Improved short-term load forecasting can be obtained by carefully selecting similar days in an integrated model for short-term load forecasting. Traditionally, similar days are selected based on experience, which can affect forecasting accuracy. Three methods were developed to classify trend-similar days from shape-similar days for short-term load forecasting in this paper. The three methods were compared to develop an effective forecasting method. Numerical tests verified the efficiency of the method.

中国实用的短期负荷预测系统普遍采用的是综合预测模型机制.文中深入研究了该机制的实现和应用原理,针对其权重不等式约束、求解方案实用化方面进行了深入探讨,提出了"最优拟合模型"不等于"最优预测模型"的观点.对该观点进行了详细的论证.在此基础上,提出了"较优预测模型"的实现策略,并以实例证实该模型提高了负荷预测的准确率和误差稳定性.

The loads collected yesterday were used to predict the loads tomorrow in conventional short-term load forecasting concept. The load data currently collected were not adopted in conventional method. In order to meet the requirement of electricity marketing in China, a new concept of Extended Short-Term Load Forecasting is defined in this paper. A new forecasting mode is designed so that the historical load information just collected can be utilized by this method to refresh the forecasted loads of the rest periods in current day. An integrated optimization model is presented and a method to select optimal weights in the forecasting model is given. Operation practice of the method applied to a practical power utility shows that an accuracy improvement of more than 2% has been achieved by the proposed method. Possible areas of application of this method are also discussed.

Using the latest historical load data, the extended short-term load forecasting method can forecast the unknown load during several hours after the current hour in that very day. The forecasting accuracy is much higher than those from conventional short-term load forecasting methods. In order to render improvement of the accuracy of short-term load forecasting, this method is applied to revise the bad data in historical load data record and to complement the unknown load data of that very day. The backgrounds and implementation principles for these two applications are presented in detail and the results of practical examples show that the above mentioned application is effective and practicable.

Aimed at unattended load forecasting,in auto-operating short-term load forecasting system for Shenzhen power network the automatic closed loop operating modes of information acquisition,load forecasting and reporting the schedules to higher authorities have been implemented.The authors integrate both intelligent self-diagnosis module and self-management module with original system,implement the functions of auto-diagnosis of error in forecasting process and auto-management of operation results,and improve the unattended mode to meet the requirement of “reducing the personnel and improving benefit” in electric power industry reformation.In addition,the contents on concepts and methods of forecasting algorithm organization and reasonable prediction precision analysis are presented,and the practical operation situation of the system is summarized.

商业性城市负荷有其典型的特点,只有充分分析这些特点,才能有针对性地修正负荷预测算法,使得这些算法在预测上达到有效、准确.鉴此 ,以深圳市为代表分析了商业性城市负荷的特征及其原理,并提出有针对性的预测算法修正模式,实现了深圳市电网负荷的较高准确率预测.

Taking account of the demand of electricity market in China. a practical design of short-term load forecasting system is presented in this paper. The design is composed of four components. The first component collects and analyzes pre-forecasting data for enhancing its validity and availability. To offer a foundation for adjusting the day load plan, a conception of extended short-term load forecasting is defined, which provides a function to re-forecast the loads from now to the end of time period in one day. As a whole, short-term, extended short-term and very short-term load forecasting sections are combined into a main component. Original load data is detected and modified by data checking section, which, with the plan checking section, forms the third component to control the forecasting system. Based upon forecasted results, the day load plan is worked out and reported by the final component. The design principles and implementation methods are described completely. The practicability of this design is demonstrated by results of applied cases.