Research Of Bus Load Forecasting Based On Clustering Analysis

Short-term load forecasting is an important foundation for the lean scheduling of power system. Because of the current decentralized model of load management currently, the research of bus load forecasting is helpful to control the operation of power system, and develop the energy-saving power generation plan.There are the large number of bus in the system with different load curve, and it’s load is small, affected by meteorological factors easily. But the other hand, due to the fixed supply area, bus load has a relatively simple and stability structure. The current research of bus load forecasting focus on three aspects:how to improve the versatility of model, the accuracy of result, and the speed of computing.The paper analyzes the difference between system load and bus load, especially the error composition and load cardinal number of them. It is clear that the most commonly used bus load forecasting method has a problem of error accumulation, this problem make the sum of bus load forecast result is not equal to system load forecast result, and the error of former is bigger. In addition, through the analysis of volatility, found that the load regularity is proportional to it’s cardinal number. So, the paper put forward to construct a kind of middle layer between system load and bus load, called cluster load.Select the daily load curve similar as clustering feature, and proposed an improved fuzzy c-means clustering algorithm based on subtractive clustering, called SUB-FCM algorithm for bus load clustering. Then adopt comprehensive data processing strategy, take a different method for the variation of different types of buses. Finally, select least squares support vector machine (SL-SVM) model to predict the cluster load. System load forecasting result in the area is equal to the sum of cluster load forecasting result, and a bus load forecasting result in one cluster load need to be allocated by ratio model.At the last of the paper, cases are given to demonstrate the effectiveness and feasibility of the proposed method, which is possible to improv the predict results both bus load forecasting and system load forecasting.