Feedback Rules for Operation of Pumps in a Water Supply System Considering Electricity Tariffs

Abstract

The cost of energy used for pumping water constitutes a large proportion of operational expenditure for a water utility. Energy saving measures in water supply systems can be realized in different ways, by design of the system to be energy efficient, by proper maintenance of equipment especially pumps and by optimal control of the system. The cost of the pumping is a product of energy consumption and an electricity tariff. The energy can be reduced by pumping less water, lowering the head against which the water is pumped and by operating pumps near peak efficiency. The cost can also be reduced by re-scheduling the pumping from expensive to cheap tariff periods. Typically the real time control for time varying tariffs is implemented in a predictive control fashion, in which an optimal time schedule is calculated ahead over 24 hours period by a solver and recalculated at regular intervals e.g. 1 hour. In order to operate the scheme in real time the solver must be sufficiently fast and this may not always be possible for big water supply systems. In this paper a method to synthesize feedback control rules is proposed taking into account a time varying tariff. The rules are calculated off-line and then implemented in local PLCs or in a control room. Once the rules are implemented the response to the changing state of the water system is instantaneous. In this paper the feedback rules are calculated by a genetic algorithm. Each pump station has a rule described by two water levels in a downstream reservoir and two values of pump speed, for each tariff period. The lower and upper water levels of the downstream reservoir correspond to the pump being ON or OFF. The approach has been applied to a large scale water supply system and compared with the traditional time schedule approach. The achieved cost for the feedback control is only slightly higher than that for the time schedule approach. However, the feedback control by its nature is more robust and performs well in the presence of uncertainty in water demands and in inaccuracy of hydraulic models.