Improving energy efficiency of individual centrifugal pump systems using model-free and on-line optimization methods

Abstract

Pump systems around the world account for a high percentage of electrical energy consumption. The energy efficiency of these systems can be improved using control algorithms to determine appropriate operating points. The problem that arises is that computed operating points deviate from real operating points. This is due to deviations of the affinity laws, inaccurate information about pump characteristics, and plant behavior. In order to address this problem, we will present model-free optimization methods that can optimize various pump systems during ongoing operation. First, we will classify different pump systems. Based on this categorization, we will then analyze characteristic systems with regard to their energetic optimum. The resulting findings will enable us to identify suitable optimization algorithms. These model-free optimization algorithms take the form of an extremum seeking control in combination with a Kalman filter, the Nelder Mead algorithm, and a dynamic optimization method. After describing these algorithms, we will optimize and validate three sample classes of pump systems using the respective appropriate optimization strategy. The three classes represent a single-pump system, a multiple parallel pump system and a pump storage system. We can see that all optimization strategies achieve the energetic optimum. We can identify savings of between 7.9% and 50%, depending on the system in question. Finally, we will present a best-fit model-free optimization strategy for each class, and system operators can employ this strategy to ensure energy-optimized operation of their specific system.