Transient study of series-connected pumps working as turbines in off-grid systems

Abstract

In the current world economic environment where reducing energy costs is a high priority, it is not surprising that sustainable water and energy management are key topics among the water and electric industry. Specifically, water suppliers, pump manufacturers, and operatives of small hydropower systems, all recognize that pumps working as turbines can be an efficient, simple, and economic approach to generating power and recover the excess of energy in water pressurized systems. In both, variable operating flow rate conditions (ex: presence of different water users) impose the need to augment the range of a proper pump as turbine system. One solution is to associate several pumps as turbines in series to increase the recovered energy. In that context, this work assesses their reliability at transient off-grid conditions by two pumps as turbines connected in series. Each one has an identical self-excited induction generator feeding a three-phase and balanced resistive load. Changes (increase/decrease in values) were applied to the resistive loads and the bank of capacitors at one group only, which allowed examining how changes affect the overall two-group system dynamics. Results show that one change in the first pump as turbine group will significantly affect the other group dynamics. For example, a decrease in load of the first group will affect the flow and head of the other system, until reaching a new equilibrium point. However, as both groups are not mechanically connected, they can achieve different equilibrium speeds. Furthermore, the highest impact occurred in the group where no changes were imposed. The first group maintained its global efficiency value, while the second had its efficiency decreased by about 8%. Similarly, an increase in the capacitance value caused a reduction in its efficiency (lesser but around 2%). Finally, a numerical model was developed and validated through experimental tests to be an applicable prediction model.