Iterative bandwidth optimization for Kernel density estimation of renewable generation uncertainty

With the rapid increase in the penetration of renewable energy, the uncertainty of renewable energy has brought huge challenges to the safe and stable operation of the power system. Kernel density estimation is widely used in modeling the probability density function of renewable energy active output, but needs bandwidth optimization. This paper deduces the integral function of optimal bandwidth with Gaussian function as the kernel function, based on minimizing Mean Integrated Squared Error of KDE. The simplification method is proposed to convert the integral function calculation into query a data table. An accuracy examination method is also given to check the impact of simplification error on the iterative result. By comparing with rule of thumb, cross validation and conventional iteration method, the proposed method is verified to have faster computational speed and better bandwidth performance.