Neural Network Based Power Forecasting And Control Of Intermittent Solar And Wind Energy Resources

Renewable energy sources(RES)seem to become essential in the electricity generating industry,owing to their capacity to mitigate global warming,reduce pollution produced by fossil fuel-based generation,and diversify the energy mix to ensure energy security and sustainability.Wind and solar energy have been widely recommended as the most effective RES solutions for reducing greenhouse gas emissions among the many RES technologies.Advancements in renewable energy technology have resulted in the establishment of more efficient PV solar panels and wind energy systems,as well as a considerable reduction in their cost.Solar radiation,cloud cover,rainfall,wind speed,wind direction,and temperature all impact the amount of energy created by these renewable energy sources.The large-scale integration of these sources into the power system is badly impacted by this fluctuation.For successful integration of solar and wind power into the electrical system,accurate forecasting of power provided by renewable energy sources is required.The goal of this thesis is to look at machine learning approaches for predicting generated electricity correctly so that these renewable energy sources may be better used.This dissertation contains two studies on renewable energy systems,one of which deals with forecasting and the other with voltage regulation in an islanded PV system.The following is a quick overview.Firstly,a novel prediction error-based power forecasting(PEBF)method for a Photovoltaic(PV)system,using Photovoltaics for Utility Scale Applications(PVUSA)model based grey box neural network(GBNN)is presented.First,the differential equation based PVUSA model is transformed into a neural network.In the proposed PEBF scheme,the neural network is set to train whenever the difference between predicted and output powers increases from a certain threshold defined based on system dynamics and requirements.The unique design of the PVUSA model based grey box neural network takes far less training time than usual black-box neural network based models.This gives the proposed prediction scheme an advantage of updating the prediction model parameters from frequent training of neural networks with the change in metrological variables.The results demonstrate that the proposed scheme predicts the PV power efficiently within the defined error tolerance level,which shows the effectiveness and feasibility of the proposed prediction scheme.The prediction accuracy of the proposed scheme has been compared with the conventional black box neural network models and reveals outperformed performance with respect to prediction accuracy improvement.Secondly,an improved Radial bases function neural network model is presented for the PV power forecasting.The Radial bases function neural network(RBFNN)has been used for few decades for PV power prediction.Various optimization algorithms have been proposed to train the linear and nonlinear parameters of RBFNN.But these algorithms suffer from local minima and slow convergence problems,making them inefficient for real-time applications.Recently a randomness operator based particle swarm optimization(ROPSO)is proposed that can handle local minima problem.In this model the ROPSO is mathematically merged with RBFNN in such a way that it also solves the slow convergence problem of RBFNN,by optimizing its search space.This improved RBFNN is applied for short term forecasting of PV System output power.Different assessment measures,such as root mean square error(RMSE)and mean absolute error(MAE),are used to analyze the effectiveness of the developed scheme.The numerical and simulation results show that the proposed scheme accurately forecasts PV power.These proposed methods are implemented on a real case study of a 20MW grid-connected PV system in Dongying city,Shandong province,China.Thirdly a power forecasting scheme for wind energy is presented.This study proposed a hybrid forecasting model combining wavelet transform,Randomness operator based particle swarm optimization and NARMAX(Hybrid WT-ROPSO-NARMAX)for power forecasting of a real wind plant based on PEBF method.The model is produced by combining interactions from the wind system's Supervisory Control and Data Acquisition(SCADA)real power record with Numerical Weather Prediction(NWP)meteorological data over one year using a 1 5-minute timestep.The wavelet is used in the proposed model to have a significant influence on ill-behaved meteorological and SCADA data,and NARRMAX methods are used to better map the NWP meteorological variables and SCADA solar power nonlinear connection.The ROPSO is used to adjust the NARMAX's settings to improve forecasting accuracy.In the proposed scheme,the Hybrid WT-ROPSO-NARMAX model is set to train whenever the difference between predicted and output powers increases from a certain threshold defined based on system dynamics and requirements.The simulation and numerical results have shown that the proposed model can accurately predict the wind output power.Lastly,as an approach to control the voltage and frequency,a nonlinear integral back-stepping controller for the voltage source inverter used in an islanded microgrid is developed.First,the dynamical model of the inverter-based distribution generations(DGs)in microgrid system is developed.Subsequently,the model based controller for the microgrid is built using dynamics of inverter based DGs and Lyapunov theory,which could eliminate the voltage and frequency deviations in the system under different uncertainties.To ensure the system stability,a control Lyapunov function is adopted.Considering the influence of irradiations and other meteorological variables fluctuations a battery energy storage(BESS)is applied on the DC side to suppress the fluctuations of output power of DGs.Furthermore,the efficiency of the designed controller was validated through simulations in the MATLAB/Simulink.environment under different scenarios and effectiveness of the proposed framework is further validated by real-time hardware in loop(HIL)experiments.In addition,the performance of the proposed controller is compared with conventional Back-stepping(BS)controller.The comparison results demonstrate that the efficiency of the designed controller in terms of obtaining steady-state operating conditions is better than that of the BS controller.