基于多尺度时序建模与估计的电力负荷数据恢复

Power Load Recovery Based on Multi-Scale Time-Series Modeling and Estimation

针对电力负荷数据缺损及失真问题,从时序数据特性分析及建模与估计的角度给出负荷数据补全与恢复的方法。运用马尔科夫链与序贯蒙特卡洛模拟联合法抽取负荷统计特性;基于电力负荷波动的年、月、周、日的多尺度时序特征分析,建立电力负荷的多尺度时序特征建模。引入B-spline基函数展开法解决负荷模型的非参、变系数问题,并给出负荷模型中关键参数的估计方法。采取误差多指标评判方法确定B-spline节点最优数量与样条最优次数。根据所得负荷恢复模型提出周尺度的电力负荷缺失数据恢复方法,并给出年度等长时段日负荷数据恢复思路。经实际算例验证,该文所提方法准确有效,具备工程应用价值。

Aiming at the defects and distortions of power load data,this paper presents a method of load data completion and recovery from the perspective of processing and characteristic analysis,modeling and estimation of time series data.Markov chain analysis and sequential Monte Carlo simulation methods were used to extract load statistics characteristics,then based on the power load fluctuation characteristics analysis of year,month,week and day,the multi-scale time-series characteristics of the load were modeled.B-spline basis function expansion method was introduced to solve the problems caused by non-parametric and variable coefficients characteristics of the load model,and an estimation method of the key parameters in the load model was provided.The optimal number of B-spline nodes and the optimal number of spline splines were determined by the multi-index error evaluation method.Based on the known load recovery model,a weekly lost power load recovery method is proposed,and the idea of daily load data recovery for the whole year or long duration is obtained.The method proposed in this paper is proved to be accurate and effective by practical examples and is of considerable engineering application value.