Implementing a real-time control algorithm of Triana on SASensor Open Platform

In terms of electricity generation, Distributed Generation (DG) using Photovoltaics (PV) panels has become more common. To deal with the potential challenges it brought on the existing electricity distribution network, Demand Side Management (DSM) methodologies are developed. Triana, developed by Computer Architecture and Embedded Systems group (CAES) of University of Twente, is one of them. It is divided into three steps. In the first two control steps, it is able to utilize simulation or measurement data off-line to forecast and plan electricity production/ consumption on the device level. Since prediction and planning errors can be caused by fortuitous behaviors of end users, it is often impossible to foresee it and prevent it from happening. It is however possible to detect these errors and solve them using a real-time control algorithm in the third step of Triana. The design, implementation and testing of this real-time control algorithm are included in this thesis. In cooperation with Locamation, an Application Programming Interface (API) is added to the Triana to detect prediction and planning errors. Power, voltage and current data of medium voltage (MV) to low voltage (LV) transformers can be gathered in real time from the SASensor open platform. From the SASensor open platform, Triana can access the real-time measurement data (RTD) to detect overloading problems, deviation from original plan, etc. Upon detection of intolerable prediction/ planning errors, the real-time control algorithm tries to perform replanning based on the deviation (measured in real-time) from its original plan. This new plan aims to quickly compensate for this deviation, preferably without violating other requirements (such as electric vehicle (EV) charging deadlines). The performance of the implemented real-time control algorithm is tested in simulation environment with data generated from a profile generator and the Triana network model of the test site of Lochem. Result shows that in most cases, the designed algorithm is capable of reducing the deviation from original planned power profile (sometimes at the cost of extending EV charging deadlines). In realistic cases (Case 3, Case 4 and Case 5), the real-time control algorithm tends to offer satisfactory improvement even compared with what is gained by the optimal solution (Table 4.6). It also helps to reduce overall power consumption upon detection of overloading problem.