Addressing the Challenges of DCOPF Based Decision-Making Algorithms in Modern Power Systems
Natural disasters have been determined as the leading cause of power outages, causing not only huge economic losses, but also the interruption of crucial welfare activities and the arise of security concerns. Because of the later, decision-making considering grid modernization, power system economics, and system resiliency has been a crucial theme in power systems’ research. The need to better withstand catastrophic events and reducing the dependency of bulky generating units has propelled the development and better management of behind-the-meter generation or distributed energy resources (DERs). DERs can assist in the grid in different manners, not only by meeting energy demand goals, but also by reducing the overall system operating cost and support the global emerging environmental objectives. By being closer to the consumer or load side, DERs avoid transmission line losses and can contribute to reduced system congestion which translates in reduced operational cost, considering that two of the three pillars that affect Locational Marginal Price (LMP) are losses and congestion. Additionally, a higher power system resilience is achieved by reducing the dependency of transmission lines that can potentially fail during a contingency event. Furthermore, DERs are primarily based on Renewable Energy Sources (RESs) which contributes to diversifying the energy generation supply and reducing the grid’s fuel dependency. As a consequence, DERs in the form of RESs have the potential of reducing the overall greenhouse gas (GHG) emissions. While RESs have demonstrated over the years their numerous benefits, they also inflict a big challenge to the power grid in the form of added uncertainty and power flow instability by demanding higher system flexibility and improved decision-making algorithms. To account for this induced RESs power output uncertainty and improve power system resiliency in the decision-making process, the implementation of stochastic optimization via Monte Carlo simulations has been widely and commonly used, where different probability-based scenarios are used to account for possible profile outcomes and renewable sources uncertainty. One of the many drawbacks of employing Monte Carlo simulations is that a power system optimization problem might be solved containing a worst-case scenario which differs significantly from the actual system condition. The latter yields consequently to a higher optimal cost, caused by the system’s uncertainty. As previously mentioned, RESs enhanced management can also be achieved by improving system flexibility. A well-studied power flow control technique consists of actively changing the power system transmission topology using Optimal Transmission Switching (OTS), leading to improved power flow and transmission congestion relief. The OTS problem is based on an DC Optimal Power Flow (DCOPF) algorithm with the addition of a big-M constraint to maintain problem linearity. In such big-M based algorithm, the selection of the M value can impose great computational complexity challenges; In extreme cases, an erroneous M value selection can potentially lead to numerical instability, long solving times, and even compromise problem optimality or feasibility. Therefore, the importance of selecting an appropriate M value is noticed and well known in literature. Finally, the increased penetration of RESs will demand a better use of the different generating resources across the network control areas. Regions with great solar irradiance capabilities are and will benefit from solar generation during sun hours. In the other hand, regions with great wind and tidal power capabilities might expect high power output during night hours. Furthermore, regions with must-run generation such as nuclear or hydro power need to maintain an almost constant power output to ensure power resilience and reliability. With the latter zone’s descriptions, different regions with diverse capabilities and characteristics can widely benefit from neighboring entities by allowing power exchange through system interconnections. As a consequence, distributed or decentralized algorithms that are able to reliably balance and co-optimize multiple control areas while sharing minimal system information are required. This type of algorithms optimizes not only the region’s own power dispatch, but also optimize the overall interconnected regions to exploit power exchange and interconnection benefits. This thesis will evaluate the cost effects of decision-making under uncertainty in the power system caused by wind power RESs, where a stochastic optimization mixed integer linear program (MILP) DCOPF based model employing Monte Carlo simulations is implemented. Furthermore, an evaluation of optimally choosing the big-M value for OTS is presented to describe some of the computational challenges of decision-making in the power system. Finally, a distributed consensus based DCOPF algorithm is presented to compare the resiliency impacts between centralized and decentralized decision-making mechanisms on power systems. The presented decentralized algorithm is based on the Alternating Direction Method of Multipliers (ADMM), founded on a penalty-based objective which implements the augmented Lagrangian method for constrained optimization problems.
Ramirez Burgueno, Luis Daniel, "Addressing the Challenges of DCOPF Based Decision-Making Algorithms in Modern Power Systems" (2023). ETD Collection for University of Texas, El Paso. AAI30522575.