基于马尔科夫链的自适应储能需求功率预测模型

Adaptive Power Demand Prediction Model of Energy Storage Based on Markov Chain

针对储能辅助火电机组二次调频时储能需求功率随机性强的问题,提出一种基于马尔科夫链的自适应储能需求功率预测模型。首先,针对火电机组响应自动发电控制(AGC)指令时功率随机变化性强且难以实时监测的特点,运用马尔科夫随机过程理论来描述储能未来有限时域内的需求功率,并采用后验功率信息实时自适应调整预测模型以适应AGC指令的周期性波动;然后,针对需求功率随机场景繁多的问题,提出一种可变预测时域的场景树生成方法来选择预测场景,该方法能够在树节点数一定的情况下更有效地选择场景;最后,进行了算例分析。结果表明,相比无自适应调整的马尔科夫模型,所提自适应预测模型的预测精度提高了8.28%;采用该文所提场景树方法的预测精度相对于固定场景树结构方法提高了6.67%,较极大似然估计法提高了4.65%。

Due to the uncertainty of power demand of energy storage system(ESS)when ESS participates in automatic generation control(AGC)with thermal generators,an adaptive ESS power demand prediction model based on Markov chain is proposed.Firstly,according to the uncertainty of the output power of thermal generators in response to the AGC command,the Markov chain is used to model the ESS power demand in prediction horizon,and a posteriori information is used to adapt to the fluctuations of the AGC command.Secondly,to reasonably select random scenarios of power demand,a scenario tree generation approach with variable prediction horizon is presented.The approach can select scenarios more effectively when the number of nodes is fixed.A simulation was implemented to validate the effectiveness of the prediction model.The results show that compared with the Markov model without adaptive adjustment,the presented adaptive prediction model can improve the prediction accuracy by 8.28%.The prediction accuracy of the presented scenario tree approach is improved by 6.67%compared with the fixed scenario tree structure method,and 4.65%higher than the maximum likelihood estimate method.