Abstract
The large-scale integration of renewable energy sources (RESs) has become one of the most challenging topics in smart grids. Indeed, such an integration has been causing significant grid stability issues (voltage and frequency control) due to the dependency of RESs on meteorological conditions. To this end, their integration must be accompanied by alternative sources of energy to attenuate the power fluctuations. Energy storage systems (ESSs) can provide such flexibility by mitigating local peaks/drops in load demands/renewable power generation. Therefore, the development of energy management strategies (EMSs) has been attracting considerable attention in the management of the power generated from the RESs associated with that which is stored/provided by the ESSs. Then, the optimization of the EMS leads to substantial savings in operation and maintenance and to correct decisions for the future. This study presents an optimized EMS for a wind farm, coupled with a pumped hydro energy system (PHES). The proposed day-ahead EMS consists of two stages, namely the forecasting and the optimization stages. The forecasting module is responsible for predicting the wind power generation and load demand. A random forest (RF) method is used to perform the power forecasting after the extraction of the weather data features using a kernel principal component analysis (KPCA) technique. Then, a nonlinear programming (NLP)-based optimization technique is proposed to define the day-ahead optimal energy of the PHES. The purpose of the optimization is to maximize the profit cost in a day-ahead horizon, taking into consideration the system constraints.