To solve the problem of difficulty in evaluating demand response resources caused by the uncertainty of photovoltaic(PV)output fluctuations and the superposition of user load uncertainty under the integrated layout of source load,this paper proposes a load-side resource response potential estimation method that takes into account the coupling between distributed PV generation and load fluctuation.Firstly,based on a rational prediction model for PV generators,the low frequency components of the PV power generation curve are separated from the user net load curve.Secondly,the separated PV output and user original load data are taken as inputs to construct a load response resource evaluation model based on time-domain convolution and upper and lower interval estimation,achieving reliable interval evaluation for load response capacity.Finally,a simulation analysis case is constructed based on real distributed PV and customer load data.The results show that the proposed method reduces the difficulty of learning load characteristics by means of the PV-load separation method compared with other conventional prediction methods.At the same time,it constructs multi temporal convolution branches with different dilation rates,effectively solving the problem of inconsistent user electricity consumption cycles in the region.The prediction results have been significantly improved in indicators such as interval coverage,improving the accuracy and reliability of load response capacity evaluation.
关键词
分布式表后光伏/负荷侧资源响应潜力感知/时域卷积/上下限区间估计
Key words
distributed behind-the-meter/load response resource assessment/temporal convolutional network/lower upper bond estimation
维普
万方数据