Agent Technologies for Active Control Applications in the Power Grid
The introduction of remotely controlled network devices is transforming the way power systems are operated and studied. The ability to provide real and reactive power support can be achieved at the end-user level. In this activity, a framework and algorithm to coordinate that type of end-user control were developed. The algorithm is based on a layered architecture that would follow a chain of command from the top layer (transmission grid) to the bottom layer (distribution grid). In the distribution network, action can be initiated locally to find solutions to certain problems. We therefore studied decentralized optimization problems to find a solution to control reactive power resources. We found that Lagrangian relaxation algorithms showed the best results. Since capacitors are another reactive power resource to be controlled, we also developed a decentralized optimization algorithm to minimize losses in the distribution network. The decentralized algorithm results were found to be similar to those using a centralized algorithm. Finally, because the decentralized optimization algorithm needs to iterate among regions to find a solution, we introduced another algorithm to find a local solution to reactive resource problems in the distribution network. The algorithm is based on sensitivities of voltages to reactive resources to estimate the top of a feeder bus voltage of a particular region inside the distribution network. The algorithm was shown to effectively find a solution to a local problem, and the results are similar to those of a centralized optimization problem. The framework and the algorithms created in this activity integrate agent-based technologies to manage the data and control actions required to operate the type of architecture we examined.
