Introducing MLOps

Chapter 11. MLOps in Practice: Consumption Forecast

Nicolas Omont

Predictions at various times and geographical scales play an important role in the operation of a power grid. They allow for simulation of possible future states of the system and for making sure that it can safely operate. This chapter will walk through a machine learning model life cycle and MLOps use case for consumption forecasting, including business considerations, data collection, and implementation decisions. Though this particular chapter is focused on power grids, the considerations and particularities of the use case can be generalized to other industrial cases that use consumption forecasting.

Power Systems

Bulk power systems are the backbone of power grids. Also called transmission networks, they form the core of the system that keeps the lights on. These systems are mainly composed of lines and transformers, which are indirectly connected with most producers and consumers through distribution networks that take care of the last few kilometers of transmission. As illustrated in Figure 11-1, only the largest producers and consumers are directly connected to the bulk system.

Figure 11-1. A sample bulk power system, to which only the largest producers and consumers are directly connected

The longer the transmission distance and the larger the energy volume to be transmitted, the higher the voltage used: ...