Design techniques for a robust wireless sensor network-based automatic weather station network
MetadataShow full item record
The operation of Automatic Weather Stations (AWSs) and especially those based on Wireless Sensor Network (WSN) technology is often disrupted by harsh environmental conditions. These conditions include dust, other extreme weather conditions, and human-related challenges among others. Hence, weather data losses and errors. These unfavorable conditions limit the efficiency of AWS components, hence limiting the life of the AWS sensors. Moreover, the cost of acquiring and maintaining the AWSs is still high for Uganda, a developing country, hence the small number of AWSs available in the country. The major aim of the study was to design mechanisms for improving the robustness of WSN-based AWSs. First, the study evaluated the performance of a WSN-based AWS (first-generation prototype), identified its challenges and made recommendations. The recommendations provided guidelines for a refined secondgeneration prototype. The first refinement was the design of an autonomous wireless sensor node application, which employs self-healing, an improved user-interface design and self-adaptive duty cycling and self-configuration to prolong the life of the wireless sensor nodes and improve the AWS robustness. Secondly, the study led to the development of condition monitoring techniques for WSN-based AWSs. Condition monitoring provides timely and reliable information for preventive maintenance and to minimize AWS down time. The condition monitor consists of a data receiver, analyzer, problem classifier and visualiser / reporter. The study proposes three anomaly detection techniques including outlier identification, using SPREDs, a new algorithm and observing anomalies in data correlations. Using a queuing algorithm, the data receiver accepts an infinite number of connections from AWSs at speeds as low as 1 millisecond without data losses. We attained a Central Processing Unit utilization of up to 75.6% at a rate of 5ms. Hence, robust condition monitoring techniques, supporting centralized monitoring of an infinite number of AWSs, which reduced the AWS downtime. The worst wireless sensor node improved its availability from 72.35% to close to 100% with the application of autonomous designs in the sensor node application. The study recommends adoption of AWSs with autonomous capabilities, coupled with improved mechanisms of AWS condition monitoring. In so doing, meteorologists shall lower operational costs, while improved preventive maintenance will reduce AWS downtime and data losses.