Similarity Clustering and Combination Load Forecasting Techniques Considering the Meteorological Factors

A similar historical load data search technique, which took the sum of meteorological load and secular trend load as clustering center, was put forward. This method can improve the similarity of the loads between forecast day and sample days, and so as to improve the reliability and precision of load forecast. Manifold load forecast methods assembled by optimal weight were also applied. Technical application manifest these techniques can represent the load character of different areas, types and weather sensitivities, so as to have robust adaptability, and then have higher precision even to those small, big amplitude of vibration and weather-sensitive load. Key-Words: Load forecast, Meteorological factors, Linear regression, Time series, Gray model, Neural network, Combination forecast