Novel Forecasting and Scheduling for Microgrid Energy Management System

Abstract

The high penetration of renewable energy resources brought new challenges to the modern grid; therefore, new solutions and concepts need to be developed. The idea of a microgrid (MG) has been introduced to overcome the upcoming issues in modern grids. MG is a small-scale grid composed of renewable energy resources, energy storage, and load demand. MG makes decisions by itself and can operate in grid-connected or islanded mode depending on functionality. The microgrid energy management system (M-EMS) is the decision-making centre of MG. An M-EMS is composed of four modules which are known as forecasting, scheduling, data acquisition, and human-machine interface. However, the forecasting and scheduling modules are considered as the major modules among the four of them. The forecasting module is required in the M-EMS to predict the future power generation and consumption. The forecast data is the input to the scheduling module of M-EMS. Employing forecasting system in the M-EMS would increase the accuracy of the scheduling module. The scheduling module is responsible for controlling the power flow from/to the main grid. Additionally, it performs optimal day-ahead scheduling of available power generation resources to feed the load demand in a grid-connected MG for economical operation. Consequently, this research work presents four contributions in the area of M-EMS for grid-connected MG. The first contribution of this research is to presents a hybrid strategy for short-term forecasting of load demand in M-EMS, which is a combination of best-basis stationary wavelet packet transform and the Harris hawks algorithm-based feedforward neural network. The Harris hawks algorithm is applied to the feedforward neural network as an alternative learning algorithm to optimized the weights and biases of neurons. The proposed model is applied for load demand prediction of the Queensland electric market and compared with existing competitive models. The simulation results prove the effectiveness of the proposed method. The second contribution of this research is to design and proposed an ensemble forecasting strategy for solar PV power forecasting. The proposed ensemble strategy is based on a systematic combination of the tunicate swarm algorithm (TSA)-based multilayer perceptron neural network (TSA-MLPNN), TSA based least-square support vector machine (TSA-LSSVM), whales optimization algorithm (WOA) based MLPNN (WOAMLPNN), and WOA based LSSVM (WOA-LSSVM). The output of all the models is combined using the Bayesian model averaging method. The proposed ensemble strategy is validated through simulation of the real-time data of building N-78 Griffith University, Queensland. The simulation results demonstrated that the proposed strategy shows excellent performance in comparison with several existing competitive approaches. The third contribution of this research is to propose an optimum scheduling strategy, using a weighted salp swarm algorithm for M-EMS, to perform the optimal scheduling of available power resources to meet consumer demand and minimize the operating cost of grid-connected MG. The performance of the proposed scheduling strategy is validated through simulation using MATLAB and compared with standard particle swarm optimization (PSO) based scheduling strategy. The comparison shows that the proposed strategy outperforms the PSO based strategy. The final contribution of this research is to propose an M-EMS using an ensemble forecasting strategy and grey wolf optimization (GWO). In the proposed M-EMS, an ensemble forecasting strategy is used to accomplish short-term forecasting of PV power and load while the GWO is applied to perform the optimum scheduling of available power resources in grid-connected MG. A small-scale experiment is conducted using Raspberry Pi 3 B+ via python programming language to validate the effectiveness of the proposed M-EMS. The experimental results of the proposed M-EMS for the selected case prove the effectiveness of the proposed M-EMS. In summary, several forecasting and scheduling strategies have been proposed and validated for the M-EMS of a grid-connected MG.