This paper presents a model predictive control of electric power systems based on the multiple Gaussian process predictors. The Gaussian process model is a nonparametric model and the output of the model has Gaussian distribution with mean and variance. | Journal of Automation and Control Engineering Vol. 3, No. 5, October 2015 Model Predictive Control of Electric Power Systems Based on Gaussian Process Predictors Tomohiro Hachino, Hitoshi Takata, Seiji Fukushima, and Yasutaka Igarashi Kagoshima University, Kagoshima, Japan Email: {hachino, takata, fukushima, igarashi}@ Abstract—This paper presents a model predictive control of electric power systems based on the multiple Gaussian process predictors. The Gaussian process model is a nonparametric model and the output of the model has Gaussian distribution with mean and variance. The multistep ahead predictors for the phase angle in transient state of the electric power system are formed by training multiple Gaussian process models in accordance with the direct approach. Based on these predictors, model predictive control is accomplished, where the input signal is optimized so that the error between the predicted future output and the reference signal becomes small taking the uncertainty of the predicted future output into account. Simulation results for a simplified electric power system are shown to illustrate the effectiveness of the proposed model predictive control. performed by using the multiple trained GP models as every step ahead predictor. Since this direct method uses not only one-step ahead predictor but also all-step ahead predictors, the prediction errors are not accumulated so much as the prediction horizon increases. Moreover, these multiple predictors give the predictive values of the phase angle and uncertainty of the predictive values as well. In the stage of control, the input signal which is the increment of excitation voltage is optimized so that the error between the predicted future output (phase angle) and the reference signal becomes small taking the uncertainty of the predicted future output into account. The information about uncertainty of the predicted future output is used as a constraint. This means that the .
