Identification and Control of Blast Furnace Gas Top Pressure Recovery Turbine Unit

Ping Wu, Chunjie YangISIJ 2012 · DOI

Abstract

Blast Furnace (BF) Gas Top pressure Recovery Turbine (TRT) unit is the most popular energy-saving equipment in iron making process. It is desired to control top pressure of BF precisely and reliably after installing TRT. The dynamic mathematical model of top pressure with TRT is the key point of controller design. However, it is difficult to establish the dynamic mathematical model of TRT from first principles. In this paper, the dynamic mathematical model of No. 1 TRT in Hangzhou Iron & Steel Group Company is first estimated by subspace identification method. For safety reasons, we adopt closed-loop identification method. Then a model predictive controller is proposed based on the identified model. Results show that the identified model can capture the dynamic properties of top pressure process accurately. Moreover, the model predictive controller gives excellent performance.

Introduction: The Challenge of Blast Furnace Efficiency

Iron and steel production is incredibly energy-intensive, with blast furnaces accounting for nearly 40% of the industry's energy consumption. To improve efficiency, many plants install a Top Pressure Recovery Turbine (TRT). This device sits downstream of gas-cleaning equipment and converts the high-pressure, high-temperature blast furnace gas (BFG) into electricity—recovering up to 20% of the power needed to run the furnace's air blowers.

Schematic of the Blast Furnace Gas Top Pressure Recovery Turbine (TRT) system.

However, the TRT's operation directly affects the "top pressure" of the blast furnace. For the furnace to operate safely and efficiently, this pressure must remain extremely stable. Establishing a mathematical model to control this pressure is notoriously difficult because "first principles" (physics-based) equations are too complex for real-time industrial use. Furthermore, performing "open-loop" experiments on a live furnace is impossible due to safety risks.

Methodology: Closed-Loop Identification and MPC

The authors propose a two-step data-driven approach to overcome these modeling hurdles:

  1. Closed-Loop Subspace Identification (PBSID_OPT): Instead of using theoretical physics, the authors used actual operational data to build a state-space model. They chose the Optimized Predictor-Based Subspace Identification (PBSID_OPT) method. Unlike traditional subspace methods that can become biased when data is collected while a controller is active, PBSID_OPT is specifically designed to handle closed-loop data accurately.
  2. Model Predictive Control (MPC): Once a reliable 3rd-order linear model was established, they designed an MPC. This controller predicts future states and optimizes the movement of the turbine's 1st-stage stator blades to maintain the pressure set point.

Data and Experimental Setup

The study utilized real-world industrial data from the No. 1 TRT at Hangzhou Iron & Steel Group Company (furnace volume: 1,250 m³).

  • Data Volume: 7,200 data points collected at 1-second intervals over 2 hours.
  • Split: The first 3,600 points were used for model estimation (training), and the remaining 3,600 points were used for validation (testing).
  • Variables: The input variable was the opening percentage of the 1st-stage stator blades, and the output variable was the top pressure (kPa).

Key Insights

  • Optimal Model Order: Using Singular Value Decomposition (SVD), the authors identified that a 3rd-order system was the optimal complexity, balancing dynamic accuracy with computational efficiency.
  • Safety-First Modeling: The paper demonstrates that high-fidelity models can be extracted through closed-loop identification without disabling existing safety systems, a critical requirement for industrial applications.
  • Linear Approximation: While the process is inherently nonlinear, a well-identified linear state-space model proved sufficient to describe the dynamic characteristics for precise control.

Results and Performance

The proposed method yielded significant improvements in both modeling and control:

  • High Modeling Accuracy: The identified model achieved a Coefficient of Determination ($R^2$) of over 95.5% on both estimation and validation data.
Comparison between predicted and actual top pressure for (a) estimation data and (b) validation data.
  • Superior Pressure Stability: In simulations based on real plant data, the MPC kept pressure fluctuations within ±1 kPa most of the time. The controller successfully tracked "pulse" changes in the set point, demonstrating robustness to operational shifts.
  • Energy Recovery: The system managed a TRT plant producing between 5,625 kW and 6,100 kW of power while maintaining high stability.

Limitations and Future Outlook

While successful, the study notes several areas for future development:

  • Operational Phases: The current model focuses on the "normal running phase" and does not account for the volatile "start-up" phase.
  • Adaptive Modeling: Future work could explore adaptive models that update over time to account for equipment aging or changes in raw material grades.
  • MIMO Expansion: Expanding the Single-Input Single-Output (SISO) focus to include variables like gas temperature or flow rate could provide even tighter control.

Summary updated

Abstract

Blast Furnace (BF) Gas Top pressure Recovery Turbine (TRT) unit is the most popular energy-saving equipment in iron making process. It is desired to control top pressure of BF precisely and reliably after installing TRT. The dynamic mathematical model of top pressure with TRT is the key point of controller design. However, it is difficult to establish the dynamic mathematical model of TRT from first principles. In this paper, the dynamic mathematical model of No. 1 TRT in Hangzhou Iron & Steel Group Company is first estimated by subspace identification method. For safety reasons, we adopt closed-loop identification method. Then a model predictive controller is proposed based on the identified model. Results show that the identified model can capture the dynamic properties of top pressure process accurately. Moreover, the model predictive controller gives excellent performance.
BibTeX
@inproceedings{wu2012identification,
  author = {Ping Wu and Chunjie Yang},
  title = {{Identification and Control of Blast Furnace Gas Top Pressure Recovery Turbine Unit}},
  booktitle = {ISIJ},
  year = {2012},
  doi = {10.2355/isijinternational.52.96},
}
Figure