07 de noviembre de 2024
Resumen:
The high penetration of renewable generators, together with the deployment of nonconventional energy storage technologies and the promotion of capacity increases in the interconnection facilities, are transforming power systems worldwide. This paradigm change entails new challenges for energy models to represent and analyze current and future market trends properly.
In this regard, medium-term fundamental models provide accurate forecasts of the operation of electricity markets in which medium- and long-term decisions like hydro management, fuel trading, or third-party access contracting must be handled. At the same time, other concerns, such as thermal and renewable generation, short-term storage, and transmission constraints, are also considered in detail.
In this context, the unit commitment problem is a powerful tool that accurately represents the technical operation of market players’ assets. However, modeling realsize power systems in medium-term horizons with a precise time granularity involves a high computational burden, especially when dealing with integer programming and uncertainty considerations.
Accordingly, this thesis proposes modeling improvements for the medium-term unit commitment problem, seeking to increase the accuracy in representing real electricity markets on an hourly basis in the most computationally efficient way. The structure of the document is briefly described below:
• Firstly, this thesis introduces the background in which it is immersed, specifying its objectives and contributions to the literature and providing the outline and structure of the document.
• Secondly, this thesis addresses an in-depth review of the unit commitment problem literature, enhancing the scope of other previous surveys by describing in detail the main different optimization techniques applied to the unit commitment and exposing their computational implications through an analysis of modeling detail, power systems representation, and computational performance of their case studies. There, research gaps are identified.
• Third, this thesis exposes the modeling simplifications that are necessarily assumed when the case studies of renowned medium-term energy models are faced. Moreover, it proposes a soft-linking methodology to overcome the limitations of some of these simplifications by taking reliable outputs and subjecting them to a more rigorous post-processing step. This phase provides feasible and optimal generation schedules for a multi-area thermal portfolio belonging to a market player, maximizing profits and considering strategic terms in a risk aversion environment. The computational burden associated with this phase is studied in detail.
• Fourth, this thesis analyzes the implications of utilizing different balance constraints in the unit commitment problem under different convergence criteria. In turn, it studies the impact of the nature of some input data, like the variability in thermal demand, generation portfolios’ sizes and composition, and duration of the time span. Furthermore, given that it is appreciated that these concerns influence the computational performance and the formulations’ tightness and compactness (T&C) do not always predict its behavior, “arduousness” is introduced as a new metric to evaluate resolution processes.
• Fifth, this thesis proposes a computationally efficient piecewise formulation for accurately representing start-up costs in the medium-term unit commitment problem. It is compared to a renowned formulation that uses stairwise functions to model these costs, and its validity is demonstrated. Later, a tighter and more compact formulation is presented to improve the computational performance further.
Additionally, the implications of continuous/integer variable declarations are studied, and the application of these detailed approaches to the short-term unit commitment is analyzed.
• Sixth, this thesis proposes a coordination methodology to achieve a computationally tractable representation of real-size multi-area power systems in the medium term on an hourly basis. It runs a first phase with modeling simplifications like time series aggregation (TSA) techniques but considers the whole horizon. Reliable medium-term decisions are taken from this step and imposed as signals to coordinate
a rolling horizon phase in which detailed time granularity and chronological relationships are managed. The outstanding result accuracy, the reasonable run times, and the modest computational requirements, even utilizing mixed integer programming (MIP), open the door to uncertainty representation.
• Seventh, this thesis summarizes the main results of the research works presented throughout the document. Besides, their original contributions to different research fields are highlighted, and future research work is exposed.
Resumen divulgativo:
Esta tesis doctoral afronta el reto de mejorar la representación de sistemas eléctricos reales en modelos horarios de medio plazo. La investigación se centra en el unit commitment problem, logrando obtener procesos de optimización eficientes a nivel computacional y perfiles de operación detallados.
Descriptores: Investigación Operativa, Tecnología Energética
Palabras clave: Unit commitment, Optimization models, Computational efficiency, Risk management, Power systems, Electricity markets, Multi-area representation, Medium-term representation, Hourly horizons, Clustering techniques, Feasible operation, Thermal generation, Hydro generation, Renewable generation, Energy storage, Fuel contracts, Third-party access, Balance constraints, Piecewise formulations, Long-term coordination, Operational research, Numerical optimization, Mixed-integer programming, Linear programming, Tightness, Compactness, Arduousness
Cita:
L. Montero (2024), Improving medium-term models to deal with the low-carbonreality of modern power systems. Madrid (España).
