Artificial Intelligence-Aided Minimum Reactive Power Control for the DAB Converter Based on Harmonic Analysis Method
Abstract : With the aim of reducing the reactive power for the dual-active-bridge (DAB) converter, this letter proposes an artificial intelligence (AI) aided minimum reactive power control scheme based on the harmonic analysis method. Specifically, as an advanced algorithm of the deep reinforcement learning (DRL), the deep deterministic policy gradient (DDPG) is used to train an agent off-line. During the training of DDPG algorithm, the three-phase-shift (TPS) modulation is adopted and the zero-voltage-switching (ZVS) constraints are considered. Thus, the trained agent of the DDPG which likes an implicit function, can provide optimal control strategies for the DAB converter in real-time with the minimum reactive power and soft switching performance in the continuous operation range. Finally, experimental results validate the feasibility and correctness of the proposed AI based optimized method.
? In vision applications, the (target) domain of interest contains very few labeled samples with limited knowledge, while an existing (auxiliary or source) domain is often available with a large number of labeled examples and useful knowledge but lying different distribution from target domain.
? Smart and Ubiquitous computing eases the existing traffic scenarios on roads.
? Comparison between existing techniques and new techniques for action recognition along with hardware applications will give greater insight for real time applications.
? Thus there is greater potential in city-industrial clusters where there are a good number of diversified institutions coexisting with equally diversified industrial setup.
? Many advanced iterative techniques have been utilized for solving optimization problems such as the Lagrange multiplier method, Newton’s method, mathematical programming methods, and genetic algorithm (GA).
? It can be seen as a mathematical optimization problem that is aimed at finding the global extremum of the function subject to equality constraints.
? Furthermore, it can be seen that the impact of the number of harmonics components is decreasing with the increasing sum terms.
? To address this issue, particle swarm optimization (PSO), a method inspired by insect swarm behavior for food search, can be easily applied for the control optimization problems such as the passive filter design, elimination of harmonics, and resonant controller parameters tuning.
• This paper proposed a method in which Gravitational search algorithm (GSA) has been used to minimize the objective function.
• The proposed approach estimates the d and q axis synchronous reactance and the open circuit flux.
• The load rejection test of a combined resistive load is performed for parameter identification where online symmetrical three phase short circuit test is proposed for the model cross validation.
• In this proposed work, error function is considered as the objective function.
• The MMC was first proposed in 2003 by Marquardt, and since then it has been an important focus of research for industry and universities.
? The performance in terms of reactive power and efficiency were investigated for a wide input voltage range and different load conditions.
? Based on above analysis, it can be seen that the POPS has very similar performance as MPS control, while MPS can slightly reduce reactive power and improve efficiency in light load condition.
? It validates the correct operation of the output voltage closed-loop control and also fast dynamic performance of the POPS algorithm.
? Especially in the high voltage-conversion-ratio condition, SPS will result in increased reactive power, high current stress and narrower soft-switching region, which finally affect the conversion efficiency.
? Moreover, the efficiency optimized control scheme including conduction losses and switching losses are also presented in.
|