双馈风机自适应神经分散协调预测控制
Adaptive neural decentralized-coordinated predictive control of double fed induction generator

目前, 双馈感应发电机转子侧励磁控制系统均依据“孤立”模型设计. “孤立”模型忽略了各子系统 之间、各控制器之间的相互作用, 因此这种控制器仅对改善本系统的控制特性有一定作用. 针对 以上情况, 提出了一种自适应神经分散协调控制策略, 并将其应用于双馈感应发电机转子侧励磁控制系统 仿真研究中. 首先, 利用电力关联测量法建立了基于本地变量的双馈风机关联测量模型. 其次, 以关联 测量模型作为预测模型, 采用多模型预测控制器对双馈风机转子侧励磁系统进行控制. 最后, 利用可在线 调整的人工神经网络作为多模型加权控制器以补偿双馈风机强非线性、工作区间变化范围大的特点. 主导特征值分析和动态仿真表明: 该控制策略不仅实现了高精度的最大功率跟踪控制, 而且在电力系统 故障时可提供持续的、充足的阻尼.
At present, all designs of rotor-side excitation control system of double fed induction generator (DFIG) are based on the stand-alone machine model, in which the interactions between subsystems and existing controllers were not considered. In this situation, those “stand-alone machine”-based controllers will have only certain effects on improving the local system dynamics. Considering the problems above, we propose the adaptive neural decentralized-coordinated predictive control (ANDPC). Firstly, the interaction measurement method is introduced to build the local signal-based interaction measurement model (IMM) of DFIG. Secondly, a multiple model predictive control scheme based on the obtained IMM is proposed to control the rotor-side excitation system of DFIG. Finally, an artificial neural network (ANN) trained online is employed as a weighting controller to cope with the nonlinearities and the large operating range of DFIG. The dominanteigenvalue analysis and dynamic simulations demonstrate that the proposed ANDPC scheme not only achieves an accurate maximum power point tracking (MPPT) control performance, but also provides a consistently enhanced contribution to network damping over the full operating range.