该文使用支持向量机中的两种核函数，采用grid-search算法、遗传算法、粒子群算法优化参数，建立对吉林市某小区燃气管网日负荷预测的支持向量机模型。将日最高温度、日最低温度、日平均温度、小区人员最高年龄、小区人员最低年龄、小区人员平均年龄作为燃气管网日负荷变化密切相关的主要影响因素，分别作为支持向量机的输入量，将小区人员临时出差、小区临时增加暂住人口等随机因素作为燃气管网日负荷变化密切相关的次要影响因素，将随机因素统一归为支持向量机的一个输入量。采用[0,1]归一化方法，对作为影响因素的输入量数据与日负荷预测输出量数据进行归一化处理。对节假日和工作日的燃气管网日负荷预测采用独立处理方法，避免了相互之间的干扰影响。试验结果表明，采用径向基核函数的支持向量机预测模型对燃气管网日负荷预测拟合程度达到90%以上。%This paper constructs daily load prediction model for community gas network based support vector machine in Jilin, which is used of two kernel function and grid-search algorithm, genetic algorithm and particle swarm optimization algorithm. lnputs of support vector machine include daily maximum temperature, daily minimum temperature, daily mean temperature, maximum age of staff, minimum age of staff, average age of staff, which are the main factors affecting the daily load of gas network. Secondary factors include some random factors such as temporary travel of staff, temporary increase of staff and so on. The paper constructs random factors as an input of support vector machine. Output of support vector machine is daily load prediction of community gas network. Normalization method of [0,1] is used to normalize inputs and output of support vector machine. The independent processing method on holidays and days of gas network daily load prediction avoids mutual interference effects. The result shows the fitting degree is more than 90% on prediction models which are constructed of support vector machine based on radial basis function kernel function.
展开▼
