Machine learning for ranking day-ahead decisions in the context of short-term operation planning

Abstract

In operation planning, probabilistic reliability assessment consists in evaluating, for various candidate planning decisions, the induced probability of meeting a reliability target and the expected operating cost over a certain future time period. In this paper, we propose to exploit Monte-Carlo simulation and machine learning to predict operation costs for various day-ahead unit commitment and economic dispatch decisions and a range of realisations of uncertain loads and renewable generations over the next day. We describe how to generate a database, how to apply supervised machine learning to it, and how to use the learnt proxies to rank candidate day-ahead decisions in terms of the expected operating cost they induce over the next day. We illustrate the approach on the IEEE-RTS96 benchmark where we use the DC power-flow approximation and the N-1 criterion to simulate real-time operation and to generate generation schedules in the day-ahead operation planning stage.

On Machine Learning-Based Techniques for Future Sustainable and Resilient Energy Systems

“…To overcome the scalability issue of the standard DRL, where the action space is proportional to the number of units and power output levels, the proposed method decomposes the collective actions into sequential Markov decision process to make fast real-time decisions. Similarly, [74] predicts the operation costs including generation redispatch cost, load shedding cost and wind power curtailment cost under the N-1 contingency through a proposed proxy. The proxy is a real-time decision making simulator with built-in MLBTs, which utilizes the Kmeans clustering and MCS methods to assess the next-day operation cost.…”

A Comprehensive Review and Comparison of the Fragility Curves Used for Resilience Assessments in Power Systems

“…Secondly, within the short-term, the approaches considering a timeframe between a day and a few hours before the event are widely known as preventive and preparative [18,21,27,36,39,40,[54][55][56]. For instance, references [21,36,56] have proposed day-ahead assessments to commit generation units and re-structure the grid considering extreme weather events. In line with these three studies, in [55], Lei et al proposed to…”

Graph Convolutional Networks for probabilistic power system operational planning

“…Alternative methods to [4] for probabilistic operational planning have been presented in the literature [5]- [9]. In [5] a DC power flow was used to include power system response in a probabilistic operational planning model.…”

“…An AC power flow and a linear approximation of frequency response were used in [7] to include frequency response in an operational planning model. More recent approaches use machine learning for generating proxy models of real-time operation to speed up probabilistic operational planning [8], [9]. In these papers, a machine learning model is trained to act as a DC-security constrained optimal power flow (SCOPF) and to predict the optimal corrective actions given a set of preventive actions.…”