PROJECTS

MSOpt –Mathematical Methods for the Optimization of Decision-making Processes based on Multi-source Data

• Summary: The last decades of human progress have witnessed a revolution in the development and implementation of information and communication technologies, leading to an explosion in the availability of data, observations, and measurements related to physical and social processes and systems. Today, the widespread use of data is crucial in diverse sectors of our society, such as communications, robotics, biology, and even in areas like politics, entertainment, and product and service commerce. In these and other sectors, data serves as a resource with enormous potential to provide value by enabling better decision-making, often in contexts characterized by high uncertainty. However, the effective utilization of data for decision-making is not without significant challenges and obstacles. These challenges are associated not only with the volume and complexity of data management but also, and especially, with its diverse origins and incomplete nature. Data often offers only a limited or biased view of the process or system from which it is obtained. In practice, however, it is not uncommon to encounter multiple decision-making processes within similar environments. These processes often face a shortage of individual-level data, yet collectively, they pool a substantial volume of valuable information. In this context, the main objective of this project is to generate knowledge, methods, and algorithms that facilitate the optimal use of information contained in data from multiple sources for decision-making under uncertainty. Additionally, we aim to create a mathematical framework that allows us to define and formalize the concept of the “value” of a dataset (understood as its capacity to improve the outcome of a decision) and develop methods for its quantification. This framework will then be used to discriminate between datasets that feed into the same decision-making process (or several analogous ones) but come from different sources (or providers). To develop these methods, we will employ advanced techniques in Stochastic Optimization, such as Distributionally Robust Optimization, and will need to generate new theoretical knowledge in this domain. Furthermore, to evaluate the mathematical methods derived from this project, we will apply them to address complex optimization and decision-making problems with high economic and/or social impact. This includes challenges arising from the necessary decarbonization of the energy sector and its imperative transition towards sustainability. Solving large-scale optimization problems will be a crucial aspect of this project, necessitating the development of tools that leverage data for the efficient resolution of such problems, potentially utilizing machine learning techniques.
• Dates: 01/01/2025 - 31/12/2028
• Funding Entity: Spanish Ministry of Innovation & Science

FlexAnalytics - Advanced Analytics to Empower the Small Flexible Consumers of Electricity (website)

• Summary: Could small consumers of electricity compete in the wholesale markets on an equal footing with the other market agents? Yes, they can and FlexAnalytics will show how. Activating the demand response, although a major challenge, may also bring with it tremendous benefits to society, with potential cost savings in the billions of euros. This project will exploit methods of inverse problems, multi-level programming and machine learning to develop a pioneering system that enables the active participation of a group of price-responsive consumers of electricity in the wholesale electricity markets. Through this, they will be able to make the most out of their flexible consumption. FlexAnalytics proposes a generalized scheme for so-called inverse optimization that materializes in a novel data-driven approach to the market bidding problem that, unlike existing approaches, combines the tasks of forecasting, model formulation and estimation, and decision-making in an original unified theoretical framework. The project will also address big-data challenges, as the proposed system will leverage weather, market, and demand information to capture the many factors that may affect the price-response of a pool of flexible consumers. On a fundamental level, FlexAnalytics will produce a novel mathematical framework for data-driven decision making. On a practical level, FlexAnalytics will show that this framework can facilitate the best use of a large amount and a wide variety of data to efficiently operate the sustainable energy systems of the future.
• Dates: 01/02/2018 – 31/01/2024
• Funding Entity: European Research Council (ERC), Starting Grant 2017

DYCON  - Data-driven Optimization Under a Dynamic Context

• Summary: The last decades of human progress have witnessed a revolution in the development and deployment of information and communication technologies. This revolution has led to, among other things, an explosion in the amount of information, in the form of data, observations and measurements, about physical and social processes and systems. The exponential growth of the amount of data available has brought about new opportunities in the field of Optimization-based Decision Making. Of all of them, in this particular project, the research team explore the possibility of using this data for two different, but closely linked, purposes, namely: the resolution of large and complex optimization problems, which, until now, have proved extremely difficult to solve; and the mathematical formulation of optimization models that allow making decisions with a higher level of confidence about their future outcome.
• Dates: 01/09/2021 - 31/08/2024
• Funding Entity: Spanish Ministry of Innovation & Science

QUALIFICA Andalusia Regional Government Excellence Programme: Mathware for the operation and planning of intelligent and sustainable energy systems

• Summary: The motivation behind this action is to grant funding, with the purpose of strengthening Andalusian university research institutes, R&D&I Centres and Infrastructures’ bids for “Severo Ochoa” or “María de Maeztu” seals of excellence. The purpose is to promote the development and implementation of actions aimed at improving already submitted applications, in order to achieve a better position for accreditation. The “Severo Ochoa” and “María de Maeztu” Centres of Excellence and “María de Maeztu Units of Excellence” accreditations recognise the best research centres and units that stand out for the relevance and impact of their results worldwide. In this framework, Mathware for the operation and planning of intelligent and sustainable energy systems is one of the research lines of the Unidad de Investigación en Matemáticas iMAT led by J.M. Morales and E. Carrizosa.
• The design, development and implementation of tools for the planning and operationof coupled gas-electric energy systems, with the possibility to transport hydrogen (mixed) through the gas infrastructure, and of active distribution networks (smart grids). These tools will be based on Mathematical Optimization and Data Science and require the design, study and implementation of algorithms and/or procedures to solve large non-convex optimization problems with integer variables. These algorithms and/or procedures may range from convex (conservative) approximations and decomposition methods to the application of mathheuristics.
• Dates: 01/05/2022 - 30/04/2025
• Funding Entity: Andalusia Regional Government (Council for Economic Transformation, Industry,  Knowledge and Universities)

Mathematical Optimization Methods for Decision Making Using Contextual Information

• Summary: Traditionally decision making under uncertainty has been a two-step process. First, a prediction model is used to generate the probability distributions that characterise uncertain future parameters. Then, in the second step, these distributions are used to solve a stochastic optimisation problem. With this in mind, the main objective of this project is to develop alternative approaches that i) take into consideration the optimisation problem when predicting uncertain parameters; ii) enable all potentially predictive information about the uncertain parameters (i.e. the context) to be exploited; iii) can be widely applied, and iv) can be solved by commercially available optimisation software. Although of general application, this project focuses on the application of the methodology developed in the electricity sector for three reasons: i) this sector has to address problems with enormous economic and social impact, ii) all decisions must be made without accurately knowing the relevant parameters, like, for e.g. the electricity demand or renewable generation, and iii) electricity networks are highly monitored systems that produce a vast amount of contextual data that can be exploited.
• Dates: 23/06/2021 - 31/12/2022
• Funding Entity: Andalusia Regional Government (Council for Economic Transformation, Industry,   Knowledge and Universities)

PowerMath - Mathematical Methods for Data-driven Power Systems

• Summary:  The massive use of data is a reality in many sectors of our society, ranging from communications, robotics, biology to politics, entertainment, shopping, etc. Despite the fact that the various agents  involved in the operation of power systems, between them, generate large amounts of diverse data on multiple spatio-temporal scales, current procedures for power system operations are not taking full advantage of this data. Power systems are huge and strongly physically constrained infrastructures that provide a basic service to society. Guaranteeing the secure functioning of power systems is, therefore, an unavoidable requisite with which any potential revision of operational procedures must comply, including those modifications aimed at making a more effective and beneficial use of the available information. The primary objective of this project is, thus, to develop a disruptive methodology that facilitates the smart use of data for power systems operations. This new methodology represents a shift in the manner in which the available information is used to make decisions.
• Dates: 01/01/2018 – 30/09/2021
• Funding Entity:  Spanish Ministry of Economy, Industry and Competitiveness

MLO-MILP - Machine Learning-Aided Optimization Applied To Mixed Integer Linear Programming

• Summary: The development of Mixed Integer Linear Programming techniques is traditionally an active field of research within the Operations Research community. The main strategies designed to solve this type of optimization problem can be divided into exact and heuristic methods. Exact methods guarantee that global optimal solutions are found, although the associated computational time may be quite high. In contrast, heuristic methods reduce the run time, but only find local optimal solutions. Hence, one of the principal challenges in Mixed Integer Linear Programming is to design novel approaches that guarantee global optimal solutions while alleviating the computational time issue. Fortunately, the era of Big Data has brought with it a large amount of data which can be successfully applied to improve existing strategies for Mixed Integer Linear Programming. Indeed, the aim of this project is to develop a Mathematical Optimization framework that  is able to exploit the information provided by tons of data through Machine Learning methodologies, so as to enhance the benchmark methods in Mixed Integer Linear Programming. The link between Mathematical Optimization and Machine Learning will give rise to computational and performance improvements, which will be demonstrated through the quality of the derived theoretical results. Moreover, the power of the new proposed framework will considerably simplify the resolution of real-life problems, thus increasing their applicability and usefulness.
• Dates: 24/05/21 - 24/05/23
• Funding Entity: University of Málaga

Prescriptive Analytics Techniques Applied to the Operation and Planning of Electric Power Systems

• Summary: This project consists in applying novel mathematical techniques related to machine learning and statistics in order to exploit the information that data can provide to solve power system operation and planning problems. The project principally focuses on the development and validation of those algorithms used in typical power system problems: optimal load flow (without and with security constraints) and power grid expansion. In fact, the massive use of data can lead to computational improvements or even improvements in the quality of the solution. The expected results are clearly useful to the power sector, related sectors on a regional level, and/or for emerging technologies.
• Dates: 24/05/21 - 24/05/22
• Funding Entity: University of Málaga

SAND — SmArt distributioN griD simulator

• Summary: Smart grid concept revolves around the power system modernization via the Information and Communication Technologies, which leads to a massive transfer of data among the different agents of the power sector. In addition, the active demand participation and the growing infrastructure based on distributed energy resources are key to drastically change (in the medium- or long-term) the paradigm on how the distribution grids are currently operated. Ensuring a resilient, secure, and reliable operation of the smart distribution grids of electric energy while making use of more sustainable and efficient energy resources is nowadays a challenging avenue of research. Therefore, simulating the operation of smart grids in controlled environments is necessary to: 1) analyze its behavior itself and that of the different players with different business models, 2) identify the challenges associated with the operation of the distribution grids from an economical and technical viewpoint, and 3) adopt emerging business models and their regulations. This project aims to develop a simulation platform for a smart distribution grid including the mathematical models able to capture the behavior of the different agents.
• Dates: 01/09/2018 – 01/09/2019
• Funding Entity: Fundación Iberdrola España

OD2ES - Decision-Making with Big Data: Applications to Renewable Energy Systems

• Summary:  The so-called Information Age has brought about the exponential growth and availability of data. These have allowed us not only to understand our environment, our bodies and our social interactions in ways we could have never imagined before, but also to shape the state of things by developing and manufacturing new systems, products and mechanisms. Humans use data to learn, build, create… and to make decisions. The classical paradigm of decision-making proceeds in a way such that data are primarily used to produce models for the processes that determine how good or bad our decisions are. We then use these models to predict the future behavior of these processes, and finally, make our decisions based on these predictions. In this project, the research team will challenge the classical decision theory and develop a new mathematical framework for decision-making whereby decisions are directly inferred from data, thus averting the need for predicting. That is, under the proposed new paradigm, data are not used to model the processes driving the decision-making, but rather to model the decisions themselves. The mathematics on which the new theory will be built will result from bridging two different disciplines, namely, those of supervised learning and optimization-based decision-making. The power of the new theory will be demonstrated in the solution to exemplary problems arising in what is, most likely, one of the major challenges of our times: the transition to green energy systems.
• Dates: 06/06/2017 – 06/06/2018
• Funding Entity:  University of Málaga

Other projects

Members of the Oasys group are also currently working on various national and international projects in collaboration with other research groups.

Models for Managing Energy Storage Systems based on Batteries, with Domestic and Industrial Applications.

• Researcher: S. Pineda. Principal Investigator: S. De la Torre
• Dates: 01/11/19 - 14/11/22
• Funding Entity: European Union Regional Development Fund

Mathematics for Imperfect Information

• Researcher: J.M. Morales. Principal Investigator: M. Ojeda Aciego
• Dates: 01/11/19 - 14/11/22