(756c) Optimal Scheduling of a Microgrid on a Steam-Assisted Gravity Drainage (SAGD) Facility

Steam-Assisted Gravity Drainage (SAGD) is a thermal recovery technique, which largely accounts for the extraction of the 1.7 trillion barrels of heavy and extra heavy oil (bitumen) present in the oil sands of Western Canada. Due to the injection of superheated steam in the oil sands formation, at the end of the production, substantial amount of waste heat is left behind in the reservoir. In this work, we explore the opportunity of utilization of this waste heat to produce electricity using a steam turbine operating on an Organic Rankine Cycle (ORC). The premise being that the electricity produced could then meet the load demands of other operational SAGD facilities, thus making the SAGD facilities less dependent on a macrogrid, more profitable, and leave a lower carbon footprint in the process. However, preliminary research has shown that the waste heat left behind in the reservoir, is by itself, incapable of sustaining the operations of an entire SAGD facility. This issue can be overcome by incorporating other renewable sources of energy generation working synergistically with the SAGD waste heat based steam turbine. This leads to a challenging scheduling problem since one has to make optimal decisions in the volatile electricity and natural gas market in the face of inherent intermittency in renewable generation and local demand.

In this work, we focus on the dispatch of a microgrid consisting of a bi-directional connection to the macrogrid, gas turbine, wind turbine, steam turbine operating on an ORC and a battery bank. We propose and formulate a Kelly Criterion (KC) based optimization problem to supply electricity (from renewables and waste heat) and make profitable decisions in the volatile electricity and natural gas pricing system. The KC is mainly used in economics to hike the return on investment and is built on the principle of maximization of the growth rate based on successive investments or bets as a fraction of the total investible capital. The technique does not have any explicit dependence on the starting capital, and provides the optimal betting or investment fraction for maximal expected return on investment, and in turn, capital growth. For this study, the starting capital is analogous to the amount of electricity generated, and the KC based optimization technique provides the optimal strategy to distribute that electricity generated as a function of the electricity and natural gas pricing markets. The electricity price was forecasted using a three-layered neural network working on a Nonlinear Autoregressive time series system identification technique with Exogenous inputs (NLARX), while the natural gas price was estimated based on a Gaussian distribution centered around the current natural gas price.

Since the KC technique does not depend on the amount of electricity generated, therefore, an accurate forecast of the parameters influencing the generators, such as wind for wind turbines, is not required. The optimization study was made possible by a bidirectional communication between MATLAB (the price forecaster) and General Algebraic Modeling System (GAMS) (the optimizer) software. The results indicate that the KC based optimization technique for microgrid scheduling outperforms the standard Dynamic-Real Time Optimization technique.