Renewable energy sources, on one hand, are environmentally friendly, but on the other, they suffer from volatility in power generation, which endangers power-grid stability. A viable solution to circumvent the intermittent behavior of renewables is the usage of energy-storage systems. In this paper, we study the energy management of a proof-of-concept system consisting of solar panels, energy-storage systems, a power grid, and household loads. Using neural networks, we identify the most relevant parameters impacting the power generation of solar panels, and then train the corresponding network to derive forecasts. We also go one step further, and propose a heuristics-based energy-management policy for the purpose of reducing curtailments. We show that our proposed policy outperforms the naive policy by 8%, which does not consider any power-generation forecasts.