Research

Methodology to generate optimal operating sequences in electricity generation plants to achieve a balance of the fluctuations of the network due to the contribution of renewable energy sources

In electricity generation systems, one of the frequent challenges is how to achieve a balance between electricity supply and demand. This balance is particularly important because producing too much or too little power can damage electrical appliances or even cause a blackout.Since demand is not constant, an electricity generation plant is designed to adjust its electricity production to changes in demand. The demand for electricity is influenced by factors such as the activities of people and industries and by weather conditions. As renewable generation plants increase, the levels of variability in electricity production increase. This variability causes the operations of traditional generation plants to be more complex, due to the fact that these plants are forced to provide the missing amount of electricity when variable renewable energies cannot satisfy the demand.

This research focuses on the generation of optimal sequences of valve operations that take the plant from an initial state to a final one in less time to be able to adjust electricity production according to changes in demand and changes in power generation. solar or wind type.

Methodology for the development of models based on data that allow more precise forecasts of electricity demand.

The impact of this research lies in better forecasting models for electricity consumption, which will make it possible to avoid overestimation or underestimation of electricity demand. The overestimation causes an excessive amount of electricity to be purchased and supplied to the system, causing power balance issues that can damage power generating equipment. Underestimation, on the other hand, leads to risky power system operation by restricting electricity production, which can lead to load outages that directly affect electricity users.

As part of this research, we developed a method for generating forecasting models based on association rules (association rules are useful for describing a model in terms of cause and effect). The proposed approach allows predicting electricity demand two hours in advance in 8 15-minute periods. The data set comes from a representative cargo zone in Mexico, which is a 15-minute cargo demand measure. The proposed method aims to estimate the prediction intervals using an artificial neural network as a model and to adjust the prediction intervals through an analysis based on association rules. The prediction intervals are constructed using Artificial Neural Network models and are subsequently adjusted by means of rules obtained with the a priori algorithm. The proposed approach was compared with the ARIMA model and a persistence model. The proposed model shows to have a better performance in all the evaluation metrics of the prediction interval.

CBR system for the selection and design of public policy during and after COVID-19

The goal of the project is to support the federal government and state governments through a computer system that suggests solutions to specific situations related to COVID-19. Specifically, this project, which is part of the THINK TANK POSTCOVID-19 initiative, consists of developing, implementing and evaluating a case-based reasoning system that allows the reuse of policies for making decisions related to COVID-19. In such a system, past and current policy descriptions are stored in a knowledge base. Once ready, the system will allow the user to search for cases with characteristics similar to a new case to apply solutions from an old case to the new one. For this, an information model has been developed that allows representing information on policies with various impacts on both health and the economy, the environment, etc. The knowledge base is continually expanded and refined by users.

Reverse engineering of 4D materials

Unlike the conventional materials development process, the Reverse 4D Materials Engineering project aims at the rapid development of high-performance materials. The objective of the project is to develop methods and tools that generate the optimal microstructure structure with minimal computational effort.

The project started in 2012 and it brings together experts in the areas of materials science, applied mathematics and optimization. The project leader is Prof. Hiroyuki Toda, director of the "3D / 4D Structural Materials Research Center" at Kyushu University. My role in the project has focused on the development of a methodology based on surrogate models that by replacing the finite element models each simulation is reduced from hours to seconds, making possible the optimization towards the design of the microstructure of the material. The challenge is to use as few simulations as possible to build the substitutes.

As a starting problem, we focus on the identification of material properties based on experimental results of nanoindentation. The problem seeks the identification of parameters with a reduced number of finite element simulations. The methodology is based on the construction of a surrogate model using artificial neural networks. However, unlike other schemes that use experimental designs with a high number of simulations, the developed methodology introduces a sampling scheme that identifies the next point in the search space during the optimization process. In other words, the sample space is built in an interleaved manner at the same time that the surrogate model is built. Optimization is achieved through the differential evolution algorithm to identify the parameters of the material that make the nanoindentation curve obtained with the surrogate model coincide and the original curve obtained with experimental data.

Reuse of past accident information

There is a great need to reduce accidents in the chemical and nuclear industries. For example, PEMEX reports that around 21 deaths occur per year at its facilities as a result of an average of 153 accidents per year. In the wake of critical accidents such as those in Chernobyl, Flixborough, Seveso, Bhopal, San Juanico, Fukushima and Texas City, companies have been concerned with maintaining and exploiting information on past accidents to avoid similar events (Batres et al., 2014 ). However, many small and medium-sized companies do not have enough data.

This research work will focus on methods based on ontologies and data visualization methods to improve the efficiency of the extraction of accident information from the process industry. Ontologies provide computer-actionable semantic definitions that define classes and subclasses of things, the possible relationships between them, and axioms that restrict the meaning of objects. Visualization methods that will be based on a "decision center" will allow manipulation of data interactively and provide an immersive and visually stimulating collaborative environment.

Therefore, an intelligent database and collaboration system will be developed. The proposed method will be evaluated by analyzing its ability to find incidents and reducing false negatives and false positives. The result of the project is a set of methods and tools for the identification, classification and formalization of knowledge associated with accidents in process industries that can be used to improve the development, maintenance and retrieval of information from databases of accidents, including a new comprehensive approach to the efficient extraction of knowledge of past incidents in large databases.