EVALUATION OF SEDIMENT YIELD AND RESERVOIR OPERATION FOR BHAKRA DAM

Abstract

Optimal utilization of available water in the context of growing demand from various sectors such as irrigation, power generation, municipal and industrial water supply and navigation due to population explosion and industrial growth has been recognized since long by the reservoir managers. Recent survey of Indian reservoirs shows that sediment yield from the catchment due to unpredicted landuse changes has been many fold than the sediment inflow considered during the design of the reservoirs. Consideration of sediment yield from the catchment area over the life period of the reservoir in view of the high sediment inflow has become important to evolve future operating policy to maximize the benefits from the water releases for various sectors. To date, empirical, black box and physics based conceptual models have been developed and applied to estimate the sediment yield. Newer techniques such as Artificial Neural Network (ANN) and Fuzzy logic have not been explored enough to overcome the difficulties faced by conventional modelling techniques. Modelling of streamflow and suspended sediment concentration using newer technique is essential to simulate the sediment yield from the catchment more accurately in the absence of measured sediment data. Assessment of performance of reservoir in maximizing the benefits from water releases for future time horizon involves the consideration of revised elevation-area-capacity curves on the account of sedimentation. The revision of elevation-area-capacity curves is done by distributing the simulated consolidated sediment volume in the reservoir. Estimation of consolidated sediment volume in the reservoir requires the consolidated unit weight of sediment. The consolidated unit weight of sediment can be computed by method proposed by U. S. Bureau of Reclamation (USBR). in The performance of reservoir for maximizing the benefit from the releases for irrigation and power can be assessed by developing reservoir operation model by optimization techniques. Dynamic Programming (DP) is the commonly used technique to account the nonlinear behavior of storage capacity and inflow. The curse of dimensionality posed by DP can be avoided by developing optimization model using optimal control theory. The present study is concerned with the development of a methodology to assess the performance of reservoir in maximizing the benefits from irrigation and power by accounting the sedimentation during future time horizon. The study area selected for this purpose is the catchment area of Sutlej Riverup to Bhakra reservoir in India. An ANN model is developed to simulate the daily suspended sediment concentration at Kasol just upstream of Bhakra reservoir by considering the daily rainfall in the catchment and daily streamflow at Kasol as inputs. The input variables are screened through correlation analysis of rainfall and streamflow with suspended sediment concentration. Back propagation feed forward neural network is used for the development of ANN model. The significant antecedent input variables are derived using the statistical properties such as auto-correlation function, partial auto-correlation and cross-correlation function of the time series. The results of ANN model during the calibration and validation process are assessed by statistical performance indices. The performance of the ANN with input variables selected by statistical properties of time series is compared with the ANN developed with input variables selected by trial and error procedure. The same procedure is adopted to develop ANN model for simulating the streamflow at Kasol. The results of ANN models are compared with Multiple Linear Regression models (MLR). The monthly sediment yield at Bhakra reservoir is modelled using ANN by considering the monthly rainfall and flow volume as inputs to the model. The data of rainfall and flow volume for next 25, 50, 75 and 100 years are generated by the time series IV modelling. The sediment yields for the same period are simulated using the best ANN model with the input data as generated series of rainfall and flow volume. The consolidated unit weights obtained by different methods are used to predict the possible range of sediment volume expected to be deposited in the reservoir for the future 25, 50, 75 and 100 years. The sediment volume computed from the simulation of ANN model is compared with sediment volume computed by empirical formula. The sediment volumes computed by different methods are distributed in the reservoir by empirical area reduction method to get the revised elevation-area-capacity curves for future 25, 50, 75 and 100 years. The optimal control theory by Pontryagin's maximum principle is applied for the development of operation model for multipurpose reservoir to maximize the benefits from irrigation and power on yearly basis. The constraints on state and control variables are incorporated in the performance index by using Sequential unconstrained minimization technique. The time dependent data (inflow and evaporation) are expressed as continuous function of time in the form of Fourier series. The resulting reservoir operation problem resembles continuous optimal control problem which is converted into two point boundary value problem by the maximum principle. Gradient technique has been utilized in maximizing the performance index. The differential equations obtained as the necessary conditions for the maximum principle are integrated numerically by fourth order Runge Kutta method. The optimal reservoir operation procedure by Pontryagin's maximum principle is applied with revised elevation-area-capacity curves for the sediment volumes for next 25, 50, 75 and 100 years obtained by various methods to achieve optimal releases and the performance of the reservoir for power generation and irrigation is evaluated. The performance of the reservoir is evaluated by taking observed inflow values from 1987 to 2003.