Hybrid methods for reservoir operation rule curve determination considering uncertain future condition

In this research, simple and practical methods have been proposed to determine the reservoir operation rule curves considering uncertain water inflow values at future conditions based on the concept of comparative reservoirs management. For this purpose, two methods are proposed. In the first method named His-Inf, by considering the historical water inflow values, a mathematical optimization model is proposed for defining the reservoir operation rule curves determination problem and solved using nonlinear programming (NLP) method. However, in the second method, three different approaches named Pre-Inf1, Pre-Inf2 and Pre-Inf3 are proposed. In the Pre-Inf1 method, at first, water inflow values into the dam reservoir are predicted using neural network (ANN) model. Then, the predicted water inflow values are used to determine and update the reservoir operation rule curves using NLP method. In the Pre-Inf2 method, the ANN model is used to simulate and update the constant coefficients of the reservoir operation rule curves considering the water demand, water inflow values into the reservoir, water release values form reservoir and time index (T) as data set. For this purpose, two different framework and structure are proposed for sorting and providing data set leading to two different cases. Finally, in the Pre-Inf3 method, at first, water inflow values, reservoir storage volumes, water demand values and time index (T) are used to simulate and predict water releases values from the reservoir. Then, the linear regression method is used to determine the reservoir operation rule curves and policies. Here, both static and dynamic ANN models are used for simulating and predicting the water inflow values and reservoir operation rule curves and policies. In order to investigate the performance of proposed methods, here, the Sefidrood dam reservoir is considered as a case study in which it is located at the intersection of the QezelOzan River and Shahroud River near the city of Manjil in the north of Iran. Finally, the obtained results are presented and compared by calculating the reliability, residency, vulnerability and sustainability indexes. Comparison of the result indicates the superiority of the proposed Pre-Inf3 method for determining the reservoir operation rule curves at uncertain future condition in which the results of the His-Inf, Pre-Inf1 and Pre-Inf2 methods are almost similar. In other words, the sustainability index of His-Inf, Pre-Inf1, Pre-Inf2 and Pre-Inf3 are 63%, 63%, 63% and 74%, respectively, in which the corresponding value of Pre-Inf3 is 17.5% bigger than other proposed methods.

Automated summary

In this research, two practical methods are proposed to determine reservoir operation rule curves considering uncertain future water inflow values. The methods involve historical data and predicted data using artificial neural networks. The results are compared using reliability, residency, vulnerability, and sustainability indexes, and the Pre-Inf3 method is found to outperform the other methods.