作　　者： ; ; ;

机构地区： 昆明理工大学电力工程学院

出　　处： 《电工技术学报》 2020年第11期2349-2359,共11页

摘　　要： 该文提出一种基于贪婪搜索的神经网络(GSNN)算法,以实现非均匀温差分布下集中式温差发电系统的最大功率跟踪(MPPT)。首先,以DC-DC升压变换器占空比为输入,系统输出功率为输出,建立双层前馈神经网络模型。然后,采用Levenberg-Marquardt法训练神经网络,得到系统的输入-输出(I/O)拟合曲线,以准确区分局部最大功率点(LMPP)和全局最大功率点(GMPP)。同时,基于拟合的曲线,设计压缩范围的贪婪策略快速逼近GMPP。恒定温度、阶跃温度和灵敏度分析三种算例下的仿真结果表明,与扰动观测法、粒子群算法和灰狼算法相比,GSNN能在非均匀温差分布下快速、稳定地输出最大功率。最后,基于dSpace的硬件在环(HIL)实验验证了所提算法的硬件可行性。 A novel greedy search based neural network(GSNN) for centralized thermoelectric generation(TEG) system under non-uniform temperature distribution(NTD) condition is designed to achieve maximum power point tracking(MPPT) in this paper. Firstly, a two-layer feed-forward neural network model is established, where input is defined as duty cycle of DC-DC boost converter while output as power output of TEG system. Then, Levenberg-Marquardt method is adopted to train neural network, accordingly the I/O curve that draws an evident distinction between local maximum power point(LMPP) and global maximum power point(GMPP) is fitted. Meanwhile, a greedy search is well performed to explore GMPP over a compressed search range. Lastly, three case studies are carried out, i.e., constant temperature, step temperature and sensitivity analysis. Simulation results demonstrate that GSNN could achieve maximum power output with speediness and stability under NTD compared to perturb and observe(P&O), particle swarm optimization(PSO) and grey wolf optimizer(GWO).Furthermore, a d Space based hardware-in-the-loop(HIL) experiment is undertaken to validate implementation feasibility of the proposed algorithm.

关 键 词： 集中式温差发电系统 非均匀温差分布 最大功率跟踪 神经网络

领　　域： [] []