Liu Zunxiong1，Zhang Deyun1，Sun Qindong1，Xu zheng2
(1.School of Electronics and Information Engineering,Xi'an Jiaotong University,Xi'an 710049,China;2.School of Electronics and Electric Engineering,East China Jiaotong University,Nanchang 330013,China)
Abstract:The mid-term electric load prediction is an existing difficult problem whose predicted solution often has a larger error.An RVM (relevant vector machine)based mid-term prediction method is proposed for solving the problem.With the practical data provided by EUNITEnetwork,the relations of before-after duration of daily maximum load, the relation between daily maximum load and holidays,and existing relation between the day's maximum load and the corresponding weeks are investigated and the model of the mid-term electric load prediction is constructed. In the model,the n preceding information related to certain day is regarded as daily maximum load of that day,and the relation information between the daily maximum load and holidays,number of weeks (the day ) is represented as two bi-values.Before training the model,the preceding 7 attribute values of input variables and the predictive goal value are normalized.Using different train sample sets,the simulation experiment result demonstrates that the RVM method has more advantages than support vector machine method.When the Gaussian kernel function width is 2.0,RVM method possesses more perfect prediction performance.
Keywords:electricity load;mid-term forecasting;relevant vector machine; model experiment