Dynamic Stochastic Modeling for Optimization of Environmental Measurements

Abstract

Minimization of energy consumption of environmental measurement systems is important to ensure their extended operational lifetime and low maintenance cost. This needs to be realized without sacrificing on data quality. One possible way to achieving this is the use of energy-aware sampling techniques such as adaptive and event-triggered sampling. In this work, new methods based on these sampling techniques have been developed. The first method produces stochastic models that accurately predict missed and future data with minimal energy. The method also determines the optimal sampling interval. The second method utilizes new type of event-triggered mechanism that adjusts sampling interval so that it adapts to the changes in measurement data. Algorithms have been developed and all methods demonstrated using field data. Obtained results have been thoroughly analyzed from the perspective of approximation error and energy savings. Models have been validated and favorable results obtained. High R-squared values and low values of mean square normalized error have been obtained. Battery lifetime is extended by more than 87% when sampling interval increases from 15 to 30 seconds. Furthermore, about 45% daily savings of energy consumption of analog-to-digital converter has been achieved in a case study analysis involving the new algorithm, an ADC and field data.