外文翻译-应用识别和人工神经网络在电力系统负荷预测.doc

1、英文部分Application of Pattern Recognition and Artificial Neural Network to Load Forecasting in Electric Power System ABSTRACT Electric power system load forecasting plays an important role in the Energy Management System (EMS),which has great influence on the operation,controlling and planning of elect

2、ric power system. A precise electric power system short term load forecasting will result in economic cost saving and improving security operation condition. With the develop-ment of deregulation in electric power system,the method of short term load forecasting with high accuracy is becoming more a

3、nd more important. Due to the complicacy and uncertainty of load forecasting,electric power load is difficult to be forecasted precisely if no analysis model and numerical value algorithm model is applied. In order to improve the precision of electric power system short term load forecasting,a new l

4、oad forecasting model is put foreword in this paper .This paper presents a short-term load forecasting method using pattern recognition which obtains input sets belong to multi-layered fed-forward neural network,and artificial neural network in which BP learning algorithm is used to train samples. L

5、oad forecasting has become one of the major areas of research in electrical engineering in recent years. The artificial neural network used in short-time load forecasting can grasp interior rule in factors and complete complex mathematic mapping. Therefore,it is world wide applied effectively for po

6、wer system short-term load forecasting. Index terms：artificial neural network (ANN)， back propagation (BP)， learning algorithm， load forecast， pattern recognition. 1. INTRODUCTION Short-term load forecasting has been useful in safe and economical planning operation of an electrical power system. It

7、has been also used in start-up and shut-down schedules of generating units,overhaul planning and load management. One of the characteristics of electric power is that it cant be stockpiled,that is,the power energy is generated,transmitted,distributed and consumed at the same time 1. In normal workin

8、g condition,system generating capacity should meet load requirement anytime. If the system generating capacity is not enough,essential measure should be taken such as adding generating units or importing some power from the neighboring network. On the other hand,if the system generating capacity is

9、of surplus,essential measure should be taken too,such as shutting-down some generating units,or outputting some power to neighboring network. Load variation trend and feature forecasting are essential for power dispatch,layout and design department of power system. Having been tried out for long tim

10、e,the load forecasting methods can be sorted to experiential qualitative forecasting and quantitative one 2. Experiential qualitative forecasting method mostly depends on the judgment from some experts. It can only make a directional idea. Artificial Neural Network and Expert System methods belong t

11、o quantitative forecasting methods 3-4. Artificial neural network,as the prototype of human brain cells,can imitate the human brain to train known information,grasp interior rule in factors and complete complex mathematic mapping 5. As is the case with time series approach,the ANN traces previous lo

12、ad patterns and predicts a load pattern using recent load data. It also can use weather information for modeling. The ANN is able to perform non-linear modeling and adaptation. It does not need assumption of any functional relationship between load and weather variables in advance 6-8. We can adapt

13、the ANN by exposing it to new data. Their ability to outperform experiential qualitative forecasting methods especially during rapidly changing weather conditions and the short time required to their development,have made ANN based load forecasting models very attractive for on line implementation i

14、n energy control centers. Therefore,it is worldwide applied effectively for the power system short-term load forecasting. Fig.1. Short-term load forecast system 2. SHORT-TERM LOAD-FORECAST (STLF) SYSTEM MODEL The short-term load forecast system is shown in Fig.1. A. Load Data Gathering The historica

15、l load data and real time load data are obtained from scan data system by computer. Straight method,median method and grey model are used to get rid of bad data. B. Input Set Choosing Using pattern recognition theory,the data that their values are highly similar to that of predicting data are chosen

16、 as parameters and Anns input sets. The parameters can be influenced by some kind of factors,such as temperature,humidity,rainfall and day types etc. Thus they should be mapped within range 0，1,according to the influencing extent to load. And then the mapping database for load sets is set up,which is the foundation of choosing similar day. C. Load Forecast The three-layered feed-forw