01.7 · Industrial project · 2023
Fuzzy control of process steam generation with a biomass boiler
For Tanin Sevnica, after major changes in production, we returned a wood biomass boiler from manual to automatic control. A fuzzy controller based on operator knowledge coordinates the fuel and primary air supply.
About the project
Tanin d.d. Sevnica produces process steam with a Galleri wood biomass boiler with a rated capacity of about 23 tonnes of steam per hour. The steam is used in tannin production, as a reserve in a steam accumulator and for electricity generation with two steam turbines. After the fuel dosing was renovated, the furnace extended from 4 to 6 zones, furfural production discontinued and a second turbine installed, the existing fuzzy controller no longer worked, so the boiler was operated manually.
From measurements during manual operation, when the operator changed the fuel and primary air supply, we developed process models in Matlab: a state-space model and models with neural networks and ANFIS. On these we designed and tested a fuzzy controller that imitates the operator's work.
The controller has four inputs (deviation and trend of steam production and of furnace temperature), two outputs (change of fuel and primary air supply) and 21 rules. Before commissioning we tested it on a real Siemens controller connected to the model via OPC (hardware-in-the-loop). We also added feedforward control to the control scheme, which reduces the steam draw for the turbine faster when the pressure in the steam accumulator drops. Automatic control has been running since 25 July 2023.
How it went
- Measurements and operator knowledge: manual changes of the influencing variables in production, three hours of measurements.
- Process models: state space (PEM), neural networks (ANN, AANN) and ANFIS.
- Fuzzy controller: design in Matlab/Simulink, transfer to Siemens FuzzyControl++.
- Testing and commissioning: hardware-in-the-loop, then commissioning in production with Sistemi INES.
An analysis of operation in September 2023 showed that steam production was lower than planned, so the controller outputs were at their lower limits. We proposed widening the limits of the manipulated variables and repeating the analysis at lower and higher steam production.
| Type | Industrial project |
| Client | Tanin d.d., Sevnica |
| Duration | June – August 2023 |
| Process | Galleri wood biomass boiler, about 23 t/h of steam |
| Controller | Fuzzy: 4 inputs, 2 outputs, 21 rules |
| Tools | Matlab/Simulink, Siemens FuzzyControl++ |
| Contractors | dr. Nenad Muškinja, dr. Božidar Bratina |