Non-parametric probabilistic load flow using Gaussian process learning

The load flow problem is fundamental to characterize the equilibrium behavior of a power system. Uncertain injections such as those due demand variations and intermittent renewable resources will change system’s unexpectedly, thus potentially jeopardizing reliability stability. Understanding solutions under uncertainty becomes imperative ensure seamless operation In this work, we propose non-parametric probabilistic (NP-PLF) technique based on Gaussian Process (GP) learning understand system for better operational decisions. can provide “ semi-explicit ” form by implementing testing steps that map control variables inputs. proposed NP-PLF leverages upon GP upper confidence bound (GP-UCB) sampling algorithm. salient features method are: i) applicable having injection with an unknown class distribution; ii) providing (PLB) which further provides over error convergence; iii) capable handling distributed generation well uncertainties. simulation results performed IEEE 30-bus 118-bus show learn voltage function subspace using small number training samples. Further, different input distributions indicates complete statistical information be obtained average percentage relative order 1 0 − 3 % 50,000 test points. • Novel uncertain injections. No need about type distribution. A solutions. Accurate flow. Probabilistic convergence criteria.