Table 1: Big Data Applications in Smart Grids – Methods and Case Studies.

Application	Ref. #	Method(s)	Case Studies
Renewable energy	‎ [28]	The means of communications through long distance or remote stations using energy efficient cellular communication networks.	Off-grid or standalone base stations powered by local small-scale renewables to not require grid power for communication.
	‎ [29]	Multiple models for current, future, and virtual energy markets used to optimize PV integration into a micro grid.	A 65 solar panel array with 15 kWH energy storage is simulated. The system operation is evaluated without any energy sales, with sales restricted to local users, and sales to both local users and the grid.
	‎ [31]	An enhanced K-means algorithm, named Time Series Clustering (T.S.C) K-means, combined with Multilayer Perceptron Neural Networks (MLPNN) for solar radiation forecasting.	Several meteorological time-series datasets are used to assess the performance of the proposed T.S.C K-means clustering method and its comparison with other clustering techniques including K-means*, K-means++, K-means, self-organizing map (SOM), fuzzy C-means (FCM), and K-Medoids. Solar radiation datasets from different US states are used to evaluate the accuracy performance of the developed hybrid forecasting method and its comparison with state-of-the-art forecasting techniques.
	[32]	A novel time-series based K-means clustering method, named T.S.B K-means, combined with discrete Wavelet Transform (DWT), Harmonic Analysis Time Series (HANTS), and MLPNN for wind power forecasting.	Wind speed, wind power, wind direction, and air temperature data from National Renewable Energy Laboratory (NREL) are used to evaluate the novel clustering and hybrid forecasting methods. A comparative analysis of the proposed hybrid method with other well-established forecasting models including Persistence, New Reference (NR), Adaptive Wavelet Neural Network (AWNN), and Phase Space Reconstruction (PSR) are also performed.
	‎ [33]	A Transformation-based K-means algorithm, named TB K-means, combined with MLPNN for solar radiation forecasting.	Several different datasets are used to evaluate the proposed TB K-means clustering and compare it with different variants of K-means algorithm. Solar radiation time series with different characteristics are used to provide a comparative analysis between the proposed hybrid forecasting and benchmark forecasting models.
	‎ [34]	A novel Game Theoretic Self-organizing Map (GTSOM), combined with Neural gas (NG) and Competitive Hebbian Learning (CHL), DWT and Bayesian Neural Network (BNN) for solar radiation forecasting.	Historical solar radiation data are used to assess the performance of the hybrid forecasting with the proposed GTSOM and other clustering methods.
Demand response	‎ [39], ‎ [40]	An extended framework of the Stackelberg game model for demand response optimization.	Homogeneous and heterogeneous generation supply quantities, generator profit and consumer welfare are evaluated in scenarios with few and many generation units and a large consumer population.
Electric vehicle	‎ [49]	Method of defining a more accurate model of electric consumption by light duty Plug-in Electric Vehicles (PEVs).	Uncontrolled home charging of EVs and uncontrolled “opportunistic” charging at public locations are simulated based on travel survey data.
	‎ [51]	A fuzzy expert method for online management of EVs’ charging demand.	An IEEE 38 bus distribution test feeder including charging stations at 4 nodes is simulated. .Different charging solutions/scenarios are implemented on the test system and compared.
	‎ [52]	A sliding horizon-based method for real-time data management and optimal coordination of EV charging with photovoltaic (PV) generation.	A 33 bus system including DG units and EV charging stations is simulated. EV charging coordination and its effect on PV power curtailment is evaluated.
	[55]	A hybrid of Auto Regressive Moving Average (ARMA), Fuzzy C-Means (FCM) clustering, Monte Carlo Simulation (MCS), and Particle Swarm Optimization (PSO) methods for optimal scheduling of EVs to increase the use of PV power for EV charging while providing economic revenues for EVs’ participation in V2G services.	A 12 MW PV system with 424 EVs is simulated. A collaborative strategy is developed between the EV aggregators and PV producers to minimize the penalty cost of PV over/under-production by charging the EVs using the PV power in excess of the scheduled output and discharging the V2G power to compensate the PV power under-production. The system performance with and without EV optimal charging/discharging are evaluated and compared.
	‎ [56]	A hybrid of ARMA, FCM clustering, MCS, and Genetic Algorithm (GA) methods for optimal scheduling of EVs to increase the use of wind power for EV charging while providing economic revenues for EVs’ participation in V2G services.	A 10 MW wind system with 484 EVs is simulated. A bilateral contract is developed between the EV aggregators and wind producers to use the extra wind power for EV charging and to discharge the V2G power during the periods of wind power deficits. The system performance with and without EV optimal charging/discharging are evaluated and compared.