基于核函数极限学习机的微电网短期负荷预测方法

Online Short-Term Load Forecasting Based on ELM with Kernel Algorithm in Micro-Grid Environment

针对以单个或集体用户为业主的用户侧小容量微电网,考虑到成本约束及用电特征的不确定性,提出了一种基于核函数极限学习机的微电网短期负荷预测方法。使用核函数极限学习机、启发式遗传算法和分时训练样本,建立了包含离线参数寻优与在线负荷预测的预测模型;通过模型参数的周期更新来保证算法最优参数的时效性,同时降低在线预测系统的计算复杂度与历史数据存储量。通过对不同容量、类型的用户侧微电网进行短期负荷预测,分析了预测结果的准确度、参数周期更新的效果、预测结果对经济运行的影响和预测方法的计算效率。

Considering the cost constraints and various electrical characteristics of the small capacity user-side micro-grid constituted by single or group users, a short-term load forecasting method based on extreme learning machine with kernel(ELM_k) algorithm is proposed. The ELM_k, heuristic genetic algorithm and time division training samples are used to establish a short-term load forecasting model, including offline parameter optimization and online load forecasting. The cycle update of model parameters guarantees the timeliness of the optimum parameters, and reduces the computational complexity and storage space of the online forecasting system. The load forecasting of user-side microgrids with different capacities and types is processed, and the load forecasting accuracy, the model performance after cycle update, the micro-grid operation costs under load forecasting result and the calculation efficiency of this method are analyzed.