基于Spark和IPPSO_LSSVM的短期分布式电力负荷预测算法

Distributed short-term load forecasting algorithm based on Spark and IPPSO_LSSVM

为了提高电力负荷预测的精度,应对单机运算资源不足的挑战,提出一种改进并行化粒子群算法优化的最小二乘支持向量机短期负荷预测模型。通过引入Spark on YARN内存计算平台,将改进并行粒子群优化(IPPSO)算法部署在平台上,对最小二乘支持向量机(LSSVM)的不确定参数进行算法优化,利用优化后的参数进行负荷预测。通过引入并行化和分布式的思想,提高算法预测准确率和处理海量高维数据的能力。采用EUNITE提供的真实负荷数据,在8节点的云计算集群上进行实验和分析,结果表明所提分布式电力负荷预测算法精度优于传统的泛化神经网络算法,在执行效率上优于基于Map Reduce的分布式在线序列优化学习机算法,且提出的算法具有较好的并行能力。

Aiming at the insufficient resource of single computer,a short-term load forecasting model based on LSSVM（Least Squares Support Vector Machine） optimized by IPPSO（Improved Parallel Particle Swarm Optimization） algorithm is proposed to improve the accuracy of load forecasting. A Spark-on-YARN memory computing platform is introduced and the IPPSO is operated there to optimize the uncertain parameters of LSSVM,which are then applied in the load forecasting. The parallel and distributed computation is adopted to improve the accuracy of forecasting algorithm and the capability of massive high-dimensional data processing. Experiment and analysis are carried out with the actual load data provided by EUNITE on an 8-bus cloud computing platform and results show that,the proposed algorithm has better accuracy than the generalized traditional neural network algorithm and better efficiency than the MR-OSELM-WA（Map Reduce-Online Sequential Extreme Learning Machine-Weighted Averaged） algorithm.