To solve the problem of low model prediction accuracy caused by noise interference and capacity augmentation in the remaining useful life (RUL) prediction of lithium batteries, an RUL prediction method combining the parameter-optimized multivariate variational modal decomposition (MVMD) and bi-directional long and short-term memory (BiLSTM) network is proposed. Multiple health factors (HFs) are extracted from the charging and discharging processes of lithium batteries as model inputs, and the northern goshawk optimization (NGO) algorithm is introduced by combining the sine cosine algorithm and the whale spiral predation mechanism, which is used to optimize the MVMD parameters. Then, the optimized MVMD is used to decompose the HF into a number of smooth components. Finally, the optimized multilayer BiLSTM is used for prediction, and the stacked summation is the final RUL result. The results of a comparative analysis of the proposed method and other methods based on NASA and CALCE data sets show that this method has higher prediction accuracy and generalized adaptability.

To ensure the safe operation of energy storage power stations and new energy vehicles, aimed at the thermal runaway phenomenon caused by internal short-circuit faults in lithium-ion batteries, a fault diagnosis method based on sparrow search algorithm optimized variational mode decomposition (SSA-VMD) and random forest (RF) algorithm is proposed. First, SSA is used to find the optimal parameter combination of VMD decomposition layer K and penalty factor α, and the internal short-circuit fault signal and normal signal of lithium-ion batteries are decomposed into multiple intrinsic mode functions (IMF). Second, the sample entropy value of each IMF is calculated as an eigenvector. Finally, the eigenvectors are input into the support vector machine (SVM) fault diagnosis model and the RF fault diagnosis model for fault diagnosis, respectively. Results show that the fault diagnosis rate of the SVM model is 66.667 0% and that of the RF model is 96.666 7%, indicating that the internal short-circuit fault in lithium-ion batteries is effectively identified.

Aimed at the difficulties and lack of precision in the online estimation of state-of-health (SOH) for cascaded utilized lithium batteries, an estimation method based on pulse voltage and hybrid neural network is proposed. Based on the battery pulse voltage response curve, this method extracts the slope of terminal voltage and the corrected standard deviation which are related to battery SOH degradation as health features. By integrating convolutional neural networks (CNN) and gated recurrent unit (GRU), the online estimation of SOH for cascaded utilized lithium batteries with high accuracy and high stability is realized. A platform for a battery testing system was established to validate the proposed SOH estimation strategy for cascaded utilized lithium batteries. Results indicate that the proposed method excels in real-time accurate estimation of lithium battery SOH, and both the average absolute error and root mean square error are controlled within 0.01, demonstrating excellent generalization capability and robustness.

To accurately predict the state-of-charge (SOC) of electric loader batteries, a prediction model combining convolutional neural networks (CNN), bi-directional long short-term memory (BiLSTM) network and attention mechanism was proposed. First, the Pearson correlation coefficient analysis method was used to filter the input features; second, the spatial features were extracted by CNN; then the extracted feature sequences were input into the bi-directional LSTM layer for time series modeling, in addition, the attention mechanism was fused to allocate weights, so as to enhance the important features; finally, the prediction results were output by the fully-connected layer. Experimental results show that the prediction accuracy of the CNN-BiLSTM-Attention model reached 99.97%, and the prediction error was within 1.00%. Compared with models such as the CNN-LSTM-Attention model, this model had higher prediction accuracy and smaller error, indicating certain superiority and reliability.

The state of health (SOH) of lead-acid batteries is an important indicator for their reliable operation. A prediction model based on convolutional neural networks (CNN) optimized by the whale optimization algorithm (WOA) was proposed to solve the problem of low prediction accuracy of SOH of forklift batteries. First, CNN was taken as the basic model for the WOA-CNN model. Then, to further optimize the performance of the CNN model, the WOA algorithm was introduced to optimize the weight and bias of CNN by searching the optimal solution in space. To verify the validity of the WOA-CNN method, experiments were conducted using a real data set of forklift batteries. Experimental results show that the WOA-CNN model reduced the mean square error, root mean square error, mean absolute error and mean absolute percentage error in predicting the SOH of forklift batteries by 29.4%, 19.7%, 41.5% and 19.5% respectively compared with the unoptimized CNN neural network, and it reduced these errors by 42.2%, 31.1%, 64.7% and 54.8% respectively compared with the traditional BP neural network, exhibiting smaller prediction errors, higher accuracy and higher stability. The WOA-CNN prediction model provides strong support for the battery health management of electric forklifts, thus extending the battery life and improving the forklift reliability.

The accurate estimation of the state-of-health (SOH) of lithium-ion batteries is a basis for ensuring the safe, stable and efficient operation of battery systems. On this basis, a method for estimating the SOH of lithium-ion batteries was proposed based on health indicator extraction and improved Gaussian process regression (GPR). First, the health indicators were extracted from the charge/discharge current and voltage curves, and those with a high correlation with capacity were screened out by Pearson correlation analysis, which were further reduced by an autoencoder deep learning algorithm. Second, GPR was used to establish an SOH estimation model, and the kernel function of GPR was constructed by combinatorial kernel functions. Third, the African vultures optimization algorithm was used to optimize the hyperparameters in the combinatorial kernel function. Finally, the NASA lithium-ion battery dataset was used for model verification, and experimental results show that the proposed method has high accuracy.

It is becoming increasingly important to carry out reliability evaluation on core devices in the electric vehicle charging equipment, so as to ensure the safe and stable operation of the charging equipment. An advanced prediction method is proposed to estimate the state of health of electrolytic capacitor in the power supply which is one of the most important components in the charging equipment, and its remaining useful life (RUL) is also predicted. According to the capacitance loss data, the Verhulst model and exponential model are combined as the final empirical model, and unscented Kalman filter (UKF) is used to generate a recommended distribution of particle filter (PF) to track the degradation path. Aimed at the problem of particle impoverishment, an improved krill herd (KH) algorithm is used to optimize the residual resampling step and improve the prediction accuracy. The results of experiments on the electrolytic capacitor degradation data show that compared with those of the traditional PF algorithm, the root mean square error (RMSE) of prediction result decreases from 2.79 to 0.37 at the prediction time point of 80 h, and the confidence interval width at the 95% level decreases from 34 h to 4 h. At the same time, compared with that of the KH-UKF-PF algorithm, the RMSE decreases by 27.5%, indicating that the proposed method can provide higher prediction accuracy of RUL.

With the continuous expansion of the scale and capacity of offshore wind power, the adoption of high-voltage direct current transmission technology has become an inevitable choice for achieving efficient long-distance transmission, in which the use of medium-frequency AC collection with diode rectifier unit (DRU) to achieve high-voltage direct current transmission is considered one of the optimal schemes. However, the nonlinearity of DRU will cause wind power grid-connected converters to generate heavy harmonic currents, which will have adverse effects on the offshore power grid and threaten its safe and stable operation. To solve this problem, a harmonic suppression strategy is proposed for the medium-frequency AC collection and uncontrolled rectification system of offshore wind power. Under this strategy, the grid-connected harmonic current is calculated using the instantaneous power theory in a two-phase stationary coordinate system, and the voltage waveform on the grid side of the wind turbine is actively adjusted by combining with model predictive control (MPC). This can effectively reduce the harmonic current generated by the DRU and ensure the safe and stable operation of the offshore power grid. Simulation and experimental results validate the effectiveness and feasibility of the proposed suppression strategy.

To deal with the over-voltage increase at the point of common coupling by the reactive power of a static synchronous compensator (STATCOM) due to its slow response speed when the grid voltage changes rapidly, a general criteria and technical indicators characterizing the over-voltage increase were proposed. Through a simplified analysis based on the magnitude of fundamental voltage, the severity of this over-voltage increase problem due to the slow response speed of STACOM was shown, and the quantitative requirement for calculating the AC voltage became clear. A fast voltage algorithm based on second-order generalized integrator and the corresponding coordination strategy were put forward. The main circuit topology of a STATCOM from a practical project and its control strategy were applied, and the digital simulation results from both a simplified Thevenin’s equivalent circuit and a complicated HVDC benchmark system showed that the proposed fast algorithm and coordination strategy can solve the over-voltage increase problem.

Island power supply systems typically exhibit weak grid characteristics, and the system frequency and voltage are prone to losing stability. The coordinated operation and control of wind power and hydrogen energy storage according to different characteristics are key solutions to this problem. On this basis, a grid-forming-based coordinated control method for an island wind-hydrogen system is proposed. By applying hydrogen energy storage into the grid-connected system of grid-forming wind turbines, the load fluctuations are suppressed and the frequency regulation capability of the system is enhanced, thereby improving the system’s frequency regulation capability, safety and stability. First, the wind-hydrogen system in island scenarios is modeled, the grid-forming control of a traditional wind turbine on the grid side is improved, and a grid-forming control strategy considering the output from hydrogen energy storage is put forward. Second, to better achieve the frequency regulation effect of hydrogen energy storage on the system frequency, an adaptive primary frequency control strategy for hydrogen energy storage is designed, so as to realize a wind-hydrogen system frequency coordinated support (WHFCS) strategy for the island wind-hydrogen system. Finally, a system model was built on the RT-LAB real-time simulation platform, and the effectiveness of the WHFCS strategy in supporting the frequency of the wind-hydrogen system was verified through comparison. Results show that the proposed control strategy has good applicability for island wind-hydrogen systems under weak grid conditions, providing a new scheme for improving the reliability and stability of island wind-hydrogen systems.