Operations - Intelligent Monitoring of Infrastructures

Twenty-four-seven monitoring of energy infrastructure, such as power plants, oil rigs or wind farms, is a challenging endeavor requiring expert operators and process engineers. Sensors are applied to equipment to send operating data back to monitoring systems.

Often 100s of variables have to be tuned and alerts are typically set point based, making use of thresholds applied using the expertise of the operator. To complicate matters further, alerts need to be attended to in a timely manner in order to ensure issues do not escalate further.

Optimization of thermal efficiency of a power plant using Machine Learning.

The large amount of historical data collected by sensors makes this scenario a perfect case for optimization using machine and deep learning. Automatic setting and adjusting of set points allows not only for optimisation of cost, but also frees up the time of subject matter experts to concentrate on responding to critical situations.

The Vistra owned Martin Lake power plant in Tatum, Texas, partnered with Mckinsey to deliver such a Machine Learning optimisation approach. In order to optimize the thermal efficiency of the plant, a neural network trained on 2 years worth of plant data was built. Once the model was deployed it was used to generate efficiency recommendations to operators every 30 minutes.

Vistra then proceeded to roll out the approach at a further 26 plants, achieving $23m in savings and 10% of its remaining carbon abatement target for 2030.