 |
Center for Nonlinear Studies |
Knowing the connectivity and line parameters of the underlying electric distribution network is a prerequisite for many grid operations. While most of the topology learning approaches proposed in the literature are based on the analysis of data passively recorded from customers, the key point presented in this talk is to actively probe the electrical grid by acting on controllable devices, e.g., smart inverters. Collecting and processing the incurred voltage deviations across nodes can potentially unveil the grid topology. Using grid probing data, the tasks of topology identification is posed as a non-convex problem. Leveraging the features of the Laplacian matrix of a tree graph, probing only the candidate terminal nodes is analytically shown to be sufficient for exact topology recovery. The non-convex problems are relaxed to convex surrogates, which are iteratively solved via closed-form updates based on the alternating direction method of multipliers. The power generation intermittency of renewables generators is one of the main causes of voltage fluctuations in distribution networks. For this reason, a necessary condition for a complete integration of distributed energy sources is the development of control algorithms whose specific goal is that of keeping the voltages within pre-assigned operating limits. Three strategies will be discussed. The first is purely local, meaning that each micro-generator updates the amount of power to be injected based only on local measurements of the voltages’ magnitude. The second one is distributed, namely, the micro-generators, to perform the updating steps, require some additional information coming from the neighboring agents. The local strategy is simpler to be implemented but it might fail in solving the voltage control problem. The distributed one requires the micro-generators to be endowed with communication capabilities but it is effective in driving the voltages within the admissible interval. Finally, a hybrid control strategy, inheriting the fast convergence from the local one, and the optimality of the steady state from the distributed one, is presented.
Host: Deepjyoti Deka