摘要
为了有效利用光伏阵列转换能量,提高并网发电效率,需要对其进行最大功率点跟踪控制(MPPT)。提出了基于二级神经网络-遗传寻优的方法,通过利用PV神经网络模型拟合光伏电池输出功率与输出电压的非线性特点,为MPPT寻优提供模型依据,采用遗传算法编码灵活的特性实行并行搜索,并采用存储器函数变换技术使得系统达到在线控制的效果。仿真及实验表明,遗传算法对PV曲线进行最大值寻优,经过52代遗传得到最大功率为135.8114W,对应的电压值为72.13824V,这与实际最大功率点相比的误差为1.45%,取得了精度较高的跟踪效果,提高了系统稳定性。并且该方法能够准确跟踪最大功率点,克服了传统爬山法等在最大功率点附近振荡引起功率损耗的问题,同时也克服了传统神经网络方法采集训练数据的难度。
In order to convert energy efficiently with photovoltaic arrays and improve efficiency of a photovoltaic grid-connected power system, it is necessary to perform the Maximum Power Point Tracking(MPPT). A novel method based on 2-level neural networkgenetic algorithm optimization is presented to fit nonlinear characteristic for output power and output voltage of photovoltaic array with PV neural network model and provide prototype proofs for optimization of MPPT. In addition, flexible encode of genetic algorithm, parallel search and the function transform of semiconductor memory technology are employed for this method to realize the on-line control. Computer simulation and experiment results prove that the genetic algorithm searches the maximum point of PV curve after 52 generations and obtains the MPP 135.811 4 W and corresponding voltage 72.138 24 V, the result error comparing to the actual value is only 1.45%, and tracking efficiency with high precision and strong robustness are obtained with the proposed method. Moreover power loss caused by oscillation at maximum power point for traditional methods such as hill climbing method is overcome as well as the difficulty of training data acquisition for traditional neural network method.
出处
《电力系统保护与控制》
EI
CSCD
北大核心
2011年第20期38-42,共5页
Power System Protection and Control