renewable, rooftop unit control, PV integration
As the United States sees the continued expansion of photovoltaic (PV) and other distributed solar generation technologies into the distribution grid, there is an increased need to find approaches to mitigate integration challenges associated with renewable resources. Depending on the renewable resource, the integration challenges will vary. Much of the challenge with integration is associated with the uncontrolled oscillations of output power, for example, from a PV array. Both solar and wind resources rely on environmental conditions to produce power. However compared to wind, solar generation resources such as PV typically produce more second to minute oscillations due to cloud patterns. With low levels of penetration, the impact is minimal. This paper focuses on developing advanced control strategies for building equipment like the rooftop units along with energy storage technologies to support seamless PV integration into buildings. A forecasting approach for PV is presented along with model-based control strategies for using load to support the integration of PV. The forecasting model takes as input solar irradiance and module temperature to estimate the output power of PV based on an interconnected voltage. The first step is to poll the cloud patterns for the day and utilize this information to project the cloud density each hour. The trained neural network defines relationship of this cloud cover to the amount of expected solar irradiance that is measured. Temperature data is collected from weather application and is inserted as an initial temperature to the PV model and thermal model. The model develops the corresponding PV curves based on the current module temperature reading and the solar irradiance data provided. The model predicts the average power output of the PV array over the next one-hour time window. A control algorithm for the rooftop unit is presented that utilizes this PV forecast to optimize the energy consumption to match the PV peak generation. The model is validated using irradiance, temperature, and PV output power measurements from Oak Ridge National Laboratory’s 50kW PV array.