In the current context of the energy transition towards renewable and sustainable energy sources, solar energy is playing an increasingly crucial role, offering a clean and abundant solution to energy needs. However, to fully exploit the potential of solar energy in a microgrid, it is crucial to accurately forecast solar panel output. On the other side, the energy demand fluctuates greatly according to consumer behaviors. Anticipating these variations is essential for achieving a production-demand balance to ensure optimal control of the whole system. This work research aims to predict the photovoltaic power production and energy consumption for living labs installed in Green & Smart Building Park (GSBP) in Benguerir, Morocco. This study compares two forecasting models based on artificial neural networks (ANN), namely multilayer perceptron (MLP) and long and short-term memory (LSTM) models. Further, the results are evaluated by analyzing statistical measures, including the mean squared error (MSE) and the coefficient of determination (R2). The MLP model shows high-quality in the overall accuracy performance compared with the LSTM model.