Rosalva Mendoza-Ramírez

Safe operation of aging dam infrastructure requires regular updates to design flood estimates and spillway operating rules. This paper presents a semi-automated methodology for updating design floods for return periods of 10 to 10,000 years, utilizing a suite of open-source software tools. The framework was applied to the El Caracol, Infiernillo, and Villita dams within the Balsas River hydroelectric system in Mexico. The methodology is based on the Engineering Institute of National Autonomous University of Mexico (IIUNAM) procedure, incorporating frequency analysis of daily mean flows and the generation of design hydrographs using the AX and IIHIDRODIS programs. Flood routing was subsequently performed under various operational policies using the Trate, HEC-HMS, and Traven programs to develop safe and effective strategies. Key recommendations include: for El Caracol dam, a staggered discharge policy for floods exceeding a 50-year return period (Tr); for Infiernillo dam, limiting discharges to 6000 m³/s during frequent floods (Tr < 100 years) and up to 12,000 m³/s for extreme events; and for La Villita dam, operating at 6000 m³/s, potentially increasing to 10,000 m³/s under extreme conditions while respecting downstream channel capacity. The proposed sequential automation markedly reduces processing times, in particular during the sequence involved in constructing a design flood. Previously, this task required approximately one or two days; it can now be completed in just a few minutes, demonstrating a significant advancement in the state of knowledge for studies on dam safety reassessment.

Se analizaron las variaciones espaciales y temporales del índice de calidad del agua en las tres capas de estratificación térmica en un cuerpo de agua endorreico de Michoacán, México. Se utilizaron dos métodos, el primero es el desarrollado por la Fundación Nacional de Saneamiento de los Estados Unidos de América, que usa nueve variables, físicas, químicas y bacteriológicas, para la estimación del índice. El segundo usa tres parámetros que son recomendados por la Comisión Nacional del Agua en México. El conjunto de datos analizado es una base de datos pública con información de los años 2012 a 2019 en cuatro sitios de muestreo. Los resultados mostraron que la calidad del agua del lago es buena y que en los meses en los que se rompe la estratificación térmica la calidad del agua en el lago es ligeramente mejor que la registrada en los meses de estratificación térmica.

The guide curves are tools that specify rules for dam operation that allow reducing the chances of unwanted spill or deficit events. In this research, we analyzed the importance of using guide curves and penalizing the deviation of such curves by the occurrence of surplus volume, or deficit, using optimization tools on an objective function that apply stochastic dynamic programming (SDP), guide curves and reservoir operation simulation all together, to obtain and evaluate biweekly operation policies for hydropower generation purposes. The main dams of the Grijalva River were taken as a case study. The optimization tool is based on sequential decision stages (time intervals) to operate a dam system that considers the restrictions on storage volumes, defined by high and low guide curves. An objective function was considered which maximizes energy generation, it uses coefficients to penalize unwanted events of spills, deficits, and operation conditions outside the limits of the guide curves. In conjunction with the agencies that manage the system, three tests with different guide curves were defined. To analyze their sensitivity in the results (mainly events of spillage) one more test was defined for an operation rule without considering guide curves. The system behavior was simulated with the historical record, where five variables were analyzed: total spills and deficit, minimum and maximum storage, as well as the average energy generated. By using the guide curves, the occurrence of spills and deficits was avoided.

Stochastic Dynamic Programming (SDP) has been used to solve reservoir management problems in different parts of the world; specifically in Mexico, it has been used to obtain operating policies that optimize a given objective function. By simulating the operation of the system with a comprehensive model, the behavior of such policies can be accurately evaluated. An optimal policy involves, on the one hand, the selection of the volume of water to extract from each reservoir of the system that guarantees the maximum expected benefit from electricity generation in the long term; and, on the other hand, an optimal policy should reduce the occurrence of unwanted events such as spills, deficits, as well as volumes exceeding the guide curves imposed by the operators of the dams. In the case of the Grijalva river dam system, SDP was applied to determine optimal operating policies considering three alternative guide curves proposed by different agencies; however, since the simulation of the operation of the system under the three alternatives with the historical record of dam inflows found that none of them showed deficits or spills, it was considered necessary to use synthetic series of inflows to increase the stress of the system. Records of synthetic biweekly series of 1000 years were then generated to simulate the behavior of the Grijalva river dam system using the optimal operation policies obtained for each alternative. By stressing the dam system by simulating its behavior with synthetic series longer than the historical record but preserving the same statistical characteristics of the historical series on the synthetic ones, it was possible to realistically evaluate each operating policy considering the frequency and magnitude of spills and deficits that occurred at each dam. For the generation of the synthetic series, a fragment method was used; it was adapted to simultaneously generate the inflow volumes to the two regulating dams (modified Svanidze method), which preserves the statistical characteristics of the historical series, including both the autocorrelations of each series and the cross-correlation. It was also verified that simulating the operation of the dam system with the generated series also preserves the average conditions, such as the average biweekly generation at each dam, which were obtained in the simulations with the historical record. Finally, an optimal policy was obtained (Test 4) by combining the guide curves used in the previous tests. Such a policy attained an average energy production of 474 GWh/fortnight, the lowest average total spills in the system (30,261.93 hm³), and limited deficits (5973.17 hm³) in the long term. This represents a relative increase of 16% in energy generated compared to the balanced historical operation scenario with respect to the few events of spills and deficits.

Stochastic dynamic programming (SDP) is an optimization technique used in the operation of reservoirs for many years. However, being an iterative method requiring considerable computational time, it is important to establish adequate convergence criterion for its most effective use. Based on two previous studies for the optimization of operations in one of the most important multi-reservoir systems in Mexico, this work uses SDP, centred on the interest in the convergence criterion used in the optimization process. In the first trial, following the recommendations in the literature consulted, the difference in the absolute value of two consecutive iterations was taken and compared against a set tolerance value and a discount factor. In the second trial, it was decided to take the squared difference of the two consecutive iterations. In each of the trials, the computational time taken to obtain the optimal operating policy was quantified, along with whether the optimal operating policy was obtained by meeting the convergence criterion or by reaching the maximum number of iterations. With each optimization policy, the operation of the system under study was simulated and four variables were taken as evaluators of the system behaviour. The results showed few differences in the two operation policies but notable differences in the computation time used in the optimization process, as well as in the fulfilment of the convergence criterion.