About our Client

Client is a state owned Indian hydel power generation company operating three power plants with generation capacity between 30 MW to 240 MW in India. These power plants have been operational since decades using equipment’s from different manufacturers. Plant’s operational team use electronic hardware deployed at plant site to monitor plant and asset (transformers, generators, turbines) parameters like voltage levels, current readings, pressure levels, fan speed & vibrations at various places. These sensor devices are localized and do not exchange data to a central place to enable centralized monitoring, analytics or various AI workloads.

 

Business Objective

As part of digital transformation journey client is looking to build cloud based IOT platform to enable

  • Centralized data collection of various sensor data in a data lake environment
  • Real time monitoring of plant & asset parameters i.e. output power generated, pressure, voltage & current levels.
  • KPI reporting for internal stakeholders at various levels.
  • Power advanced AI workloads like predictive maintenance in future.

 

Solution

After discussion with client’s technology team and evaluating popular cloud vendors, AWS was chosen as a platform of choice to develop the solution. We then worked with AWS technology team to develop technology architecture comprising data lake environment on S3, Kinesis streams to ingest incoming sensor streams, lambda functions to pre-process sensor data before storing in S3 and athena queries for reporting. Entire solution was developed in following steps

  • Different plant assets (transformers, turbines, generators) were equipped with sensors to enable real time monitoring. These sensors were also configured to send data to cloud via centralized local gateway device. Total of 123 sensors were integrated across 3 transformers, generators, transformers each plus a single water conducting system.
  • Ingestion of streaming data through Kinesis data stream and pre-processing & enrichment through lamda functions.
  • Data from kinesis streams was stored in S3 in Json and parquet format. Data in S3 was aggregated using athena queries to compute KPI’s and other operational metrics. These metrics were stored in relational database for downstream BI applications.
  • Web based application with multi-user & role-based access to enable information access for different audience. Plant’s team wanted to monitor real time data & historical data points to observe & study any anomalies in asset’s health. Management wanted to review each power plant’s output generation, downtime.

 

Outcome
  • In a month’s time we went live with the IOT platform across single power plant. This enabled operational team real time access to the asset data, historical trends & patterns. Management could also view metrics on output power & downtime.
  • Platform is now being expanded to two more power plants.