风光互补发电系统的分布式模型预测控制

Distributed Model Predictive Control of Wind-solar Hybrid Power Generation System

风光互补发电系统中,风力和光伏独立发电且两者在地理上相隔较远,彼此没有通讯交流。对此问题,提出用分布式模型预测控制的方法去解决。首先,在风光互补发电系统中存在大量的非线性环节,运用神经网络线性逼近,训练得出各个子系统的神经网络线性化模型。然后,在此基础上,基于风力优先发电、光伏配合、蓄电池必要时输出的原则,设计出满足相应的约束条件且合理的目标函数并利用二次规划的方法求解出最优值。最后,研究结果表明,针对系统负荷端的需求,分布式模型预测控制方法可以在理论及实际真实天气条件下去合理地分配各个子系统的输出功率。相对于其他的传统方法,分布式模型预测控制在优化系统性能方面速率高且效果好。

In the wind-solar hybrid power generation system, wind power and photovoltaic power generation are independent and the two system are far apart, and there is no communication with each other. To solve this problem, a distributed model predictive control method is proposed. First of all, there are a large number of nonlinear links in the wind-solar hybrid power generation system. Using the ability of neural network linear approximation, the neural network linearization model of each subsystem is trained. Then, based on the principles of wind power generation priority, photovoltaic cooperation and battery output when necessary, a reasonable objective function that satisfies the corresponding constraints is designed and the optimal value is solved by the method of quadratic programming. Finally, the research results show that, in response to the needs of the load side of the system, the distributed model predictive control can reasonably allocate the output power of each subsystem under theoretical and actual weather conditions. Compared with other traditional methods, distributed model predictive control has a high rate and good effect in optimizing the performance of the system.