To ensure the long-term safe and stable operation of electric vehicles and reduce the failure rate of lithium-ion batteries, aimed at the issue of whether the battery management system of electric vehicles can accurately and effectively detect the two important parameters of batteries, i.e., state-of-charge (SOC) and state-of-health (SOH), an improved convolutional neural networks-sequence-to-sequence (CNN-Seq2Seq) neural networks joint estimation method for the SOC and SOH of lithium-ion batteries based on CNN-long short-term memory (CNN-LSTM) neural networks was proposed. The results of comparative experiments on the public data set show that the proposed method improves the stability and accuracy of SOC and SOH estimation results for lithium-ion batteries.

The state-of-health (SOH) of batteries is an important indicator for evaluating their performance. Aimed at the difficulty in accurately estimating the SOH of batteries, a novel SOH estimation method for lithium-ion batteries was proposed by combining the hunter-prey optimizer (HPO) algorithm and long short-term memory (LSTM) neural network. Through the analysis of influences of current and temperature on the incremental capacity (IC) curve, the peak value of the highest peak in the IC curve and the corresponding voltage, temperature and current were introduced as the model input, and the dynamic parameters of LSTM neural network were adjusted using the HPO algorithm. Finally, the battery charging and discharging data under the condition of peak cutting and valley filling in the environment of energy storage were used for experimental verification. Results indicate that the SOH estimation method for lithium-ion batteries based on HPO-LSTM neural network has higher estimation accuracy compared with the traditional LSTM network methods, and it also has good robustness under different network training amounts.

With the widespread applications of energy storage plants in new power systems, improving the prediction accuracy of lithium-ion battery state-of-charge (SOC) is of significance for enhancing the safety and stability of energy storage plants and power systems. A battery SOC prediction method is proposed, which combines a multi-scale convolutional neural network (CNN) and a residual long short-term memory (LSTM) network. First, the multi-scale CNN is employed to automatically extract different types of features from the sensor data. Then, the features are fed into an LSTM network to mine the correlation in the time dimension. Meanwhile, residual connections are added at both ends of the LSTM network to avoid model overfitting. The comparative experiments on the measured data of the energy storage system show that the proposed model achieves the best prediction results in terms of all the indicators. In particular, the key features can be extracted better when the battery is under complex operating conditions, resulting in a better prediction performance.

During the normal utilization of a lithium-ion battery, its capacity will be regenerated in a short term due to factors such as internal chemical reactions and external environment. Aimed at the problems including the inaccurate estimation of state-of-health (SOH) which results from the ignorance of capacity regeneration, an optimal variational mode decomposition (OVMD) strategy based on the number of optimal layers is proposed. The low-frequency stable capacity sequence reflects the overall aging trend, while the high-frequency non-smooth capacity sequence reflects the local capacity regeneration. With the consideration of high- and low-frequency components and by introducing the weighted mean of vectors (INFO) algorithm to improve the support vector regression (SVR), a short-term SOH estimation model for lithium-ion battery is established. With the selection of NASA battery aging data sets, contrast schemes for short-term lithium-ion battery SOH estimation experiment are designed based on BP (back propagation), SVR, INFO-SVR and OVMD-INFO-SVR. Results indicate that in the case of reducing the capacity training sets, the OVMD-INFO-SVR model based on high- and low-frequency components has higher accuracy when it is used in short-term SOH estimation of lithium-ion batteries.

As a black-start power supply, distributed generation exhibits higher flexibility in terms of structure, and it is also more rapid in response speed. However, during the system restoration and construction process, the PV-storage black-start power supply based on self-synchronous voltage source control may experience frequency fluctuations, posing a risk of failure in the subsequent black-start process of power grid. To enhance the suppression effect on frequency fluctuations during the construction process, an improved self-synchronous voltage source control strategy for an interface inverter in the PV-storage black-start power supply is proposed. This improved control strategy improves the stability of the PV-storage black-start power supply, laying a foundation for the smooth implementation of subsequent black-start process. Meanwhile, simulation analysis and experimental verification were also presented, providing a detailed validation of the proposed control strategy and theoretical results.

Towards the goal of a high percentage of friendly consumption of offshore wind power, a novel control method for the parallel operation of a high-voltage direct current (HVDC) transmission system based on diode rectifier (DR) and an HVDC transmission system based on voltage source converter (VSC) is proposed. Both of the two transmission lines are connected to offshore wind farms (OWFs). The VSC-HVDC controls the voltage and frequency of the offshore AC power grid, i.e., the voltage of OWFs is controlled based on the active power balance of the AC busbar of VSC-HVDC. When the injected power of OWFs exceeds the active power reference value of VSC-HVDC, the voltage of OWFs will increase, and the DR-HVDC will be activated. In this case, the offshore AC power grid transmits the required power value of OWFs through the DR-HVDC line, while the system frequency is controlled by VSC-HVDC by meeting the reactive power balance. When the power transmitted by VSC-HVDC is the same as the set value, the DR-HVDC will be automatically shut down. The significant advantage of the proposed method is that the capacity of VSC-HVDC is only used to support the normal operation of the system. Once the OWFs are put into operation, the DR-HVDC will be automatically activated. Finally, simulation results demonstrate the effectiveness and super- iority of the proposed method.

To improve the test and evaluation technology system for fuel cell buses and effectively ensure their widespread applications, a comprehensive performance evaluation method for the power train of fuel cell buses is proposed, which combines the combination weighting method and subjective scoring method. First, a comprehensive performance evaluation system for the powertrain of fuel cell buses is constructed, and the test methods for each index are given. Then, the analytic hierarchy process and coefficient of variation are used to determine the subjective and objective weights of evaluation indexes, and the combination weighting method based on the game theory is used to obtain the comprehensive weights. On this basis, the subjective scoring method is used to score the test results of vehicles to be evaluated. At last, the comprehensive scores are obtained by weighted calculation with the comprehensive weight, which are further rated. The calculation result of an example shows that the comprehensive weight combining the subjective and objective weights over-comes the limitation of calculating the weight by only one single method and makes the evaluation results more objective and accurate, thus verifying the scientific nature and rationality of the evaluation system and method.

To improve the responsiveness, system efficiency and operation economy of a fuel cell/supercar-pacitor/lithium-ion battery hybrid energy storage system for trams, a multi-threshold energy management strategy for the hybrid energy storage system based on the operation states of trains is proposed. First, according to the actual operation states oftrains, different working conditions of tram driving are divided and multiple thresholds are set, so as to realize multi-media and multi-scenario power exchange and achieve real-time control of power allocation among energy sources of the energy storage system. At the same time, according to the physical characteristics of each energy storage medium, the supercapacitor is set to start as priority, thus reducing the number of cycles of lithium-ion batteries and extending the service life of the energy storage system. Second, the life cycle cost of the hybrid energy storage system is taken as an optimization objective, and the power of the energy storage system, energy demand and state-of-charge (SOC) of the lithium-ion battery and supercapacitor are used as constraints, so as to optimize the capacity configuration of the energy storage system and obtain the optimized capacity configuration parameters. Finally, the proposed strategy is compared with the fixed-threshold strategy and traditional equivalent consumption minimization strategy (ECMS). Results show that the proposed strategy is more reasonable in terms of power allocation of the hybrid energy storage system, and it can improve the system response speed. The SOC offset ranges of lithium-ion battery and supercapacitor are significantly improved. The hydrogen consumption is reduced by 1.47 kg and 0.81 kg compared with those under the fixed-threshold and ECMS strategies, respectively, indicating that the fuel economy is improved.

The power supply system in a data center is its critical infrastructure, which provides stable and reliable power support for all the power-driven equipment therein to ensure itsnormal operation. With the rapid development of technologies such as big data, cloud computing and 5G, the number and scale of data centers in China are increasing rapidly, requiring higher reliability of power supply and increasing the overall power consumption. In addition, the problem of high carbon emissions is even more obvious because the data centers at present mainly use electricity from the traditional energy sources. Therefore, how to achieve a green, energy-saving and low-carbon power supply and distribution mode is the focus of the development of data center power supply systems. The functions and characteristics of typical power supply architectures of data center power supply system are analyzed, and the development and evolution of power supply architectures of data center are introduced. On this basis, “carbon peaking and carbon neutrality” is taken as the background, and a prospect for the power supply architectures of data center in the future is given combined with the problems encountered in the development of power supply system of data center, providing reference for its in-depth research and subsequent development.

In the medium- and high-speed operation ranges, the sliding mode observer (SMO) usually obtains the rotational speed and rotor position of a permanent magnet synchronous motor (PMSM) based on a formula method or phase-locked loop. However, the phase delay caused by a low-pass filter, the parameter variation in the phase-locked loop itself and the sliding mode chattering remain the key issues that need to be addressed. Under this background, a position sensorless control method for PMSM based on an improved SMO is proposed. First, a variable sliding mode gain related to the rotational speed is designed, which can achieve adaptive adjustment of the sliding mode gain parameter when the rotational speed changes, thereby reducing the system chattering and improving the system stability and dynamic performance. Second, based on the mechanical motion equation, a method for estimating the rotational speed and rotor position is put forward, which not only simplifies the estimation process but also effectively reduces the influence of changes in rotational speed on the estimation accuracy and enhances the system’s anti-interference capability. Finally, the feasibility and effectiveness of the proposed method were verified based on bench experiments.