Short-term Load Forecast Of Mountainous Area Contain Small Hydro-power Unit

Short-term load forecasting is an important part of distributed clean energy, which is due to low investment, short cycle, easy maintenance, price stability, etc. is now being widely used. Abundant water resources in some areas, small hydropower generating capacity is more abundant, not only to meet local compliance, also incorporate a large grid. However, as more and more small hydropower installed capacity, its power system, especially on the local power grid has increased year by year. Because of its poor performance adjustment, power generation wet period and dry season changes the difference is great, the instability caused by the high power make it difficult to manage. Therefore, the study of the corresponding load forecasting model has a very important practical significance according to the local water and electricity in line characteristics.This paper introduces the whole society electricity load factors as well as small hydro power load, and through various factors discussed in detail in the comparative analysis, by the average annual load data and the corresponding standard deviation, coefficient of variation analysis, We were forecast for the next load of data processing, as well as the relevant factors to consider selecting a prediction model of the foundation. Subsequently, the self-organizing feature map feature clustering to achieve load curve, in line with historical sample data for cluster analysis and identified four decomposition effects than two exploded make data more realistic. Finally, a load forecasting model, which uses a combination of forecasting methods for small hydro power load forecasting using neural network to predict the load on society, then the predicted value obtained by adding the two final full-load electricity network and in 2015 as the sample data Liu’an City, Anhui Province, by the methods described in this article, use MATLAB software programming load forecasting their day’s data, the predicted value of the error is relatively small, which proves this method feasibility.