Power quality disturbances(PQDs)are becoming the primary concern as serious issues for the emerging renewable energies to a distribution and transmission network.In this thesis,some novel combination of feature extraction and classification methods based on discrete wavelet transform(DWT),wavelet packet decomposition(WPD),multivariate singular spectral analysis(MSSA),Empirical mode decomposition(EMD),improved principal component analysis(IPCA),discrete orthogonal S-transform(DOST)with statistical parameters analysis and 1-dimensional convocational neural network(1-D-CNN),deep stacking neural network(DSNN)and Compact multiclass support vector machine(CMSVM)classifiers are investigated for the detection and classification of PQ disturbances.The first approach,the impact of statistical parameters,is proposed for optimal feature extraction and selection.Four different decomposition techniques(DWT,WPD,MSSA,and EMD)are examined for optimal selection of signal processing techniques to classify the Power Quality(PQ)disturbances.The combination of these selected features is divided into three groups,and each group is employed to examine the effect on the selected decomposition technique.These experiments are well explained that the higher-order statistical(HOS)features extracted from Multiclass singular spectral analysis and wavelet packet decomposition has better computational speed and the highest classification accuracy,respectively.At last,optimally selected features are fed to a convolution neural network(CNN)based soft-max classifier for classification of PQ disturbances.This study highlights the key properties of each technique and compares them to find out the best-suited approach for signal decomposition.WPD based method has shown a higher classification accuracy of 99.8 and 99.12 % under without noise and noisy environment,respectively.However,the MSSA based approach has less computational complexity.For the second and third approaches,the renewable energy resource is introduced to generate the PQ disturbances data and tested with a novel algorithm based on improved principal component analysis(IPCA)and 1-Dimensional Convolution Neural Network(1-D-CNN)for detection and classification of PQDs.IPCA is decomposed into four levels.The principal component(PC)is obtained by IPCA,and it contains a maximum amount of original data as compare to PCA.The statistical analysis is employed for optimal feature selection.Secondly,these improved features of the PQDs are fed to the 1-D-CNN-based classifier to gain maximum classification accuracy.Finally,the proposed IPCA-1-D-CNN algorithm has been tested with noise(50 dB to 20 dB)and a noiseless environment.The average percentage of correct classification of single and multiple PQ events is 99.92%,99.85,and 99.75% without noise,different noise levels(20 dB,50 dB),and simulated dataset,respectively.The comparative results show that the proposed method gives significantly higher classification accuracy.The third methodology,a novel algorithm is proposed based on multivariate singular spectral analysis(MSSA),wavelet packet decomposition(WPD)and 1-dimensional convolution neural network(1-D-CNN)for monitoring,mitigation,and classification of power quality disturbances(PQDs).MSSA and WPD are used to decompose the signal into four levels to extract the statistical features.The optimally extracted features are fed to a CNN based soft-max classifier to classify the signal.The proposed algorithm is tested under no noise and 20 dB to 50 dB noisy environments.WPD has higher classification accuracy(99.8% and 99.12%)as compared to MSSA(99% and 98.3%)for noiseless and noisy conditions,respectively.The results show that the proposed framework has obtained reliable highest classification accuracy.The fourth method,the complex PQ disturbances problems have been analyzed using a novel algorithm that comprised of discrete orthogonal S-transform(DOST),compressive sensing(CS),and deep stacking network(DSNN)for automatic monitoring and classification.DOST-CS based method is employed for feature extraction and reduction of power quality event data.Moreover,compressive measurements of 24 types of multiple and nine types of single PQDs events are fed to a proposed DSNN classifier for PQD recognition.The proposed method is tested with 6680 numbers of real and synthetic,single,and multiple PQDs data.The average classification accuracy of single and combined PQ disturbances for noiseless and noisy conditions is(100% and 99.88%),and(99.8% and 99.57%),respectively.The high classification results show that DOST-CS feature extraction and DSNN classifiers have high precision to recognize multiple power quality events data even in noisy conditions.The fifth approach represents a novel classification algorithm based on the multi-scale singular spectrum analysis(MSSA),Grey wolf optimizer(GWO),and compact multiclass support vector machine(CMSVM).Initially,the power quality disturbance(PQD)waveforms are decomposed based on proposed MSSA into various trends,oscillations,and noises.These decomposed oscillations and patterns are used as a feature vector.The swarm intelligence(SI)based GWO algorithm is utilized for optimal feature selection.CMSVM is employed for the classification of sixteen types of single and combined PQDs.The proposed method is tested using real and synthetic datasets with(20 dB to 50dB)and without noise levels.Moreover,a deep convolution neural network(DCNN)with MSSA is also tested and compared with the proposed algorithm in terms of classification accuracy,convergence,and computation time.The proposed algorithm is trained for 20 iterations and has the highest classification accuracy of 99.97% and 99.8% for without noise and noisy PQ disturbances.The results showed that a DCNN based algorithm has a slightly better classification accuracy,while the CMSVM base method has better convergence time and computational speed.