【摘要】 基于对负荷时间序列高阶矩时变特征的研究,提出了一种基于自回归条件密度模型的短期负荷预测新方法.该方法通过引入含时变参数的有偏分布,对负荷时间序列二阶以上矩信息进行了分析和描述.基于南京地区日用电量实际历史数据,分析了该负荷时间序列的时变高阶矩特征,建立了自回归条件密度模型.使用条件对数极大似然估计对模型参数进行了估计,实现了短期负荷预测,验证了该方法的可行性和有效性.结合算例中自回归条件密度模型时变参数的取值范围,推导了时变参数与条件高阶矩的数理关系,给出了一种刻画时间序列时变高偏度(三阶矩)、时变高峰度(四阶矩)的途径.算例分析表明,基于有偏t分布的自回归条件密度负荷预测模型的预测效果良好.更多还原

【Abstract】 By analyzing the time varying characteristics of the high order moments of load time series,a novel short term load forecasting method is proposed based on the auto-regressive conditional density( ARCD) model. By introducing the skewed distribution with time-varying parameter,the proposed method can analyze and describe the implicated information in moments higher than the second order of load time series. Based on the historical daily practical power consumption data of Nanjing,the time-varying high order moments of the load time series are examined,and the proposed ARCD load forecasting model is established. By using the conditional maximum likelihood estimation( CMLE),the parameters are estimated,and short term load forecasting is provided. As a result,the feasibility and effectiveness of the proposed model is validated. The mathematical relationship between time-varying parameter and conditional high moments is deduced considering the range of the time-varying parameter in the ARCD model. Specifically,the methods for describing the third order moment and the fourth order moment of time series are illustrated. Numerical results indicate that the ARCD model with skewed t distribution provides satisfying forecasting results.更多还原