CEMIG: up to 46.6 hm³ less water per month in the Três Marias plant schedule, in a study

Up to 46.6 hm³

less water per month in the Três Marias plant study, compared with usual unit dispatch rules

≈ 7×

the useful volume of a plant such as Baguari, at 6.8 hm³

Automatic

correction of the constraints violated in the DESSEM schedule, previously done by hand

Half-hourly

daily generation schedule, with flows and reservoir levels

Every day, each hydropower plant must decide how many generating units to start, how much each one will generate, and how this affects the reservoir and the downstream plants. The schedule must follow the load curve and the demand set by Brazil’s National Grid Operator (ONS) without wasting water:

• Different turbines. Each unit has its own efficiency curve and operating ranges, which change with the net head.
• Cascaded plants. Water released by one plant takes time to reach the next, and the water balance must close at every one of them.
• Many constraints at once. Environmental flow, flow rate of change, operating levels, unit operating ranges, and scheduled maintenance.
• Manual corrections. Manually fixing the constraints violated in the DESSEM schedule takes the team’s time every day.

With HPO, daily scheduling is now simulated and optimized on a faithful model of each plant and of the cascade:

1. Models the plants as they operate

   efficiency curve of each turbine and generator, operating ranges by net head, hydraulic losses, reservoirs, and water travel time between plants.

2. Optimizes dispatch with multiple objectives

   how many units to start, how to split generation among different turbines, how to fit in maintenance, and how to reduce starts and stops, following the load curve and the desired reservoir level at the end of the day.

3. Corrects and delivers the schedule

   checks the physical, environmental, electrical, and ONS constraints, automatically corrects violations, and produces the half-hourly schedule, which the scheduler accepts or adjusts before submission.

HPO integrates with inflow forecasting, machine availability, and reservoir volume systems, runs in the cloud, and serves several teams at the same time.

In the Três Marias plant study, HPO’s dispatch was compared with three usual rules for starting generating units. To generate the same energy, the usual rules used more water: between 5.9 and 46.6 hm³ per month, depending on the rule. The largest value is nearly seven times the useful volume of a plant such as Baguari.

At the plants where HPO operates, analysis of the DESSEM schedule shows where the violated constraints were that the system now corrects automatically. More than half involved environmental flow:

With the system’s corrections, the team no longer adjusts the schedule by hand and receives a schedule that already respects environmental and operating constraints.

• More energy per cubic meter. The combination of units chosen by HPO uses less water for the same generation, preserving the reservoir.
• Environmental constraints respected by default. Environmental flow, rate of change, and levels enter the schedule as rules, with less risk of violation.
• Fast, auditable scheduling. The daily schedule comes out optimized, with corrections logged, and the team decides before submission to ONS.
• Less wear on the machines. Dispatch aims to reduce starts and stops of the generating units.

• You operate hydropower plants with different generating units or cascaded plants.
• Daily scheduling and submission to ONS depend on manual adjustments and a few specialists.
• Environmental and operating constraints make the schedule hard to close in the time available.

We start from your data: we customize the twin, validate the result in scenarios and a pilot, and only then scale. The results of this case reflect the conditions of the plants studied.