physical model structure to achieve more efficient condition monitoring and parameter optimization;

[0063] Step 2: Build a digital model based on the physical structure and properties of the BESS, including digital models of battery cells and battery packs;

[0064] Step 3: Build a cloud platform based on machine learning algorithms to collect and store a large amount of BESS operation data, optimize the model parameters using machine learning algorithms, and feed the optimization results back to the digital model;

[0065] The present invention comprises the following steps:

[0066] Step 3.1: Selection and principles of machine learning algorithms;

[0067] Specifically, the long short-term memory network (LSTM) selected by the present invention is an improved version of the traditional recurrent neural network (RNN), which effectively alleviates the gradient vanishing problem and the long-term dependency capture problem that occurs as the time series grows. Figure 3 The LSTM unit structure diagram shown in the figure; the LSTM unit consists of three main parts: input gate, forget gate and output gate. The input gate is responsible for filtering and memorizing the input data at the current moment. It is necessary to appropriately scale the input data and set the memory control sequence; the forget gate is responsible for filtering and forgetting the output data at the previous moment. It is necessary to set the forget control sequence; the output gate is responsible for determining the output content at the current moment. It is necessary to set the output control sequence; its related function expressions are as follows: Among them, z n is the input data after scaling; z i Select memory control sequence for information; f Select forget control sequence for information; o Select the output control sequence for the information; W n , W i , W f , W o is the weight matrix; b n 、b i 、b f 、b o is the bias matrix; x t is the input data at time t; h t-1 is the short-term memory output of the LSTM unit at time t-1; c t ′ is the long-term memory data information after the input data at time t is processed; c t 、c t-1 are the long-term memory outputs of the LSTM unit at time t and t-1 respectively; h t is the short-term memory output of the LSTM unit at time t.

[0068] Based on this, in order to improve the LSTM unit structure and increase the model training speed, the three parts of the input memory gate, forget gate, and output gate can be coupled and reorganized to obtain a new control gate. The improved LSTM unit structure is shown in the figure below: Figure 4 shown.

[0069] In summary, compared with the traditional LSTM unit structure, the optimized LSTM structure contains only one coupled control gate, which can reduce the number of parameter updates to a greater extent and simplify the variable calculation process, thereby speeding up the training efficiency of the model; the function expression of its coupled control gate is as follows: Among them, z i ′ is the input data for information scaling processing.

[0070] Step 3.2: Building the cloud platform framework.

[0071] Specifically: A cloud platform can be built to collect and store a large amount of BESS operation data, and use machine learning algorithms to optimize model parameters and feed the optimization results back to the digital model. The cloud platform can receive data and information from decentralized local BESS stations, including the moving average of battery charging / discharging power, DC side current or voltage ripple, and then aggregate and compare the data and information with the data from the digital twins running at the same time, and judge them by their error covariance or other indicators to evaluate their differences: |X k -Y k |<ε, where X k is the operating data of the BESS station; Y k is the operating data of the digital twin model; ε is the error threshold.

[0072] If the difference between the physical system and the digital twin exceeds a threshold, the parameters of the digital twin are optimized through a machine learning program. The optimization objective function is to minimize the difference between the operating data of the BESS station and the operating data of the digital twin model, that is, minX k -Y k |.

[0073] Step 4: Use the digital twin model to detect the SOC and SOH of the BESS system.

[0074] In the present invention, based on this, the digital twin model can be used to realize the SOC and SOH detection of the BESS system. The specific process is as follows: Figure 5 shown.

[0075] Among them, SOC is defined as the current state of charge and the nominal battery capacity C N Ratio of: In the formula, SOC t is the state of charge at time t, indicating the current state of charge of the battery; SOC0 is the state of charge at the initial time t0, that is, the initial state of charge of the battery; η is the Coulomb efficiency of the battery, indicating the ratio of the actual amount of charge stored or released by the battery during the charging and discharging process to the theoretical amount of charge; I is the current of the battery, indicating the current intensity during the charging and discharging process of the battery; C N It is the nominal capacity of a battery, usually expressed in ampere-hours or ampere-hours, and is the maximum amount of charge that a battery can store or release under standard conditions.

[0076] In actual scenarios, SOC cannot be measured directly, so the digital twin model of BESS can be used to identify the parameters of the battery equivalent circuit to estimate the battery SOC. oc The principle of linear relationship between U and SOC is to detect oc To realize the indirect detection of SOC, according to the equivalent circuit, Uoc Satisfies the following relationship: t =U oc -U1-U2-IR0, where U t is the terminal voltage of the battery at time t, that is, the actual voltage of the battery in the working state; U oc Open circuit voltage refers to the voltage of a battery when no current flows, that is, the electromotive force of the battery. It has a linear relationship with the state of charge (SOC) of the battery, and the SOC is usually detected indirectly by detecting the open circuit voltage. U1 and U2 represent the voltage drop in the battery equivalent circuit model, which is related to the polarization phenomenon of the battery. I is the current flowing through the battery, which indicates the current intensity during the battery charging and discharging process. R0 is the ohmic internal resistance of the battery, which represents a part of the internal resistance of the battery, is related to the charge and discharge state of the battery, and may change with the use and aging of the battery.

[0077] When the battery is in idle state, U t Can be considered as U oc However, due to the weak chemical reactions and ion interactions at the microscopic level, this approximate relationship is satisfied only after a long time, which is not conducive to the real-time status detection of BESS. However, through the machine learning algorithm in the digital twin, the U can be identified through the operation data collected in the cloud platform. oc This provides great convenience for accurate battery open circuit voltage, which can be obtained by looking up the table or SOC-U oc The SOC value is obtained from the curve.

[0078] Similarly, the SOH of a battery is defined as its aging level, which can be expressed as capacity decay and internal resistance increase: Among them C N and C current is the nominal capacity of the battery and the actual capacity at the current time t; R 0,0 and R 0,t are the internal resistance of the battery when it leaves the factory and at the current time t, respectively; therefore, the current battery capacity and internal resistance can be used to estimate the battery SOH.

[0079] Similarly, the current battery capacity and its ohmic resistance can be identified through a machine learning program, and the SOH in terms of capacity fade and resistance increase can be calculated according to the above formula.

[0080] In order to further illustrate the method of the present invention, an example analysis is performed, as follows:

[0081] ① Example setting

[0082] In the field of electric vehicles and renewable energy storage, the performance of battery energy storage systems (BESS) is crucial. In order to improve the reliability and safety of BESS, the status of the battery needs to be accurately monitored. The experiment of the present invention aims to verify the effectiveness of the proposed BESS status detection method based on digital twins. By performing charge and discharge tests on the selected batteries under different temperature conditions, collecting data and using the digital twin model for state estimation, the accuracy and reliability of the method in practical applications are evaluated.

[0083] Meanwhile, the hardware condition of the simulation platform is 11th Gen Intel(R)Core(TM)i7-1165G7@2.80GHz, and the software condition is matlab2023b.

[0084] ② Example results and analysis

[0085] In this study, a detailed simulation analysis of the state detection method of electric vehicle battery energy storage system based on digital twin was carried out. Figure 6-7 The SOC and SOH estimation results and their errors after 150 cycle tests at 40°C are demonstrated; simulation data show that the maximum error of the proposed method for SOC is 2.587%, and the average error is only 0.5519%, indicating that the proposed method can maintain a high estimation accuracy under high temperature conditions; similarly, the SOH estimation results can be obtained: the maximum estimation error of the proposed method is 4.29%, and the average error is only 1.78%, which also proves the superiority of the proposed method in SOH estimation.

[0086] ③In summary, the estimation error of the proposed method is significantly lower than that of the control group, which verifies the high accuracy and stability of the proposed method in SOC estimation under high temperature conditions; the lower error rate means that the real-time status of the battery can be monitored more accurately, thereby providing reliable data support for the battery management system; at the same time, under the same test conditions, the proposed method also shows high accuracy in SOH estimation, which shows that digital twin technology can effectively capture changes in battery health status and provide a more accurate basis for battery maintenance and replacement; therefore, the proposed method shows extremely high accuracy in both SOC and SOH estimation, which is mainly due to the fact that digital twin technology can use real-time data to dynamically update and optimize the model to adapt to the dynamic changes in battery performance.

[0087] The present invention is a state detection method for an electric vehicle battery energy storage system based on digital twins. During use, the present invention establishes a digital twin model of an electric vehicle BESS, processes the operating data collected on a cloud platform through a machine learning algorithm, identifies the model parameters of the battery module equivalent circuit model, and then realizes the state monitoring of the BESS; compared with the prior art, the present invention has the following advantages and beneficial effects: ① The model parameters can be calibrated in real time using the operating data, eliminating the influence of parameter degradation and environmental factors on the model parameters during operation; ② The digital twin model data is collected on the cloud platform, which solves the problem that the local controller cannot process a large amount of data, and also improves the accuracy of state monitoring; the present invention has the advantages of constructing a digital twin model, optimizing through a machine learning algorithm, improving accuracy, and realizing precise monitoring.

Claims

1. A state detection method for an electric vehicle battery energy storage system based on digital twins, characterized in that: The method comprises the following steps: Step 1: Build the BESS digital twin and physical model structure to achieve more efficient condition monitoring and parameter optimization; Step 2: Build a digital model based on the physical structure and properties of the BESS, including digital models of battery cells and battery packs; Step 3: Build a cloud platform based on machine learning algorithms to collect and store a large amount of BESS operation data, optimize the model parameters using machine learning algorithms, and feed the optimization results back to the digital model; Step 4: Use the digital twin model to detect the SOC and SOH of the BESS system.

2. The state detection method of the electric vehicle battery energy storage system based on digital twin according to claim 1, characterized in that: The BESS digital twin and physical model constructed in step 1 consists of two parts: the physical BESS and the cloud digital model.

3. The state detection method of the electric vehicle battery energy storage system based on digital twin according to claim 2, characterized in that: The physical BESS consists of a plurality of battery cells and a battery management system.

4. The state detection method of the electric vehicle battery energy storage system based on digital twin according to claim 3 is characterized in that: The physical BESS also includes a local controller, which is responsible for performing data acquisition, processing, analysis, state estimation and protection; the local controller works in conjunction with the BMS, the power conversion system PCS, the energy management system and auxiliary equipment to monitor the state and environmental conditions of the physical BESS in real time.

5. The state detection method of the electric vehicle battery energy storage system based on digital twin according to claim 4, characterized in that: The cloud-based digital model identifies and processes key parameters of battery performance through machine learning algorithms, receives data from the local controller in real time and performs parameter optimization. The optimized parameters and models are sent back to the local controller to adjust the operating status of the BESS, thereby realizing data-based battery system modeling and status estimation.

6. The state detection method of the electric vehicle battery energy storage system based on digital twin according to claim 1, characterized in that: The digital model based on physical structure and properties constructed in step 2 is an equivalent circuit model of the battery module based on the Thevenin model. The digital twin model of the battery pack can be obtained by connecting the equivalent circuit models of each battery module in series.

7. The state detection method of the electric vehicle battery energy storage system based on digital twin according to claim 1, characterized in that: The step 3 comprises the following steps: Step 3.1: Selection and principles of machine learning algorithms; Step 3.2: Building the cloud platform framework.

8. The state detection method of the electric vehicle battery energy storage system based on digital twin according to claim 7, characterized in that: The selection and principle of the machine learning algorithm in step 3.1 are as follows: a long short-term memory network LSTM consisting of three main parts: an input gate, a forget gate, and an output gate is used. The input gate is responsible for filtering and memorizing the input data at the current moment, and the input data needs to be appropriately scaled and a memory control sequence needs to be set; the forget gate is responsible for filtering and forgetting the output data at the previous moment, and a forget control sequence needs to be set; the output gate is responsible for determining the output content at the current moment, and an output control sequence needs to be set; The relevant function expressions are as follows: Among them, z n is the input data after scaling; z i Select memory control sequence for information; f Select forget control sequence for information; o Select the output control sequence for the information; W n , W i , W f , W o is the weight matrix; b n 、b i 、b f 、b o is the bias matrix; x t is the input data at time t; h t-1 is the short-term memory output of the LSTM unit at time t-1; c t ′ is the long-term memory data information after the input data at time t is processed; c t 、c t-1 are the long-term memory outputs of the LSTM unit at time t and t-1 respectively; h t is the short-term memory output of the LSTM unit at time t; The input memory gate, forget gate, and output gate can be coupled and reorganized to obtain a new control gate, thereby obtaining an improved LSTM unit structure. That is, the optimized LSTM structure contains only one coupled control gate, and the function expression of the coupled control gate is as follows: Among them, z i ′ is the input data for information scaling processing.

9. The state detection method of the electric vehicle battery energy storage system based on digital twin according to claim 7, characterized in that: The construction of the cloud platform framework in step 3.2 is specifically as follows: the cloud platform collects and stores a large amount of BESS operation data, and uses machine learning algorithms to optimize model parameters, and feeds the optimization results back to the digital model. The cloud platform can receive data and information from decentralized local BESS stations, including the moving average of battery charging / discharging power, DC side current or voltage ripple, and then aggregates and compares the data and information with the data from the digital twins running at the same time, and judges them by their error covariance or other indicators to evaluate their differences: |X k -Y k |<ε, where X k is the operating data of the BESS station; Y k is the operating data of the digital twin model; ε is the error threshold; if the difference between the physical system and the digital twin exceeds the threshold, the parameters of the digital twin are optimized through a machine learning program, and the optimization objective function is to minimize the difference between the operating data of the BESS station and the operating data of the digital twin model, that is, minX k -Y k |.

10. The state detection method of the electric vehicle battery energy storage system based on digital twin according to claim 1, characterized in that: The detection of the SOC and SOH of the BESS system in step 4 is specifically as follows: SOC is defined as the current state of charge and the nominal battery capacity C N Ratio of: In the formula, SOC t is the state of charge at time t, indicating the current state of charge of the battery; SOC0 is the state of charge at the initial time t0, that is, the initial state of charge of the battery; η is the Coulomb efficiency of the battery, indicating the ratio of the actual amount of charge stored or released by the battery during the charging and discharging process to the theoretical amount of charge; I is the current of the battery, indicating the current intensity during the charging and discharging process of the battery; C N is the nominal capacity of the battery; since SOC cannot be measured directly, the digital twin model of BESS can be used to identify the parameters of the equivalent circuit of the battery to estimate the SOC of the battery, that is, using U oc The principle of linear relationship between U and SOC is to detect oc To realize the indirect detection of SOC, according to the equivalent circuit, U oc Satisfies the following relationship: t =U oc -U1-U2-IR0, where U t is the terminal voltage of the battery at time t; U oc Open circuit voltage; U1 and U2 represent the voltage drops in the battery equivalent circuit model; I is the current flowing through the battery; R0 is the ohmic internal resistance of the battery; Similarly, the SOH of a battery is defined as its aging level, which can be expressed as capacity decay and internal resistance increase: Among them C N and C current is the nominal capacity of the battery and the actual capacity at the current time t; R 0,0 and R 0,t They are the internal resistance of the battery when it leaves the factory and at the current time t; Therefore, the current battery capacity and internal resistance can be used to estimate the battery SOH.

Images