论文部分内容阅读
针对支持向量机在电力系统短期负荷预测中,预测模型的精度易受训练样本数据的影响,且训练时间长的问题,本文提出1种基于离散Frechet距离和支持向量机相结合的预测方法,通过建立离散曲线相似性的数学模型,找出与基准日负荷曲线形状相似的历史日负荷曲线,以相似日的负荷数据及相应的气温、星期类型等影响因素作为训练样本对支持向量机进行训练,有效地减少了训练数据量,使得训练样本更具针对性。采用East-SlovakiaPowerDistributionCompany提供的负荷数据对提出的模型进行验证,并与标准支持向量机的预测结果对比,本文的方法能够科学合理地选取相似日,提高了支持向量机短期负荷预测的精度。
Aiming at the shortcoming of support vector machine (SVM) in power system short-term load forecasting, the accuracy of prediction model is affected by training sample data and the training time is long. In this paper, a prediction method based on discrete Frechet distance and support vector machine is proposed. The mathematical model of the similarity of discrete curves was established to find out the historical daily load curve similar to that of the baseline daily load curve. The support vector machine was trained with the load data of similar days and corresponding factors such as air temperature and day of the week as training samples, Effectively reduce the amount of training data, making the training samples more targeted. The load data provided by East-Slovakia Power Distribution Company is used to verify the proposed model. Compared with the prediction results of standard support vector machines, the proposed method can select similar days scientifically and rationally and improve the accuracy of SVM short-term load forecasting.