A two-state deterministic dynamic programming model for optimising the joint operation of Mangla and Tarbela reservoirs in Pakistan

• Main Author: Khan, Asim Rauf
• Language: English
• Published: 2009
• Online Access: http://hdl.handle.net/2429/4620

A two-state deterministic DP (Dynamic Programming) model is developed to derive the optimal reservoir operation policy for the Mangla and Tarbela reservoirs in Pakistan. The analysis is carried out with two separate objective functions, (1) maximisation of energy generation while treating the irrigation demands as constraints, and (2) maximisation of combined benefits from energy production and irrigation water supply in monetary terms. Historic data for a period of five years (1985-90) has been used in this research. The Mangla and Tarbela reservoirs are built on the Jhelum and the Indus Rivers, respectively, in northern Pakistan. Both the reservoirs are multipurpose reservoirs and are operated on a ten-day time step. Water is released from these reservoirs to meet irrigation demands of the agriculture sector as a first priority and generate electricity as a second priority. The maximum live storage capacity of Tarbela reservoir (9.986 MAF) is almost twice as that of Mangla reservoir (5.365 MAF). The current maximum plant output at Tarbela (3500 MW) is more than three times of that at Mangla (1000 MW). Two of the four irrigation regions served by these reservoirs are common to both Mangla and Tarbela, which emphasizes the need to operate the two reservoirs in conjunction with each other in order to maximise benefits from irrigation and energy production. The optimisation results from the DP model are compared with the actual operation of both the reservoirs during the period 1985-90. The model shows an increase in energy production over actual energy production during the same period. However, the model shows deficits in irrigation water supply in the months of May and June, which are critical from the point of view of irrigation, when 'maximisation of combined benefits in monetary terms' is used as the objective function. This problem can be overcome by assigning a greater monetary value to benefits from irrigation. The important characteristic of the model, when run using this objective function, is that it maintains a more or less constant discharge through the turbines during most of the one year period of operation which is quite important from the point of view of energy generation. When run using the objective function of 'maximising energy production with constraints on irrigation water supply', the model not only gives higher energy production but also deals adequately with the irrigation demands. The results also show that more water can be released for irrigation from the reservoirs during early Kharif period (Apr-Jun) because both reservoir can fill to their respective maximum conservation levels during the monsoon season (Jul-Sep). This would also help in flood mitigation by providing more storage and reducing flood peaks. The model can be used, with some modifications, for optimising the real-time operation of Mangla and Tarbela reservoirs. These modifications would involve finer quantisation of the state variables (reservoir levels), determining the limits on minimum discharge outflows necessary to maintain suitable turbine efficiency, criteria to prevent salt water intrusion and meet the requirements of hydro projects downstream of Mangla and Tarbela, and consideration of the losses due to evaporation.