Publications
342
Thesis
14
Department
29
Researchers
285

Recent Additions
- ThesisRepresentación discreta de variables independientes multi-dimensionales en la operación de sistemas eléctricos de potencia para estudios de mediano y largo plazo(Universidad Técnica Federico Santa María, 2013-03)La presente Tesis analiza, propone y aplica metodologías de representación reducida (discretas) para variables multi-dimensionales de entrada en modelos de optimización de sistemas eléctricos de potencia. Estas variables, que se presentan a la vez en distintos puntos físicos del sistema de potencia, bien pueden aparecer en el lado derecho (términos constantes) de las restricciones o como coeficientes de la función objetivo de los problemas de optimización económica de sistemas de potencia. En modelos de sistemas de mediana a gran envergadura no es posible considerar todas las variables de entrada en forma desagregada. Por ejemplo, en modelos de optimización de mediano y largo plazo se agrupan instantes de características similares en cuanto a las variables de entrada, representando cada grupo con un sólo valor para las variables. Del mismo modo, en modelos complejos y/o de gran envergadura de programación estocástica se reducen los escenarios para obtener una formulación cuya resolución sea computacionalmente factible. Lo anterior lleva en forma casi natural a intentar encontrar metodologías que permitan realizar la agrupación temporal o la reducción de escenarios, problemas que son equivalentes, de forma óptima. En este contexto, “óptima” se refiere a que la solución a los problemas simplificados sea lo más similar posible a la solución del problema de optimización original. Para encontrar las representaciones reducidas se propone usar técnicas de clustering y otras basadas en reducción de escenarios. El desempeño de las técnicas propuestas es contrastado con las metodologías actualmente usadas en la industria y la academia en tres casos de estudio: (1) la expansión del Sistema Interconectado Central (SIC) usando programación estocástica para considerar la variabilidad hidrológica; (2) el dimensionamiento óptimo de la interconexión entre sistemas aislados con características de carga diversas; y (3) la operación de mediano plazo del Sistema Interconectado Norte Grande (SING) considerando niveles elevados de penetración eólica. En los tres casos las metodologías propuestas superaron a las actualmente usadas en términos de los errores en los resultados, en especial en lo que respecta a los flujos en los últimos dos casos de estudio. La razón para esta superioridad recae en que los métodos propuestos son capaces de considerar la estructura real de los datos durante la construcción de las representaciones reducidas de mejor forma que los métodos actualmente utilizados. El caso de estudio más que llevó al problema de optimización más complejo fue la expansión del SIC, puesto que se resolvió el problema de expansión conjunta de generación y transmisión considerando la variabilidad hidrológica. Los resultados dieron lugar a planes de expansión que disminuyen los costos totales cerca de un 5% respecto a los planes indicativos de expansión determinados por la autoridad (plan de expansión de generación de la Comisión Nacional de Energía, CNE, y plan de expansión de transmisión del Centro de Despacho Económico de Carga del SIC, CDEC-SIC).
- ThesisMachine Learning Aided Column Generation for Solving Transmission Network Expansion Planning(Universidad Técnica Federico Santa María, 2024)Many countries around the world are experiencing an energy transition towards low-carbon economies, in order to reduce greenhouse gas emissions and combat climate change. One of the main sectors called to lead this transition is the electricity sector, mainly through the massive incorporation of renewable energies and the decommission of coal-fired power plants. In Chile, according to the long-term Chilean energy policy, the target is to reach net zero emissions by 2040 and supply 100% of the electrical demand from renewable energies by 2050. The Chilean commitment to combat climate change already started in 2011 by promoting the introduction of renewable energies in the electrical sector. Since then, the installed capacity of variable renewable energies (VRE) increased from 4% of the total capacity in 2011, to more than 40% in 2023. Under this global trend, the transmission infrastructure is key for achieving a cost-effective and secure energy transition. Indeed, to have an adequate transmission capacity is fundamental to accommodate new renewables and facilitate their integration into the market, supply the growing demand in a cost-effective way and enhance competition to ensure market efficiency. Increasing transmission capacity is not an easy task, due to growing difficulties in acquiring new rights-of way, the opposition of local communities to the construction of new lines and long processes of environmental licensing, among others. The aforementioned challenges considerably extend the development time of new transmission line projects. Delays in the commission of expansion projects can have significant technical and economic repercussions. Key concerns include unexpected curtailment of renewable energy sources (RES), where new RES investments can experience unexpected, diminished returns if the transmission capacity if not available on time for the delivery of their energy into the grid, increasing energy costs for consumers and greater transmission losses caused by the sub optimal operation of the transmission network due to underdevelopment. Flexible technological solutions can alleviate these issues by enhancing the network capacity and reliability in shorter time and without the need of installing new transmission lines. Amongst these alternative technological solutions, we can find line reconductoring, HVAC to HVDC reconversion, Flexible AC Transmission Systems (FACTS), Battery Energy Storage System (BESS) installed as grid boosters and others. What all these solutions have in common is shorter development times, lower investment cost and less challenging regulatory processes compared to the investment in new transmission lines. Consequently, nowadays there are plenty of technological options to increase the transmission capacity. However, the main challenge consists of determining which technology, when and where should be incorporated into the system to minimize total system costs while maintaining security, social and/or environmental constraints. To answer this question, a usual approach is to address this challenge as an optimization problem, the transmission network expansion planning (TNEP) problem. The TNEP is usually formulated as a mixed-integer linear problem (MILP), where integer variables are related to the investment decisions, while continuous variables are related to the power system operation. In more complex formulations, integer variables can also comprise operational variables such as the generator's unit commitment. The TNEP can also be formulated as a nonlinear problem (MINLP), for example to consider power system losses and/or AC power flows [6]. Furthermore, the use of stochastic and/or robust models has increased in the last years, due to the growing uncertainty involved in the network planning. In the past, when power systems were vertically integrated and the generation matrix was dominated by conventional fossil-fuel based generation, there was little uncertainty regarding future generation capacity and power feed-in. Therefore, the majority of TNEP formulations were deterministic. However, the deregulation of the electricity markets, along with the massive introduction of renewable energies, brought new uncertainties regarding future generation capacity and availability. The level of uncertainty regarding future generation capacity increased even more with the introduction of renewable energies, because of their short construction lead times (1 - 2 years), compared to conventional fossil-fuel based generation (3-5 years), thus increasing the need of adopting stochastic or robust TNEP formulations. Furthermore, it is expected that the use of stochastic and robust model will increase in the future, in order to incorporate new sources of uncertainties, for examples, those related to long-term effects of the climate change. In all cases, the TNEP problem is NP-hard, which means that it cannot be solved in polynomial time. For large-scale realistic-size power system models, solving the TNEP problem demands significant amounts of computational resources and could even require several days to be solved using a state-of-the-art optimization algorithm. Furthermore, the appearance of emergent technologies that increase the operational flexibility of power systems have introduced new challenges for the TNEP. Examples of such technologies are FACTS, storage devices, special protection schemes, energy storage-based grid booster and HVDC links, as well technological option to increase the use of existing transmission network assets such as line uprating and HVDC conversion. For each new technology, novel formulations capable of capturing their value in the power system are required, which usually means having more complex formulations, more variables, and constraints, thus resulting in models which are even more challenging to solve. In particular, proper models of energy storage systems require the introduction of time-coupling constraints to account for the energy balance, thus significantly increasing the complexity of the TNEP problem. Note that, to capture the benefits of short-term storage systems such as batteries, requires increasing the time resolution of the models (for example, representative days instead of isolated operating conditions), which in turn increase the computational burden of the TNEP formulation.
- ThesisLocalización óptima de baterías en sistemas con alta penetración de energías renovables considerando criterios de robustez(Universidad Técnica Federico Santa María, 2024)Durante los últimos años, los sistemas eléctricos de potencia (SEP) alrededor del mundo han experimentado un crecimiento constante de las energías renovables no convencionales (ERNC). Los números son claros, entre los años 2010 y 2021, la capacidad instalada de las ERNC a nivel global aumentó en un 150%, pasando de 1.2 TW el año 2010 a 3.0 TW al año 2021. De hecho, solo en el año 2021 la capacidad instalada aumentó un 9.1% respecto al año anterior. De particular importancia ha sido el crecimiento de las tecnologías de generación variable con convertidor (TGVCC) como la solar fotovoltaica y eólica de velocidad variable, cuya capacidad instalada respecto al año anterior aumentó en un 19% y 13% respectivamente. En Chile, el aumento de las TGVCC ha sido similar al resto del mundo. El país cuenta actualmente con 5,1 GW de capacidad instalada en centrales fotovoltaicas y 3,7 GW en generación eólica, que en conjunto representan un 30,0% de la capacidad instalada total de generación en el país. Las soluciones para mitigar los problemas de estabilidad en redes débiles con altos niveles de TGVCC cubren un amplio espectro de alternativas que van desde refuerzos de red clásicos hasta cambios en el sistema de control de los convertidores. Adicionalmente, la robustez en áreas débiles de una red se puede mejorar localmente mediante la incorporación de equipos basados en convertidores como SVC, STATCOM, BESS, etc. En este contexto, entre los dispositivos que más se han investigado en el último tiempo se encuentran los sistemas de almacenamiento de energía, en particular, las baterías (BESS). Diversos estudios y experiencias practicas han mostrado que los BESS son una buena alternativa para mejorar la estabilidad de los SEP debido a sus tiempos rápidos de respuesta y capacidad de inyectar potencia activa y reactiva de forma independiente. Entre las aplicaciones más estudiadas se encuentran la incorporación de lazos de control para que los BESS puedan aportar a la estabilidad de frecuencia y tensión en los SEP. En estos trabajos se destaca la incorporación de inercia virtual control de reactivos, respuesta rápida de frecuencia, regulación de frecuencia y regulación de tensión. Si bien las diversas maneras en que los BESS pueden aportar a la estabilidad de los SEP han sido ampliamente investigadas durante los últimos años, la localización óptima de estos equipos en escenarios con altos niveles de TGVCC aún no ha sido muy investigada. Tradicionalmente, los trabajos en esta línea se centran en localizar los BESS considerando criterios económicos buscando, por ejemplo, minimizar los costos de un sistema en el marco de la operación o de la planificación. Si bien este tipo de trabajos aprovechan las ventajas económicas que los BESS pueden entregar, no consideran sus posibles aportes a la estabilidad de los SEP. En el contexto anterior, en este trabajo se propone una metodología para determinar la localización óptima de BESS en sistemas eléctricos débiles con bajos SCL, considerando criterios de estabilidad.
- ThesisMulti-agent Deep Rainforcement Learning for Efficient Multi-Timescale Bidding of a Hybrid Power Plant in Day-Ahead and Real-Time Markets(Universidad Técnica Federico Santa María, 2022)Effective bidding on multiple electricity products under uncertainty would allow a more profitable market participation for hybrid power plants with variable energy resources and storage systems, therefore aiding the decarbonization process. This study deals with the effective bidding of a photovoltaic plant with an energy storage system (PV-ESS) participating in multi-timescale electricity markets by providing energy and ancillary services (AS) products. The energy management system (EMS) aims to maximize the plant’s profits by efficiently bidding in the day ahead and realtime markets while considering the awarded products adequate delivery. EMS’s bidding decisions are usually obtained from traditional mathematical optimization frameworks. However, since the addressed problem is a multi-stage stochastic program, it is often intractable and suffers the curse of dimensionality. This document presents a novel multi-agent deep reinforcement learning (MADRL) framework for efficient multi-timescale bidding. Two agents based on multi-view artificial neural networks with recurrent layers (MVANNs) are adjusted to map environment observations to actions. Such mappings use as inputs available information related to electricity market products, bidding decisions, solar generation, stored energy, and time representations to bid in both electricity markets. Sustained by a price-taker assumption, the physically and financially constrained EMS’s environment is simulated by employing historical data. A shared cumulative reward function with a finite time horizon is used to adjust both MVANNs’ weights simultaneously during the learning phase. We compare the proposed MADRL framework against scenariobased two-stage robust and stochastic optimization methods. Results are provided for one year round market participation of the hybrid plant at a 1-minute resolution. The proposed method achieved statistically significant higher profits, less variable incomes from both electricity markets, and better provision of awarded products by achieving smaller and less variable energy imbalances through time.
- ThesisEstrategia de Control Óptimo para plantas Fotovoltaicas de Gran escala conectadas a Redes trifásicas mediante Convertidores Puente H en Cascada(Universidad Técnica Federico Santa María, 2022)Un pilar fundamental para superar la crisis climática global es la reducción drástica en las emisiones de gases de efecto invernadero producto de la actividad humana. En el sector eléctrico, grandes bloques de energía provenientes de fuentes contaminantes han sido reemplazadas por la instalación de plantas fotovoltaicas de gran escala. Para las cuales, el desarrollo de nuevas topologías de convertidores basados en electrónica de potencia busca incrementar su eficiencia y rendimiento. Particularmente, el convertidor puente H en cascada (CHB) ha sido objeto de interés debido a que su implementación posee una alta granularidad del esquema MPPT, un menor requerimiento de filtrado en las corrientes de línea y la posibilidad de ser conectado a una red de media tensión mediante el apilamiento de módulos de menor tensión. Naturalmente, los desbalances de generación entre fases han sido el mayor desafío técnico para propiciar la adopción de esta topología. No obstante, la solución documentada de este problema radica en la inyección de una tensión de modo común en las tensiones de fase del convertidor CHB. En este trabajo de tesis se elabora un modelo matemático que permite determinar una tensión de modo común con mínimo contenido armónico para lidiar con desbalances de generación entre fases manteniendo un flujo de corrientes de red balanceado y con bajo contenido armónico. Por lo demás, se confeccionan expresiones analíticas para la tensión de modo común óptima y las regiones factibles de esta técnica. De esta manera, se desarrolló un análisis en la implementación de la estrategia de inyección de una tensión de modo común que derivó en la construcción de un algoritmo iterativo y distribuido para calcular esta tensión en tiempo real. Por su parte, un controlador de tipo Phase-Shifted Model Predictive Control (PSMPC) se empleó para el control en tiempo real de las corrientes inyectadas a la red, cuyas soluciones óptimas se calcularon mediante un algoritmo de tipo Active-Set Method (ASM). Finalmente, las pruebas experimentales en laboratorio con un convertidor CHB de siete niveles permitieron verificar la efectividad de la estrategia propuesta logrando desbalances de corriente de, a los más, un 1% y un THDe máximo de 2.8% en un conjunto de experimentos con severos desbalances de potencia entre fases.
Most viewed
- PublicationThe experimental facility for the Search for Hidden Particles at the CERN SPS(2019-03-25)The Search for Hidden Particles (SHiP) Collaboration has shown that the CERN SPS accelerator with its 400 GeV c proton beam offers a unique opportunity to explore the Hidden Sector [1–3]. The proposed experiment is an intensity frontier experiment which is capable of searching for hidden particles through both visible decays and through scattering signatures from recoil of electrons or nuclei. The high-intensity experimental facility developed by the SHiP Collaboration is based on a number of key features and developments which provide the possibility of probing a large part of the parameter space for a wide range of models with light long-lived superweakly interacting particles with masses up to O¹10º GeV c2 in an environment of extremely clean background conditions. This paper describes the proposal for the experimental facility together with the most important feasibility studies. The paper focuses on the challenging new ideas behind the beam extraction and beam delivery, the proton beam dump, and the suppression of beam-induced background.
- PublicationNew aspects of muon-electron nuclear conversion(2002-09-19)We found a new important tree-level contribution to muon–electron nuclear conversion from neutrino exchange between two quarks in the same nucleon and demonstrated that this process, contrary to common belief, can be observed in the near future experiments if there exists a mixed sterile-active neutrino state νh heavier than the quark confinement scale Λc ∼ 1 GeV. From the present non-observation of muon–electron conversion we derive new experimental constraints on νh − νe,µ mixing
- ThesisTorque Ripple Reduction of a Switched Reluctance Machine by Implementing an Optimization Based Control Strategy(Universidad Técnica Federico Santa María, 2023)In recent times switched reluctance machines have started to draw attention as a feasible alterna tive to be implemented in electrical vehicles, due to its low-cost manufacturing and reliability at high speeds. However, the usual operation of these machines involves a high torque ripple, which can bring unwanted mechanical consequences. This is why the control strategy of these machines is vital to overcome these drawbacks. This work proposes a novel control strategy for torque–ripple minimization of a switched reluctance machine drive. The strategy is composed of two stages: (i) an outer flux–linkage reference generation layer delivers optimized patterns obtained offline via mixed-integer quadratic programming; (ii) an inner flux-linkage control loop that tracks references of the outer layer by applying an optimal switching sequence model predictive control algorithm. Each stage is tested in simulation and the results show that both the proposed reference genera tion technique and flux control algorithm can produce lower torque ripple and better flux–tracking performance than their state-of-the-art counterparts, them being the torque sharing function ap proach for reference generation, and the finite control set model predictive control and the deadbeat predictive control algorithms for flux–tracking. The proposed strategy and its components are im plemented in an experimental setup, which consisted of a 2.32 kW switched reluctance machine being fed by three Asymmetric Bridge Converters. The switched reluctance machine is coupled to a 2.23 kW induction machine working as a regenerative brake and is controlled by an external AC drive. The experimental outcome validated the proposed approach and the results obtained by simulation
- PublicationSoot Volume Fraction Measurements by Auto-Compensating Laser-Induced Incandescence in Diffusion Flames Generated by Ethylene Pool Fire(2021-11-08)The main characteristics of pool fire flames are flame height, air entrainment, pulsation of the flame, formation and properties of soot particles, mass burning rate, radiation feedback to the pool surface, and the amount of pollutants including soot released to the environment. In this type of buoyancy controlled flames, the soot content produced and their subsequent thermal radiation feedback to the pool surface are key to determine the self-sustainability of the flame, their mass burning rate and the heat release rate. The accurate characterization of these flames is an involved task, specially for modelers due to the difficulty of imposing adequate boundary conditions. For this reason, efforts are being made to design experimental campaigns with well-controlled conditions for their reliable repeatability, reproducibility and replicability. In this work, we characterized the production of soot in a surrogate pool fire. This is emulated by a bench-scale porous burner fueled with pure ethylene burning in still air. The flame stability was characterized with high temporal and spatial resolution by using a CMOS camera and a fast photodiode. The results show that the flame exhibit a time-varying propagation behavior with a periodic separation of the reactive zone. Soot volume fraction distributions were measured at nine locations along the flame centerline from 20 to 100 mm above the burner exit using the auto-compensating laser-induced incandescence (AC-LII) technique. The mean, standard deviation and probability density function of soot volume fraction were determined. Soot volume fraction presents an increasing tendency with the height above the burner, in spite of a local decrease at 90 mm which is approximately the position separating the lower and attached portion of the flame from the higher more intermittent one. The results of this work provide a valuable data set for validating soot production models in pool fire configurations.
- Organizational Unit