|dc.description.abstract||In recent years, more distributed energy resources (DERs) such as solar PV, wind and more are being installed to compensate for the growing energy demand. It is vital to properly monitor and integrate these new energy resources. An Energy Management System (EMS) can be used to achieve this objective. This thesis is focuses on an EMS an infrastructure that can be used for a residential solar PV system. An EMS should be manufacturer independent, affordable, and easy to integrate.
This thesis presents an EMS consisting of a Monitoring and Control Platform (MCP), a data management infrastructure, and a client application. The MCP includes a Smart Energy Controller (SEC) and Motes that are placed at the output of the solar PV system, the input to the battery, and the input to the load. The user can send commands to the SEC via the web application. Then, the SEC sends commands to the Motes via Zigbee. The Motes can block or allow the flow of power using a power relay. Thus, the Motes allow precise control over the flow of power. The Motes also contain sensors that can measure the power. The measurements are sent to the SEC via Zigbee. Commands and alarms are logged by SEC. The system uses a neural network to predict the power production of a solar PV energy system. The SEC uploads the power measurements, the predicted power production, and system logs to a database where it is stored.
The client application allows users to access the Information stored in the database. The client application displays the measured power from each Mote, the predicted power production, alarms, and system control logs. The EMS was tested on a solar PV system with a single panel, a 12 V battery, and a DC load. It demonstrates the potential for low cost and independent energy management systems.
The thesis hasIed an operational Iardware infrastructure for an EMS thIt can be integrated into the existing solar PV systems. The EMS infrastructure is manufacturer independent, affordable, and easy to integrate. These features will help with faster integration of solar PV energy systems into a Smart Energy Network. Furthermore, it provides a platform that can host different energy scheduling and energy prediction algorithms. The functionality has been demonstrated through integration into a solar cart. Furthermore, a software and hardware architecture consisting of a sensor data collection system, control system, and a neural network has been successfully developed. The performance of these components was evaluated through multiple experiments. The sensor data collection system and control system were each tested separately and then tested concurrently. For the sensor data collection, 500 sensor samples were collected and analyzed to demonstrate the reliability of the sensor collection system. The testing results indicate that the system is able to consistently and accurately collect the data. The performance of the remote energy flow control has also been tested. 500 commands were sent from the web application. The response times from the system were evaluated. The results show that the system executes the command and responds within 0.9 to 1.4 seconds. The neural network was trained and tested on actual data collected from a 250-kW solar PV installation in Gaithersburg, Maryland, USA. The mean absolute error of the power production forecasts was 6 kW or 2.4%. Furthermore, the cost of the presented system is estimated to be almost 18 times cheaper compared to existing energy management systems. This thesis provides valuable information for research and development of future energy management systems.||en