Utilities can spend a lot of time understanding the current health of their pumps. The best case is regular multi-point pump tests that track degradation in pump head and efficiency curves, and use of that information to revise pumping policies and maintenance priorities. On the other side, pumps are run to failure. In determining how frequently to test pumps, there is an inherent calculation of the net value of that information -- its benefits from more efficient pump selection, operation, and asset life, minus the cost of the testing program. We aim to increase the net benefits of a pump testing program by reinventing it as an automated software service, reducing the incremental cost of a pump test to near zero while increasing the frequency of testing. We will describe a pure software approach to automation, leveraging standard SCADA instrumentation and telemetry that the utility has probably already paid for. We specifically want to avoid the installation of special instrumentation to measure individual pump discharge flows, as those can be expensive and sometimes impractical due to space limitations.
The first question we answer is "is it possible to accurately estimate pump head and efficiency characteristics using ordinary SCADA data streams?" Results comparing field pumping tests on individual pumps at four different pumping stations, to results obtained independently from the automated services, suggest that automated pump analysis is indeed feasible and accurate. The approach uses historical measurements of suction and discharge head, total station flow, and individual pump status, speed, and power, to estimate a polynomial representation of the head and efficiency curves. The estimation approach minimizes the sum of squared deviations between measured and predicted total station flow or input power, for head and efficiency curve estimation, respectively. Variable frequency drives are modeled using the typical affinity laws.
The second question is "how can updated and accurate pump characteristics be used by operators to increase the efficiency of operations?" The team has worked collaboratively with operational staff to design a user interface that allows an operator to interactively access the automated pump test results, explore the impacts of pump selection, and see suggestions about the optimal pumps to run given a desired operating point. This solution has recently been deployed to a production site that is accessible by operational staff. At the conference, the team will present practical evidence and information collected about how operators are using this new capability, and how it was adapted over time to meet their needs.
The final question is "how can long term trends in pump characteristics be used by maintenance staff to better prioritize and calculate the value of their work?" This ongoing work will determine the best metrics to prioritize maintenance activities, and will be completed by the time of the conference.
