Fuzzy Interval-Based Fault Detection for Water Consumption Profiles From Isolated Communities

The availability of water consumption data is a significant concern regarding the design of water management systems. Furthermore, when the control system relies on model-based predictive strategies, this data becomes essential to generate a predictive model and achieve an accurate controller design. Unfortunately, this type of data for isolated communities is rarely available online, making it more difficult to develop research related to this topic. Therefore, this work proposes a Markov process-based methodology to generate a synthetic water consumption profile. Based on this new dataset, a fuzzy prediction interval model is implemented to study its accuracy when modeling future water consumption data and its corresponding uncertainty. Finally, an interval-based fault detection method is implemented using the information provided by the fuzzy prediction intervals in a simulated case where abnormal behavior is introduced in the synthetic water profile. The simulation results reported in this work show the effectiveness of the proposed strategy for generating a new synthetic water profile that re-sembles the behavior of previous works. Moreover, the results confirm the good performance of the fuzzy prediction interval for correctly approximating the profile dynamics by reaching the expected coverage performance for different prediction steps ahead. Regarding the implementation of the interval-based fault detection algorithm, the reported results show that the number of false negative events can be reduced by increasing the prediction steps considered for the interval evaluation.