Globally, the total cost of water losses amounts to $15 billion with low and middle-income countries being a part of the two-thirds of these water losses. Leaks in the water distribution network, unauthorized consumption, and poor metering costs over $4 billion each year.
With the help of classic machine learning algorithms and geospatial data, Smartterra built an AI-powered operational intelligence system to reduce water losses. The co-founders of Smartterra company include Gokul Krishna Govindu and Giridharan Sengaiah. This technology also ensures water availability in developing cities in India. The company will also help in minimizing the electricity bills of consumers by reducing utilities’ carbon footprint.
Challenges faced by Smartterra
Initially, it was impossible to combine the disparate data sources which included SCADA, CRMs, revenue/billing systems, geospatial data stores, meter data management systems, maintenance records to extract patterns from data. Identification of subtle patterns of loss in both the network and in water consumption was made easy with the help of developed algorithms.
Secondly, the sources contained critical data gaps that need to be filled before feeding into the AI algorithm. The data gaps were filled using the combination of analytical techniques and subject matter expertise. Thirdly, external data sources such as open-source mapping with traffic, satellite imagery, soil analysis, urban activity need to be integrated to gain a 360-degree view of the problem being faced.
The tech behind Smartterra
Deficiencies in the data, identifying connections, and network elements like pipes, pumps, valves, and stores to investigate anomalous behavior is targeted by customized AI-powered analytics and monitoring system. The company offers a comparison to similar consumers with working meters when the artificially intelligent machine tags a faulty meter.
Techniques like SHAP, CORELS, and RuleFit is used by the company to improve the decision-making such that it can either directly build explainable models or explain existing models. Data based on various aspects of operations, like supply, consumption, connection details, geospatial maps, and network information are provided by a unified data model that lies underneath the analytics layer.
The quality of data is ensured by custom data cleaning and pre-processing modules. The nuances in city activity, socio-economics, and local variability are captured by incorporating urban dynamics through land use, satellite imagery, and terrain/soil data. Interpretation of water consumption behavior, demand projections, and network planning is done based on geospatial data.
To design workflow engines, the company leveraged the MEAN (Mongo – Express – Angular – Node) stack such that users can now create tasks for on-ground verification of the given predictions. A full-cycle from data to models to on-field verification and back to data of the AI system is ensured by the workflow engine.
The company aims at helping the operators to reduce water losses in their network and to ensure that the residents of the developing cities receive consistent, high-quality water supply.
