Study Of Power Load Forecasting Assessment And Its Application

Power load forecasting is an important part of the operation of electric powersystem, and plays a key role in its planning, designing, security, reliability, economyof operation and market trade. Under the special background of saving energy andreducing emissions, power enterprises put forward higher requirement for powerload forecasting precision. In consideration of the randomness and particularity ofpower load, power load forecasting models or analysis of load forecasting results arethe effective ways of improving precision of load forecasting.The prior assessment, the post-fact assessment, and the combination of the twoabove will be explored respectively in this thesis.A prior assessment algorithm based on meta-learning is researched. Thisalgorithm introduces the idea of meta-learning to the prior assessment, aiming atusing the forecasted values on the basis of feature attributes of load sequence andsingle models as meta-learning to evaluate reasonably the single models before loadforecasting, from which the results are applied to load forecasting. The analyticalresults show that the forecasting precision of this algorithm is higher, having morepractical value.The assessment of relation between total transformation and local transformati-on of power load based on virtual prediction, the load property assessment based onK diagram, the post-fact assessment of probability forecasting and fuzzy clusteringanalysis of power load are studied. The quantitative conclusions from these fouralgorithms offer the load forecasting precision useful references.This paper discusses the combined forecasting model integrating the priorassessment and the post-fact assessment, considering the combined forecasting. Thismodel takes advantage of the prior assessment and the post-fact assessment to form afore-and-aft combined model. The results from this model indicate that itsforecasting precision is higher than the prior assessment and common combinedmodels and ensure the adaptivity and robustness of load forecasting.On the basis of researching the prior assessment and post-fact assessment ofload forecasting systematically, this thesis develops an algorithms library for loadforecasting assessment. This library has been operating in Hunan power grid and theoperational results manifest that the algorithm meets the demand of the load development in Human electric power grid and have certain directing meaningstoward the load forecasting of Hunan electricity network.