DESIGN OPTIMUM PROPORSIONAL-INTEGRAL CONTROL IN INTEGRATED BOILER-HEADER DURING PARTIALLY TRIP WITH GENETIC ALGORITHM, FLOWER POLLINATION ALGORITHM, AND DEEP REINFORCEMENT LEARNING
Realibility and efficiency from steam power plant is growing as the development in knowledge of control techniques and optimization methods. The Integration between two boilers and one steam header as a steam reservoir and two turbines as a power generator, makes steam power generation more relia...
Saved in:
| Main Author: | |
|---|---|
| Format: | Theses |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/69793 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Realibility and efficiency from steam power plant is growing as the development in
knowledge of control techniques and optimization methods. The Integration
between two boilers and one steam header as a steam reservoir and two turbines
as a power generator, makes steam power generation more reliable. In this study,
a simulation was carried out with the condition of one boiler trip, the pressure on
the header would decrease along with the loss of steam rate from the tripping
boiler. Boilers that are still running will increase their steam production to
compensate for the lost steam rate so that there is no reduction in the power
produced by the turbine.
The mathematical model of the boiler-turbine system used is a nonlinear model of
the boiler-turbine system that was introduced by R.D. Bell and K.J. Astrom in 1987
which is consisting of three inputs and three outputs. The three inputs are the fuel
flow rate, the steam rate, and the feed water rate to the drum, while the three outputs
are the boiler steam pressure, the power generated by the turbine, and the deviation
of the water level in the steam drum.
This research was conducted based on model of a steam power plant in Riau using
a Proporsional-Integral controller. These PI controller are usually tuned manually
by experienced people and require a long observation time. In this study, the PI
controller was optimized using the genetic algorithm (GA), flower pollination
(FPA) algorithm, and deep reinforcement learning (DRL), specifically the deep
deterministic policy gradient (DDPG) algorithm to optimize the PI controller
automatically.
From the simulation results, it was found that the flower pollination algorithm has
a very good test time of 1 minute to perform optimization with the results of steam
pressure control and power control being generated by 5% overshoot with a
settling time of 6.3 seconds and 1.9 seconds, respectively. while the water level
control the best optimization method is DRL with an overshoot of 9.3% and a
settling time of 47.6 seconds. |
|---|