永磁同步风电机组神经网络滑模多目标优化控制

Multi-objective optimization of neural network sliding mode control for permanent magnet synchronous wind turbine

将永磁同步风力发电机组中的变流器和电网用等效负载代替并对控制回路进行简化,得到非线性仿射形式的机组模型,利用反馈线性化方法对系统进行精确线性化。固定参数离散指数趋近律滑模控制算法主要缺陷是如两个参数匹配不当,可能会使求得的控制量过大,同时系统在滑模面附近剧烈的高频抖振会导致机组所要承受的机械应力增加,动态性能变差,利用神经网络的自适应学习能力对这两个控制参数进行实时优化,根据机组控制目标定义一个综合性能指标,通过优化该指标得到网络权值修正算法。仿真结果表明,该方法可以使系统快速到达滑模面,实现了机组对最优转速的快速跟踪;同时有效抑制了系统的抖振,减小了额外的疲劳载荷,实现了多目标优化控制。

By replacing convertors in permanent magnet synchronous wind turbine and power grid with equivalent load and simplifying control loop, a nonlinear affine model of the unit is obtained and the system is exactly linearized by feedback linearization method. However the sliding mode control method based on discrete exponent reaching law with fixed parameters has a main defi- ciency of too much control output if two parameters don＇t match perfectly ,meanwhile the high-fre- quency buffeting can enhance the mechanical stress that the unit is born and cause worse dynamic performances of the unit. To solve this problem, the adaptive learning ability of neural network is used to perform a real-time optimization for the two control parameters, an integrative performance index is defined according to control aims of the unit, and then the performance index is optimized to get a modified algorithm of the network weight values. The simulation results show that the pro- posed approach can make the system reach the sliding mode plane rapidly, track the optimal speed quickly, effectively prevent system from buffeting, reduce extra fatigue load and achieve multi-ob- jective optimization control.