Solar power forecast is a much needed means for grid operators, particularly in residential microgrids, to manage the produced energy in a dispatchable fashion. Deterministic methods are unable to accurately forecast the intermittent solar power generation since they depend on unique sets of inputs and outputs. Therefore, stochastic methods and artificially intelligent (AI) strategies are utilized for solar power forecast. In this work, a neural network (NN)-based numerical weather prediction (NWP) model is developed for a residential microgrid in San Diego, California considering all key weather parameters such as cloud coverage, dew point, solar zenith angle, precipitation, humidity, temperature, and pressure in the year 2016. The developed weather model is then used to predict the generated power in the residential smart microgrid. To validate the accuracy of the model, the solar irradiance and generated solar power in the residential microgrid are predicted for the year 2017 using the obtained NN-based model. The results are compared with the actual solar irradiance and power in 2017 to evaluate and validate the accuracy of the developed model. Furthermore, to showcase the effectiveness of neural networks in forecasting solar power and the accuracy of the NN-based model, the results are compared with those of two other methods including multi-variable regression (MVR) and support vector machine (SVM) approaches using mean absolute percentage error (MAPE) and mean squared error (MSE) criteria.
All Science Journal Classification (ASJC) codes
- Civil and Structural Engineering
- Building and Construction
- Safety, Risk, Reliability and Quality
- Mechanics of Materials