基于ν-支持向量机的边际电价预测及置信区间估计

System marginal price prediction and confidence interval estimation with ν-support vector machine

引入ν-支持向量机,通过构造和求解一个凸优化问题,同时实现了对边际电价的预测和对置信区间的估计,且无需假定预测偏差的概率分布.在ν-支持向量回归中,当ε>0时,ν是错误样本的个数占总样本个数份额的上界.利用该性质,边际电价预测的置信度和置信区间可以很自然地用参数1-ν和变量ε来表示,这为发电公司进行竞价方案的风险分析打下了很好的基础.算例仿真表明,该方法具有较好的泛化性能和较高的预测精度.

u-support vector machine is employed to achieve system marginal price prediction and confidence interval estimation simultaneously by constructing and solving a convex optimization problem. Hypothesis for prediction-error distribution is unnecessary in this method. In v-support vector regression, v is an upper bound on the fraction of errors if the resulting ε is greater than zero, therefore prediction confidence level and confidence interval can be expressed with the parameter 1 - v and variable 8 naturally, which will play important role in risk assessment of bidding strategy. Simulation results demonstrate that this method has better generalization performance and prediction accuracy.