Day-Ahead Generation Scheduling of Power System in Presence of Fast Generation Resources under Uncertainty of Renewable Generation Units.

Extended Abstract
Introduction: Increasing penetration rate of renewable energy resources will face operation of future power systems with serious technical and economic challenges. Due to uncertain output generation of these resources, the need to cover these uncertainties led to the emergence of a concept known as flexibility in the power system. One of the serious challenges in the field of operational flexibility of the power system is providing the required ramping capacity by the power system. Therefore, in order to solve this challenge, CAISO market introduced a product called Flexible Ramping Product (FRP). One of the most important resources of this product is dispatchable generation units. Indeed, increasing the penetration of non-dispatchable generation resources has provided an opportunity for the presence of fast generation resources such as gas units in wholesale markets such as day-ahead market. Fast resources, with the ability to change the status from inactive to active and vice versa during operation time, play an important role in providing the ramping capacity required by the power system. In order to solve day-ahead scheduling problem, robust optimization has been recently considered compared to other methods due to its accuracy and trackability. Robust approach provides a solution that enables the system to provide the required flexibility in face of the worst-case scenario. In most studies, the base-case of system has not been considered; this fact has led into a deterministic reserve allocation that is in contrast to the uncertain behavior of non-dispatchable resources. In fact, in those studies, the reserve and the clearing of energy are not possible together, and it is contrary to the acceptable approach of the electricity markets. In order to modify these abovementioned studies, many studies were conducted with the aim of jointly energy and ramping capacity reserve clearing to maximize social welfare. The noticeable point about these studies was their lack of attention to the potentials of fast resources in providing the flexibility required by the power system. Actually, in the existing literature, solving methods with high computational complexity and low convergence speed such as Benders decomposition method have been used for scheduling problem solving. In this paper, for day-ahead scheduling, the potentials of fast resources in providing operational flexibility of the power system have been considered. Also, the scheduling problem, which has been presented in a two-stage and three-level robust model, is solved by the column and constraint generation (CCG) method. In fact, the CCG method, due to the existence of optimality primal cuts in the master problem, makes reduction in computational complexity and an increase in the speed of convergence during the problem solving process. Furthermore, contrary to the existing literature, ramping capacity reserve is provided due to the ramping limit of slow and fast generation units.
 Materials and Methods: In this paper, a two-stage, and a three-level method along a mixed integer adaptive robust programming has been used and solved by means of CCG method. The master problem involves the first level problem which includes the minimization of operating costs, and the sub-problem includes the max-min model of the second and the third level problems. In order to identify the worst-case of power imbalance in the power system, it is necessary to solve the sub-problem in form of maximization problem. Therefore, in order to convert the max-min model of sub-problem to a single-level maximization model, it is necessary to use the dual theory to convert the minimization of the third level problem to its maximization equivalent. In each iteration by solving the master problem, the minimum cost of providing operational flexibility is determined. Then, according to the identified uncertainties in the system, the sub-problem is solved according to the decisions in the master problem; after that the flexibility of the system against the fluctuations of renewable generation resources at this stage is identified. According to the required flexibility identified by the sub-problem, the master problem is resolved and the exchange of information between the master problem and the sub-problem continues until the system will be flexible against the uncertainties of wind-based generation units. It should be noted that in order to model the uncertainty in the sub-problem, polyhedral uncertainty set has been used, which in relation to its linear nature, reduces the computational complexity of the problem.

Results: The proposed approach in this paper is implemented on the IEEE 24 bus RTS. The results are as following:
a) By classifying generation-side resources as slow and fast units, reduction in operating costs and robustness costs are obvious; these costs are reduced by 0.85% for the worst-case scenario. It is because of the utilization of fast resources during operation in such a way that these resources with the ability to provide non-spinning ramping capacity reserve can significantly reduce the need of the operation of expensive slow dispatchable generation units. These slow units impose a fixed and variable cost of generation on the system to provide the required flexibility of the power system.
b) The use of the proposed approach based on the column and constraint generation method to solve the proposed operational planning problem has had so good efficiency in the problem solving process that the problem has converged in a maximum three iterations and an average of 64 seconds computational time.

Discussion and Conclusion: By classifying generation-side resources as slow and fast units, it can be realized that fast dispatchable units, due to their response speed in start-up and shutdown and due to their ability in non-spinning operating, can reduce the need for spinning operation of expensive fast and slow units to provide the required ramping capacity in case of uncertainty realization. The results show that this approach has reduced operating and robustness costs by an average of 0.52% for different uncertainty budgets. Furthermore, this paper illustrates that the efficiency of proposed approach based on CCG method is sufficient suitable for solving day-ahead scheduling problem in such a way that the problem converges to the optimal answer in a maximum of 3 iterations and an average of 64 seconds computational time.