Machine Learning Assisted Stochastic Unit Commitment during Hurricanes with Predictable Line Outages
Abstract : Stochastic unit commitment is an efficient method for grid operation in the presence of significant uncertainties.An example is an operation during a predicted hurricane with uncertain line outages.However, the solution quality comes at the cost of substantial computational burden, which makes its adoption challenging.This paper evaluates some possible ways that machine learning can be used to reduce this computational burden.First, a set of feasibility studies is conducted. Results suggest that using machine learning as an assistant to the stochastic unit commitment solver is more advantageous than using it as a standalone solver. In particular, the machine learning model is trained to facilitate solving the problem by determining the unnecessary constraints that can be removed from the original problem without affecting the final accuracy.The variables that can be used as input features/predictors or outputs for the machine learning model are determined through feasibility studies.Then, an algorithm to train and utilize a machine learning model is proposed.The method is tested on a 500bus synthetic South Carolina system.Various test cases show an average reduction in solution time by more than 90% by using the trained machine learning model to assist the stochastic unit commitment solver.
? The method should be able to solve the SUC for a largescale realworld size network with multiple scenarios.
? The method should be able to solve within an acceptable computational time with standard hardware.
? As multiple line outages are allowed, the network topology can change over the duration of the study (a day), and the model should be able to handle this.
? As changes in the topology may lead to inevitable load shedding or overgeneration (outage of a radial line), the model should allow relaxation of nodal power balance.
? SUC is an optimization problem defined over a set of scenarios that represent realizations of the uncertain future.
? The goal of SUC is to minimize the objective function, often operation cost, subject to physical and reliability constraints of the network.
? With high levels of uncertainties, the objective function should include not only generation costs, but also penalized load shedding (unserved load) and overgeneration.
? The power transfer distribution factor is among the most efficient methods to solve the power flows in UC problems. However, when it is used for largescale networks with multiple outages, the number of variables and constraints becomes extremely large.
? There are a few studies on the modeling of uncertain equipment failures in the transmission network.
? A method to solve the securityconstrained UC for large networks with one line outage possibility is developed in .
? However, the proposed method is not valid when there are multiple line outages.
? As one application for the proposed algorithm, a preventive dayahead unit commitment model was developed for Texas 2000bus test system, affected by a hypothetical hurricane.
? Stochastic Unit Commitment (SUC) has a primary advantage of being simple to model the uncertainty explicitly, thus offering a reliable solution
? Due to strict SUC constraints, and considering the fact that it is unlikely to train an ML model to perfectly predict every output exactly the same as solving original SUC, a potential solution is to use a machine learning as an assistant to SUC to facilitate the solution process.
? This way, not only is the result accurate similar to the original SUC, but also perfect ML performance is not necessary.
? This would translate in a reduced need to training data
