Solar home systems (SHSs) have become a popular solution to solve the problem of energy access in areas with no electricity or poor electricity service. However, they face the challenge of daily and seasonal variation in solar irradiance and load usage leading to suboptimal energy use. Solar forecasting techniques are used to solve this problem and are often used in larger systems such as solar farms. In larger commercial systems, solar forecasting involves predicting solar irradiance either in a hour/day-ahead fashion. This technique works well for large systems where the solar power is often exported back to the utility grid. However, with smaller solar systems such as SHSs, the solar energy is often used to charge batteries (which are then used during the night) and the solar output from the panels depends not only on the weather but also the battery utilization. In addition to this challenge, most SHSs lack computing and communication systems making deployment of optimization techniques challenging. This study proposes a smart solar home system and a forecasting methodology which accounts for battery utilization leading to more optimal energy usage. Smart solar home systems have been deployed in real homes and field data is evaluated with the proposed forecasting technique.