基于粒子滤波和主成分分析优化的极限学习机在风电场功率预测的应用

Application of Extreme Learning Machine Based on PF and PCA Optimization in Wind Farm Power Prediction

摘要: 风力发电的波动性给实现电网安全运行带来了挑战，因此，需要准确预测风力发电量。提出了一种基于粒子滤波和单隐层神经网络极值学习机的短期风电预测方法，并引入主成分分析对数据进行降维。极限学习机具有快速的训练速度和良好的泛化性能，它可以在同一时间和空间学习风速、风向等气象因素与海上风电场发电量之间的关系，再加上粒子滤波对数据的预处理，更新训练数据和网络结构。实现气象数据的快速实时校正和风电机组输出功率预测。与现有预测规模型相比，本实验模型平均绝对百分比误差最低为5.81%，结果验证了所提预测方法的可行性和优越性。

Abstract: The volatile nature of wind power generation creates challenges in achieving secure power grid op-erations, therefore, necessary to accurately predict wind power. This study proposes a short- term wind power prediction method based on Particle Filter and single hidden layer neural network Ex-treme Learning Machine and introduces Principal Component Analysis to reduce the dimensionality of the data. ELM has a fast training speed and good generalization performance, it can learn the re-lationship between wind speed, wind direction and other meteorological factors and power genera-tion in offshore wind farms at the same time and space, coupled with the preprocessing of the data by the Particle Filter, update training data and network structure. Realize rapid real-time correc-tion of meteorological data and wind turbine output power prediction. Compared with the existing prediction scale model, the average absolute percentage error of the experimental model is 5.81%, comprehensive results validated the feasibility and superiority of the proposed prediction ap-proach.