RenewableEnergyCo optimizes wind power generation with Machine Learning
RenewableEnergyCo, founded in 2009 is an independent American renewable power company. They are wind and solar powerhouses with operations in the United States, Canada, Mexico, Chile, and Japan. Their global footprint enables them to manage power projects of 5000 MW capacity. They aim to double their power generation capacity by the end of 2020.
RenewableEnergyCo envisions a transitioned world in terms of renewable energy. But the lack of a robust IT infrastructure limited their organizational potential. There were certain other challenges as well which created roadblocks while attaining their goal.
Lack of diagnostic features
RenewableEnergyCo’s wind turbines went through undetected faults. They could not diagnose the failed component nor its cause. To achieve maximum uptime of turbines and generate optimum power output, knowing failure cause and severity was important. But a limited infrastructure could not record whether the fault took place in the blades, brake pad, gearbox, shaft, etc. With this historical data, they had opportunities of predictive maintenance for the wind farms. It would enable planned shutdowns & scheduled maintenance cycles of wind farms and their operations. Instead, they had to experience emergency shutdowns whilst maintaining an expensive spare part inventory. This also created lags in energy transmission. There were categories for energy loss such as turbine outage, BOP issues, climate issues, curtailment, Force majeure, etc. It became difficult to measure and note how much energy was lost and why.
Opportunities for Machine Learning
RenewableEnergyCo had the potential to predict power generation and enhance its value. To do that, they needed to leverage their existing database and implement machine learning algorithms on it. These algorithms would compile all aspects of power generation such as turbine operations, wind and weather history, electricity demands, etc. This would help optimize wind farm operations and predict power output from every asset.
Power forecasting would optimize asset performance, promote operational costs and meeting electricity demands. RenewableEnergyCo could certainly enjoy heightened confidence from utilities and customers.
Multiple data sources
For RenewableEnergyCo, the multiple data sources- SCADA system, OPC Server & an API server created challenges. This affected operational effectiveness. The SCADA system captured turbine data from sensors on various components. The wind farm operations were stored and maintained on an OPC server. The historical data of the wind farm maintenance was stored in an API Server. Together, they stored data but failed to provide a unified view of the windfarm operations. To visualize the combined data became a complex task. It became increasingly cumbersome to make smart decisions for power generation, transmission and distribution.
Lack of accurate data
For asset performance, output and maintenance data, it was challenging to rely on a single data source out of the many. This affected the data-driven decision making. In the case of data loss due to outages, they could not recover the lost data. With asset data lost at 1-minute intervals, it became a challenge to fill in these gaps.
Solution: Cloud-native data warehousing & machine learning based data visualization
RenewableEnergyCo’s chaotic IT infrastructure challenged optimized power generation. This triggered the journey of migration. They eliminated the dependency on on-premise & API servers and migrated the wind farm data to Azure Cloud. This process ensured a robust, scalable, and secure IT infrastructure. This enhanced the quality, speed, and reliability of decision-making. The operations team had 24/7 access to data. Data recovery became easier too.
Earlier, it was an effort-intensive task to collate data together and visualize it. Extracting necessary data to detect asset faults was trouble. It hindered the process for predictive maintenance. RenewableEnergyCo had opportunities to cut costs but ended up losing them. With a cloud-native data warehouse, data visualization became easy. This data was used to train the machines and run algorithms. This ML model contributed towards scheduled maintenance of assets. Planned shutdowns could then be executed and revenue losses could be avoided.
Between the SCADA data, on-premise, OPC and ML data, the best source was chosen. It helped make data-driven decisions to enhance operational efficiency.
With this collaborative solution, RenewableEnergyCo could allocate time and resources towards attaining their goal- working for a better, cost-effective and a greener world.
Want to optimize power generation for your wind farm?
Talk to our ML Expert