Virtual synchronous generator (VSG) has become an effective means to solve the problem of grid-connected frequency fluctuation of new energy by simulating the inertia characteristics of traditional generators. However, in the traditional VSG method, the moment of inertia and damping coefficient are usually fixed, which may lead to frequency deviation when the load changes. Therefore, a method combining adaptive parameter adjustment and model predictive control (MPC) is proposed to optimize the frequency performance of VSG. By analyzing the small-signal model of VSG, an adaptive adjustment strategy is proposed to suppress frequency fluctuation. At the same time, MPC is used to dynamically adjust the input power of VSG to enhance the response of power generation units to the change of power demand. Simulation results show that the proposed method effectively improves the frequency stability of the system.

Accurate estimation of the state of health (SOH) of lithium-ion batteries can effectively ensure the safe operation of battery systems. To address the difficulties in selecting health indicators (HIs) and the problem that computational methods are prone to falling into local optimal solutions in the process of estimating the SOH of lithium-ion batteries, a lithium-ion battery SOH online estimation model based on incremental capcity (IC) of quantum genetic algorithm optimized back-propagation (QGA-BP) neural network is proposed. This algorithm first extracts key indicators from the IC curves during battery aging, subsequently conducts a correlation analysis between the extracted HIs and the SOH, then uses the extracted indicators as input to train the QGA-BP neural network to achieve online estimation of the SOH of lithium-ion batteries, and finally verified by the experimental data measured from the experimental platform and the NASA dataset. The results show that the maximum values of the mean absolute error (MAE), mean absolute percentage error (MAPE), and root mean square error (RMSE) of the proposed algorithm are 0.013 12, 0.012 94 and 0.017 86, respectively, compared with the traditional back-propagation (BP) neural network and other algorithm, with good convergence and robustness, and can achieve the accurate online estimation of SOH of lithium-ion batteries.

The precise identification of model parameters for lithium-ion batteries is crucial for the observing of remaining battery capacity and extending the service life. However, traditional parameter identification methods are often affected by parameter coupling, getting trapped in a local optimum, and other factors, which reduce the estimation accuracy and subsequently lower the accuracy of the subsequent joint estimation. Therefore, an SOC-SOH joint estimation based on a three-stage parameter identification method is proposed. In the first stage, an RC model with fast and slow dynamic modules is constructed based on the internal dynamic characteristics of the battery. In the second stage, the fast dynamic (FD) module uses the forgetting factor recursiveleast squares (FFRLS) adaptive parameter identification method to identify the parameters R₀, R₁, and C₁ with a forgetting factor. In the third stage, the slow dynamic (SD) module uses the adaptive extended Kalman filter algorithm (AEKF) for online parameter identification, and combines the zebra optimization algorithm (ZOA) to solve the local optimum problem and further improve the accuracy of the parameter identification of U_OC, R₂, and C₂. Considering that battery capacity and battery state have different time-varying characteristics, a multi-time scale method is used for the SOC-SOH joint estimation to improve the estimation accuracy. The simulation results show that the improved joint estimation method has better estimation characteristics than traditional methods in terms of average error and maximum error of SOC estimation under different temperatures and operating conditions.

Accurate estimation of state of charge (SOC) and state of health (SOH) of lithium-ion batteries is of great significance in improving the performance of battery management system (BMS). In order to solve the problem that the unscented Kalman filter (UKF) algorithm has low accuracy in estimating the SOC of lithium-ion batteries, a double-multi-new information adaptive UKF (DMIAUKF) algorithm considering battery aging is proposed based on the UKF algorithm. The algorithm uses two multi-new information adaptive unscented Kalman filters (MIAUKF) to simultaneously estimate the SOC and SOH of the battery. Taking the ternary lithium-ion battery as the experimental research object, under the hybrid pulse power characteristic (HPPC) condition, the offline parameter identification method is used to identify the parameters of the second-order RC equivalent circuit model. The experimental results show that the DMIAUKF algorithm achieves high accuracy in estimating the SOC and SOH of lithium-ion batteries under constant-current discharge and HPPC conditions.

Aiming at the shortcomings of low estimation accuracy and poor robustness of extended Kalman filter (EKF) state of charge (SOC) under non-Gaussian noise, an adaptive extended Kalman filter algorithm based on dynamic generalized maximum correntropy (DGMCC-AEKF) optimization is proposed. The algorithm improves the robustness by replacing the minimum mean square error criterion with the generalized maximum correntropy criterion, and proposes a strategy of dynamically adjusting the kernel width parameters based on the exponential form combined with the iterative method, which improves the accuracy and stability. Under dynamic stress test (DST) and federal urban driving schedule (FUDS) conditions, the mean absolute error and root mean squared error of the new algorithm are lower than those of the existing adapative extended Kalman filter (AEKF), adapative iterated extended Kalman filler (AIEKF) and maximum correntropy criterion extended Kalman filter (MCC-EKF) algorithms. Compared with other adaptive kernel width methods, it has better performance under four noise schemes.

Aiming at the problems of difficulty in extracting feature parameters of the life assessment model of retired power batteries and low accuracy and low assessment efficiency in model improvement, the life assessment model combining convolutional neural network (CNN) and long short-term memory (LSTM) network is improved by adding input feature parameters that are easy to extract, without increasing the computational burden. The equal voltage drop discharge time, equal temperature rise discharge time, and charging time during cycling of retired batteries are selected as the input feature parameters of the CNN-LSTM life assessment model, and the proposed model has improved accuracy and stability and is more efficient compared with the existing CNN-LSTM life assessment model. In addition, in order to improve the efficiency of the model in evaluating the life of retired batteries, the effect of reducing the proportion of training resources on the assessment results is explored, and it is found that when only 50% of the total number of battery cycles is selected as the training set of the model, the accuracy and stability of the model assessment can still be maintained at more than 99.00%.

To address the issue of single-cell voltage inconsistency in energy storage battery packs caused by manufacturing processes and other factors, a balancing control scheme capable of interleaved operation in multiple working modes is proposed. This scheme consists of a T-type transformer-based active balancing topology and a corresponding multi-mode interleaved balancing control strategy. The balancing scheme uses the terminal voltage of the battery as the balancing variable and can operate in an interleaved manner under three working modes: partial cell-to-cell, cell-to-pack, and cell-to-cell. This improves the balancing speed. Simulation results from MATLAB/Simulink platform show that, under three operating conditions—idle, charging, and discharging—the balancing time is significantly reduced compared to traditional transformer-based balancing circuits. Specifically, compared to the partial cell-to-cell balancing circuit, the balancing time is reduced by 45.2%, 38.4% and 41.1% respectively. Compared to the transformer-based cell-to-pack balancing circuit, it is reduced by 37%, 30.7% and 36.1% respectively. Compared to the transformer-based cell-to-cell balancing circuit, it is reduced by 21.2%, 25.8% and 32.6% respectively.

With the widespread application of lithium-ion batteries in industry and daily life, aiming at the existing problems of low energy utilization efficiency and long equalization time during battery pack equalization, the existing modular and reconfigurable circuits are improved, and a modular reconfigurable battery equalization circuit with multi-mode static charge-discharge operation is proposed. First, modularization processing and mathematical modeling of the equalization circuit are carried out. A bidirectional Cuk converter is used within the module; and a reconfigurable method is used for charging equalization between modules; discharge and static equalization dynamically modularize the battery pack with a bidirectional Buck-Boost converter. The battery equalization circuit is mathematically modeled, the working principle of the circuit is introduced, and the corresponding control strategies are developed for different equalization situations. Then, the simulation model of the proposed equalization circuit is constructed and compared with the traditional equalization circuit, and the equalization time, equalization efficiency and energy loss ratio of the circuit are analyzed. Finally, in combination with the hardware-in-the-loop simulation platform Typhoon, the battery equalization circuit is built, and the effectiveness of the proposed equalization circuit is verified.

The inconsistency problem of lithium-ion batteries is one of the key factors affecting the overall performance output of battery packs, and accurate estimation of inconsistency parameters is crucial for the health management and optimization control of battery packs. The traditional complex single battery cell parameter estimation algorithm faces the problem of huge computational complexity when extended to all cells within the battery pack. A method for estimating inconsistent parameters based on an offline single battery cell simulation database is proposed. First, by building an equivalent circuit model to simulate and generate a comprehensive battery sample database. Then, segment voltage characteristic curves are extracted from the target battery charging data and compared with the sample library. Based on similarity measurement, the best matching sample is locked, thereby achieving rapid estimation of the capacity and internal resistance of all battery cells in the group. Subsequently, considering the interference of interpolation errors in the method and voltage sampling errors in practical applications, a correction mechanism based on sliding window search is proposed to improve estimation accuracy. Finally, method validation was conducted on 60 single battery-cell series connected battery packs set up in a laboratory environment. The results showed that the mean absolute percentage error (MAPE) of the capacity did not exceed 0.40%, and MAPE of the internal resistance did not exceed 2.60%, and maintained good stability on simulated sampling error noise data.

To solve the problem of high temperature rise and poor temperature uniformity of energy storage battery module under high-rate operation, considering the heat generation of batteries and aluminum bars, an electric-thermal-fluid coupling simulation model for battery module was established, and the influence of cold plate arrangement scheme and flow rate on the temperature characteristics of the battery module was analyzed. The results indicate that the scheme of arranging double large surface or side cold plates can control the maximum temperature rise within 11.0 K, which is better than the scheme of arranging single large surface cold plates. The scheme of arranging side cold plates is superior to the scheme of arranging large surface cold plates in controlling the temperature rise range. Increasing the flow rate can effectively reduce the maximum temperature rise and temperature rise range of the module when the flow rate is within 3.0 L/min. Adding bottom cold plates in the scheme of arranging large or side cold plates can improve the temperature characteristics of the module when the flow rate is within 2.0 L/min. The comparison between experimental and simulation results shows that the maximum temperature rise error is within 0.9 K, which verifies the accuracy of the coupled simulation model.