Research On Load Forecasting Colloboration Based On Maximum Entropy Principle

Urban power network planning is an extremely complex task. The traditional single planning approach is no longer suited to the fast development at present, demanding of multi-department collaboration of the grid company. As an important basis of power network planning, load forecasting has a number of typical collaboration problems. This paper carried out the following studies about load forecasting collaboration.Firstly, a collaborative load forecasting model based on maximum entropy principle is proposed. Taking the expectation and second order central moment of forecasting results into account and using the1 maximum entropy principle, the new model receives the probability distribution function of forecasting values and obtains the only final high, middle and low forecasting scheme and reflects adaptability of actual network planning results. Case studies show that this model has a broad applicability. Different original data can be fused by this model and only one forecasting results is obtained. This model is able to adapt to the development of load forecasting and achieves collaboration of every aspects.Secondly, a practical multi-path load forecasting method is given based on maximum entropy principle collaboration load forecasting model according to the actual situation for power networking planning. For load forecasting has a characteristic of layering and zoning in multi-path, this process is quiet complex. As a collaboration of hierarchical partition problem, multi-path forecasting is defined in this paper and method of multi-path collaboration is proposed. In case of multi-path of data, this forecasting method makes each two sets of data integration until each group is used more than once. The example gives fusion method in this situation and shows the result that different permutation and combination has different entropy. The low-frequency data has higher entropy. So the final forecasting scheme can be chosen according to the entropy in the context of actual power network operation.Finally, based on maximum entropy principle, this paper proposed a method to solve data conflict in each step of the whole process of load forecasting and make great progress in load forecasting precision. How to deal with this data problem is important to load forecasting results. Entropy principle, as an effective information processing method, is used to solve some local problems in load forecasting. This paper analysis and abstracts uncertainty information in the whole process of load forecasting and sums up information fusion and filtering method based on entropy principle, which applied in the whole process of load forecasting. Case studies have shown that information entropy method has a good application in preprocessing, model evaluation comprehensive and multi-path decisions.