Application Of Soft Computing Techniques In Microgrid Automation:Renewable Generation Forecasting,Demand Forecasting And Microgrid Protection

Microgrids,distributed renewable energy sources and energy efficient technologies are regarded as a key feature of modernizing the power system.Microgrids have become an attractive proposition for integrating renewable generation and smart grid features such as energy efficiency.Recent years saw a growing number of microgrids installed in different islands and remote areas.The need for higher level of reliability also provoked larger penetration of microgrids in electrified areas.However,issues such as the highly intermittent and site-specific nature of renewable energy sources,the unpredictability of load situations and added complexity in protection system of microgrids have also increased the level of uncertainty in the operation of such power systems.Soft computing techniques offer an effective solution to model and study the stochastic behavior of renewable generation,load demand and operation and protection of microgrids integrating renewable sources.The tolerance of Soft Computing techniques to uncertainty,imprecision,and approximation make them handful tool to address issues in microgrid automation.This thesis work makes use of Soft Computing tools in the form of different classes of Artificial Neural Network(ANN),Support Vector Machine(SVM),Adaptive Neuro-Fuzzy Inference System(ANFIS),Non-linear autoregressive neural networks(NAR and NARX)and Bagged Decision Trees(BDT)for generation forecasting,load forecasting and protection of microgrid.In the effort to devise stable and accurate forecasting models,the paper devises strategic input parameter selection and output processing schemes.The use of modified signal processing tools of Windowed Fast Fourier Transform(WFFT)and Wavelet Transform(WWT)is also involved for feature extraction in the developed microgrid protection scheme.An input parameter importance metric which combines correlation analysis,sets of sensitivity analysis techniques and Garson's algorithm is formulated for selection of input parameters in forecasting applications.Forecasting models for generation from PV arrays and wind turbines are developed based on soft computing tools such as ANN(Radial Basis and Conventional Feedforward NNs),SVM and ANFIS.Outputs of more than one forecasting models are ensembled using techniques of Simple Averaging,Regression and Outperformance.The load demand forecasting model devised in this study uses wavelet analysis for pre-processing of the load record is broken down into an approximate and detail components through wavelet decomposition that will be treated and forecasted independently.A unique step of determining the optimal feedback delays for each target parameters through autocorrelation analysis is adopted.A cross correlation analysis is employed to decide on the usage of external input parameters and hence choose either of the NAR or NARX models.The individual forecasts are later merged trough wavelet reconstruction to give the load forecast.The paper also addresses the issue of microgrid protection through a novel approach which incorporates soft computing based scheme with detailed analysis of current and voltage waveforms through windowed fast Fourier and wavelet transforms.The fault detection scheme involves bagged decision trees which use input features extracted from the signal processing stage and selected by correlation analysis.The proposed protection strategy is evaluated for its performance in both island and grid-connected modes of operation.The techniques for demand and generation forecasting proposed in the thesis are tested and verified based on practical data from Goldwind Smart microgrid in Beijing,China.The microgrid protection scheme is tested on a microgrid model developed using PSCAD/EMTDS,which is inspired from the same operational microgrid.All soft computing techniques are implemented in MATLAB2017b environment.The results attested applicability of the soft computing tools in forecasting of generation and demand as well as protection of microgrid.The forecasting models were seen to outperform the respective methods in the literature in terms of forecast accuracy.The devised protection scheme proved to be sensitive and selective enough to effectively identify faults from other disturbances and faults in different zones of the microgrid.The comparisons of the devised technique with conventional over-current,directional and differential protections is also presented.