Operating Rules Derivation of Jinsha Reservoirs System with Parameter Calibrated Support Vector Regression

A Bayesian model averaging method for the derivation of reservoir operating rules

“…LS-SVM converts a quadratic optimization problem into a system of linear equations (Okkan and Serbes, 2012). It is vital that LS-SVM can find a linear function f ðXÞ ¼ W Á /ðXÞ þ b that best interpolates the training points, where W is the coefficient vector and /(X) is a projection that maps the original independent variable to the ''feature space'' (Ji et al, 2014;Karamouz et al, 2009;Suykens et al, 2002).…”

“…These include linear regression (LR) (Liu et al, 2011bYoung, 1967), fuzzy models (Russell and Campbell, 1996), genetic programming , two-dimensional surface models (SURF) (Celeste and Billib, 2009;Celeste et al, 2005), Bayesian networks (Malekmohammadi et al, 2009) and support vector machines (SVMs) (Ji et al, 2014;Karamouz et al, 2009;Suykens et al, 2002;Suykens and Vandewalle, 1999). Piecewise linear regression (PL-REG), SURF http://dx.doi.org/10.1016/j.jhydrol.2015.06.041 0022-1694/Ó 2015 Elsevier B.V. All rights reserved.…”

“…By analyzing the optimal trajectory, Liu et al (2014) concluded that LR operating rules were suitable for the hydropower operation of China's Three Gorges Reservoir, and implemented an uncertainty analysis for future parameter values. Celeste and Billib (2009) determined that the SURF model was the best in terms of overall performance for the Epitácio Pessoa Reservoir, whereas Ji et al (2014) proposed SVM operating rules for the Jinsha Reservoirs system, demonstrating the effectiveness of SVMs for such scenarios. Although the above LR, SURF and SVM operating rules have been successfully used, they are black models and should have model uncertainty.…”

Total output operation chart optimization of cascade reservoirs and its application

“…Due to the different operation efficiency of water turbines and tight water volume and hydraulic relationship between Three-gorge and Gezhouba hydropower stations, Ma et al [20] proposed two operation rules which were applied to improve optimization model with more exact decision and state variables and constraints. Ji et al [21] used Support Vector Regression (SVR) to derive optimal operating rules, parameters in SVR model were calibrated with grid search and cross validation techniques to improve the performance of SVR. By considering both generalization and regression performance, the trained SVR model overcomes local optimization and over fitting deficits.…”

A New Reservoir Operation Chart Drawing Method Based on Dynamic Programming

“…So, the validity and rationality of the proposed method are further verified by the simulation results.Energies 2018, 11, 3355 2 of 17 rules are needed because of the inflow uncertainties and forecasting level [12,13]. It is realistic to get the satisfactory solution via pre-obtained operation rules [14,15].Reservoir operation rules are varied [16][17][18], and an often-used rule is the reservoir operation chart nowadays. A reservoir operation chart is the graphical result of operation rules, and it contains three main kinds of operation curves and operation zones.…”