Solar electric generation is the fastest-growing and lowest-cost form of electric generation today. Since solar power generation is variable, nonlinear, and unpredictable, it is posing technical and economic challenges to both grid operators and energy traders. Grid operators are concerned about voltage violation, reverse power flow, and penalty fee due to overproduction or underproduction of solar power. Because most electricity is traded in the day-ahead market, energy traders are interested in long – term power forecasting, specifically day – ahead forecasting, and power system operators are interested in short – term power forecasting: the higher the forecasting accuracy, the higher the profit to energy traders and the lower the cost to customers. Due to the easy availability of historical solar power generation and associated weather data, solar power forecasting using a machine learning (ML) technique is becoming an attractive option. There are different ML techniques. Power system operators must choose suitable ML techniques for the right forecasting horizon. This thesis compares relevant ML techniques: support vector regression (SVR), kernel ridge regression (KRR), least absolute shrinkage and selection operator (LASSO), and ridge regression (RR); also included in the comparison is one time-series technique: autoregressive moving average (ARMA). The comparisons are for different forecasting horizons in terms of R2_Score, root mean squared error (RMSE), and mean absolute error (MAE). Results show that the kernelized machine learning techniques (SVR and KRR) outperformed other techniques.