To meet the requirements of long life, high efficiency, low ripple and intelligence of a high-power LED light source, based on an in-depth analysis of the characteristics of low-frequency secondary ripple compensation of a Boost asymmetric half-bridge (AHB) converter LED constant-current driver without an electrolytic capacitor, an adaptive low-frequency ripple compensation control algorithm for an AHB LED constant-current driver in digital predictive current mode was proposed. MATLAB simulation results show that the proposed control algorithm can adapt to the changes in input voltage and load and effectively reduce the low-frequency secondary ripple current flowing through the LED in the whole dimming range.

The grid-forming cascaded H-bridge energy storage system is an optimal solution for mitigating large-scale renewable energy power fluctuations and providing grid support. However, the power control of the system is coupled with line impedance, and the coupling effect intensifies under weak grid, leading to output power inaccuracy and increasing risk of instability. To solve this problem, an adaptive power decoupling control strategy based on reactive voltage compensation is proposed. First, based on the small-signal model of a virtual synchronous generator and through the introduction of power coupling coefficients which take into account closed-loop feedback, a quantitative analysis is conducted on the power coupling effects between line impedance angle and power angle, revealing the mechanism of effect of active power on reactive power control. Second, an adaptive reactive power compensation channel is constructed, and a proportional-integral voltage compensation link in the reactive power control is designed, which dynamically modifies the output voltage references to mitigate the power coupling without the identification of line parameters. As a result, dynamic decoupling control is realized in a wide range of impedance angle with strong adaptability. Finally, validation tests based on a simulation platform and an RT-LAB platform verified the proposed method.

Aimed at the problem of insufficient inertia and damping of a microgrid resulting from a large number of power electronic converters connected to power grid, a virtual DC machine enhanced inertia control strategy suitable for DC microgrid is proposed. First, based on the frequency division control of the hybrid energy storage unit, the virtual DC machine technology is introduced to control the energy storage converter. Second, according to the instantaneous power balance characteristics of DC microgrid, the fluctuation of load or photovoltaic output controlled by first-order high-pass filtering is taken as an additional amount of rated power controlled by the conventional virtual DC machine. Third, a regulation factor related to the auxiliary voltage regulation power and the state-of-charge of the supercapacitor is introduced to adaptively optimize its working state. Finally, the control strategy was verified by constructing an experimental platform, and results show that this strategy can improve the transient stability of DC bus voltage.

Aimed at the problem that the traditional droop control allocates load power based on a fixed droop coefficient without considering the influence of line impedance or the information about energy storage unit capacity, as well as the problems of high system communication pressure and complex control structure, a control strategy for the coordinated operation of multi-energy storage without voltage observers is proposed. First, the proposed strategy introduces a convergence function including the capacity of distributed energy storage units (DESUs) and state-of-charge (SOC) to correct the initial droop coefficient, and it gives constraints on the adjustment coefficient, thereby promoting the rapid SOC equilibrium of the energy storage unit. Second, by introducing a virtual rated power, the unit output power is configured and balanced to overcome the effect of line impedance on the load sharing of DESUs. Through the construction of a product of maximum allowable bus voltage deviation and average unit power, the bus voltage compensation amount is obtained, so as to realize voltage regulation without the need of communicating and exchanging the node voltage information about the energy storage unit or the design of the corresponding voltage observer and PI controller. Finally, the effectiveness of the proposed strategy was verified by simulation tests using MATLAB/Simulink.

With the development of offshore wind power in China advancing to the deep sea, there are no buoys in the deep sea area, the wind speed near the coast is not large and unstable, and the flotation device is affected by complex factors including extreme environmental conditions (e.g., waves and ocean currents), so there exist problems such as the poor control effect of floating offshore wind turbines due to the low accuracy of wind speed measurement data. Under this background, an intelligent power control method for the converter in a deep-sea floating offshore wind turbine is proposed, which is based on a neural network effective wind speed estimator. The control system is composed of a proportional-integral-derivative (PID) controller, lookup table, a neural network estimator and a virtual sensor, in which the neural network estimator is trained online to adapt to the transformation of the environment, so as to estimate the effective wind speed of the floating offshore wind turbine and predict its future value. In addition, the current and future wind speed data are combined and further provided to the controller. The virtual sensor can also provide measurement of the effective wind speed based on other available factors (e.g., the pitch angle and angular velocity of a fan) related to the power generation of the fan. Finally, the effectiveness of the proposed control method is verified by simulation analysis. Compared with the traditional PID, this method can improve the accuracy of sinusoidal wind measurement by about 16%, and the average improvement rate is about 8%.

An optimized control scheme for a single-phase grid-forming photovoltaic-supercapacitor system considering dynamic response improvement and double-line frequency power decoupling is proposed. Compared with the conventional scheme, this scheme integrates the supercapacitor energy storage with double-line frequency power decoupling, and there is no need for additional power decoupling circuits. The DC-link proportional integral resonant control, virtual rotor kinetic energy feedforward control and photovoltaic power feedforward control suitable for this topology are studied, so that the capacitance of the DC-link capacitor and supercapacitor is reduced. In addition, a small-signal linearized model for the second-order generalized integrator-based power calculation part is established, and a sector-based state feedback damping control scheme is put forward to improve the dynamic performance of the system. Experimental results verified the effectiveness of the proposed control scheme.

Aimed at the problem of over-smoothing in the process of wind power fluctuation smoothing by means of traditional wavelet packet decomposition, an improved wavelet packet decomposition method is proposed based on the time-frequency analysis by introducing the smoothing coefficients of high-frequency components. First, the set of time when the power fluctuation crosses the limit is found, and the components ineach frequency band are obtained using the traditional wavelet packet decomposition. Then, the optimal smoothing coefficients are determined based on an objective function and constraints, and the power components in the set of limit-crossing time are suppressed in turn, starting from the highest-frequency components until the fluctuation limits under the two time windows are satisfied. Finally, the smoothing effect is evaluated. Through the simulation analysis of an example, it is verified that under the premise of satisfying the fluctuation limits, the improved wavelet packet method can reduce the charge and discharge accumulation of energy storage and maintain the power output level of energy storage in a better state under different scenarios.

Based on the equivalent circuit model of proton exchange membrane fuel cell (PEMFC) and the electrochemical reaction theory, a fuel cell power generation efficiency calculation model is established to study the influences of operating conditions (e.g., working temperature and working humidity) on the internal resistance characteristics, output characteristics and power generation efficiency of the fuel cell stack. The principle of minimizing the total DC internal resistance inside the stack is adopted to obtain the optimum operating conditions under the output of different current densities. Experiments on AC impedance characteristics and U-I output characteristics are designed to verify the accuracy and effectiveness of the model. Simulation and experimental results indicate that the internal resistance characteristics and output current density are important factors affecting the variation in fuel cell power generation efficiency. The operating conditions corresponding to the minimum DC total internal resistance inside the stack can optimize the output performance of the stack and improve the power generation efficiency. The fuel cell power generation efficiency model established can accurately reflect the internal characteristics of the fuel cell stack, and the calculation results are effective and reliable.

To solve the inconsistency problem of series battery packs, a dual-layer topological structure with an improved Cuk chopper circuit in the outer layer and a single-inductor balanced topology in the inner layer is proposed. The improved Cuk chopper circuit is adopted, and a dual-layer switch is used to connect the balanced battery pack to the Cuk equalizer, so as to shorten the energy transfer path and improve the energy transfer efficiency. The use of the zero voltage switch technology ensures continuous balanced current and small fluctuations therein. The single-inductor balanced topology is used, which has a simple structure and high accuracy. The proposed dual-layer balanced topology can achieve energy transfer between any individual cells within one pack and that between any battery packs. With the state-of-charge of the battery as an equilibrium evaluation indicator, a balanced circuit simulation model is set up in MATLAB/Simulink, which is further compared with the traditional balanced topology in terms of static equilibrium and discharge equilibrium. Results show that the proposed balanced topology shortens the balancing time and improves the balanced efficiency. A battery balancing experimental platform was established, and discharge experiments were carried out on 12 batteries, verifying the accuracy of the hierarchical balanced topology.

To accurately determine the battery capacity, long short-term memory (LSTM) neural network was used to estimate the battery capacity. First, the life-cycle curves for different battery parameters were analyzed. The Pearson correlation coefficients between the parameters and battery capacity were calculated, and the battery voltage, internal resistance, equal voltage drop time and other parameters were selected as health factors to construct a battery capacity estimation model. Then, the Center for Advanced Life Cycle Engineering (CALCE) battery data set was used for model training and capacity estimation, and the average percentage error of the estimation model was 1.19%. The causes of estimation errors were analyzed, and the model was optimized by revising the battery’s initial capacity and aging parameters. The optimization results show that the estimation error of the model constructed using battery voltage, internal resistance, constant-current charging time and 4.0–3.4 V equal voltage drop time was about 0.55%.