Uncertainty analysis, modelling and prognosis for power system operation and planning

Abstract

The introduction of smart grid, renewable energy, and electricity market, etc., has led to tremendous changes to modern power systems. Particularly, as one of the most important renewable energy, wind power has experienced dramatic growth worldwide. However, due to the intermittency, wind power generation also brings challenges to many aspects of power system operation and planning. Wind power is directly related to wind speed. Proper probability model for describing the stochastic wind speed would be essential for wind farm planning, power system analysis, and so forth. Traditional Weibull and Rayleigh distributions cannot accurately capture the wind speed properties that exhibit significant non-stationarities and complexities. In this thesis, an advanced generalized Lambda distribution is proposed to statistically model wind speed that can achieve a superior performance. Considering the short-term operational planning horizon, such as from 1 hour to 48 hours, accurate forecasting of wind power generation becomes crucial to ensure secure and reliable management of power system operation. Traditionally, research efforts are paid to developing point prediction methodologies of wind power, which corresponds to the exact mathematical expectation of the stochastic wind power series at a given prediction horizon. Because of the chaotic nature of weather systems, wind power prediction errors cannot be avoided and can be significant in some conditions. Therefore, transformation from traditional point forecasts to probabilistic interval forecasts can be of great importance to quantify the uncertainties of future forecasts, thus effectively supporting the decision making activities against uncertainties and risks ahead. To this end, this thesis has proposed novel probabilistic forecasting methodologies to quantify the uncertainty involved in wind power forecasting. Theoretical background of probabilistic forecasting and the state-of-the-art approaches for wind power prediction are thoroughly reviewed. Subsequently, a parametric probabilistic forecasting approach, bootstrap-based extreme learning machine, is developed to generate prediction intervals, which can be trained at an extremely fast speed. In addition, two nonparametric probabilistic forecasting approaches using extreme learning machine based forecaster are also developed, and possess significant advantages such as a much simplified problem formulation and no need of prior knowledge about point forecasting errors. The first one is the direct interval forecasting approach that is proposed to directly produce optimal prediction intervals in terms of the novel cost function combining reliability and interval score. The second one is the Pareto optimal interval forecasting approach which aims to construct optimal prediction intervals through reaching the Pareto front of two quality index reliability and sharpness. The electricity market has important influence on the management of modern power systems, e.g., encouraging the development of renewable energy and the participation of consumers in smart grids. Precise electricity price prediction would assist market participants to properly deal with various decision making problems. Similar to wind power, electricity price also demonstrates significant nonstationarities and is fairly difficult to accurately forecast. An advanced hybrid approach is proposed for electricity price forecasting in this thesis to estimate prediction uncertainty of electricity price. Probabilistic load flow computation is implemented considering the power system integrated with large wind farm. Based on the developed generalized Lambda distribution model for wind speed, the probability property of wind power can be described accordingly. Comprehensive numerical studies demonstrate that the point estimate method can give accurate estimation of probabilistic load flow in the environment with wind power integration, which can help to investigate the impacts of wind power and facilitate wind farm planning, power system operational and expansion planning, etc.