Time Series Forecasting Method And Its Application In Power System

Forecasting plays a crucial role in decision-making processes.Time series forecasting is a solid foundation for power system planning,scheduling and control.The advent of the era of future smart grid and power big data,and the increase of uncertainties in the power system put forward higher requirements for the forecasting breadth and precision of power system time series.The cumulative effect is widely found and considered in time series forecasting.Accumulation of influencing factors during several consecutive time periods makes the variation of target parameters lags behind the variation of their influencing factors.The fact that the effects of cumulative effect are not addressed in time series forecasting has a tendency to result in larger forecasting errors.In view of the problems above,this paper take into account the dynamics of the time series variation in time series forecasting to more accurately reflect the cumulative effect and improve the forecasting accuracy.Specific studies are as follows:The characteristics of time series,influence mechanism of influencing factors and cumulative effect are revealed,which provide the basis for time series forecasting.In the analysis of the general characteristics of the time series,trends,seasonality and residuals of the time series are revealed.In terms of power system load analysis,the annual/monthly,weekly and daily load characteristics of the power system are revealed based on the actual load data,and factors such as meteorology and day type that affect power system load are summarized through correlation analysis;In terms of electricity price analysis,the fluctuation and periodicity of electricity price are summarized based on actual electricity price data,and factors such as load and primary energy price that affect electricity price are analyzed and summarized through correlation analysis.In addition,the typical phenomena and effect of the cumulative effect on the time series are also analyzed with examples and the general effect of cumulative effects is revealed.In order to reflect the influencing process of the cumulative effect as accurately and truly as possible in time series forecasting,so as to obtain a better forecasting result,the dynamic similar sub-series method is proposed in this paper.The internal variations of target parameters sub-series and each influencing factor sub-series are integrated into the selection rules for the dynamic similar sub-series by the weighted sum method and the fuzzy clustering method.The corresponding forecasting algorith m is designed according to the variation of the dynamic similar sub-series.The proposed method is applied to the short-term daily average load forecasting and the electricity price forecasting,and the accuracy and effectiveness of the proposed method in time series forecasting are verified in comparative experiments.Dynamic similar sub-series method for variable time window sizes is proposed based on the dynamic similar sub-series method.Quantitative and qualitative analysis of the influence of different factors on the selection of the size of the time window is carried out,and the conclusion is drawn that the cumulative effect different influencing factors exert on target parameters may not be the same in the actual situation.The optimal time window sizes are selected for different sub-series in the proposed method.Through testing and comparison in short-term daily average load forecasting,the advantage of this method in forecasting accuracy is verified.This dissertation is focused on the cumulative effect of the influencing factors on the target parameters in time series forecasting and two dynamic similar sub-series forecasting methods are proposed.The comparison tests in different scenarios show that the proposed method can improve the forecasting accuracy of power system time series.