Short-term operation of the power system and the natural gas system considering uncertainties

Hu, Dan
Major Professor
Sarah M. Ryan
Committee Member
Journal Title
Journal ISSN
Volume Title
Research Projects
Organizational Units
Journal Issue
Industrial and Manufacturing Systems Engineering

Electricity generation increasingly relies on natural gas for fuel. The competing demands for gas by natural gas-fueled generators and other users, the differences in the timing of short-term operations and markets between the natural gas system and the power system, and the deepening penetration of variable renewable energy in the power system cause difficulties in operating the two systems economically and reliably. This dissertation consists of three papers which present different models and methods for the short-term operation of the natural gas and power systems considering different sources of uncertainty.

From the viewpoint of a centralized system operator, who can operate the power system and natural gas system simultaneously, we first compare two approaches to addressing the uncertainty in the joint scheduling of a combined natural gas and power system. A stochastic programming model and a deterministic model with reserves are formulated to minimize the daily operational cost and investigate the hourly unit commitment and economic dispatch decisions in the power system as well as the hourly working schedule of the natural gas system while satisfying all the operational constraints. In the deterministic model, the reserves proportional to the wind energy forecast are used to mitigate the effect of the uncertainty in wind energy, whereas in the stochastic programming model the day-ahead decisions are made while explicitly considering the wind energy uncertainty. To tackle the nonlinear constraints on the gas flows in pipelines, we approximately linearize those nonlinear constraints by adding multiple binary variables and constraints. Through numerical experimentation, the number of piecewise linear segments is chosen to balance accuracy and computational efficiency. The simulation results of two case studies indicate that, when the total wind capacity exceeds 15\% of the conventional generation capacity, the stochastic programming model produces schedules with comparable or lower cost and energy shortages than the deterministic model with reserves.

The centralized system operator modeled in the first paper does not exist in the real U.S. energy market. From a more realistic viewpoint of the power system operator, in the second paper, we quantify the effect of the uncertainty in the gas spot price on power system dispatch cost in the absence of wind energy. The influence of the natural gas system is considered in terms of fixed or uncertain parameters in the power system daily economic dispatch problem. A benchmark distribution of the dispatch cost is generated by Monte Carlo simulation conducted with the gas price fixed at its expectation while sampling from the marginal distribution for the load. For comparison, another dispatch cost distribution is generated by sampling from a joint distribution for the gas price and the load. The risk from uncertainty in the gas price is quantified by the distance between dispatch cost distributions or, alternatively, by the difference between the values of a risk measure applied to each distribution. We demonstrate that this risk quantification method helps to select from among alternative risk-mitigation strategies, such as providing dual-fuel capability or adding gas storage facilities at the system level.

In the third paper, we investigate the use of a reliability unit commitment (RUC) conducted after the day-ahead market unit commitment to manage the natural gas cost in the power system operations, where the operations of the two systems are separately optimized to minimize their own net cost. This separately optimized model incorporates the interruptible contract and the real-time market for gas, where an iterative process between the electricity and gas operations determines the real-time gas flows and prices. An ideal co-optimized model, where a centralized system operator optimizes the two systems simultaneously, is taken as a benchmark for comparison. By numerical studies, we demonstrate the ability of the RUC step to reduce power system cost, maintain a low real-time gas price, and avoid real-time gas supply deficiency.