The Tüpraş experimentation aimed to improve the energy efficiency and reduce the environmental impact of the Tüpraş refinery in Izmit, Turkey, in which 8.5 million tons of crude oil was processed in 2010.
The Izmit refinery’s power systems produce electricity and steam (with four different types of pressure) to provide utilities to the refining processes. With the recent connection to the Turkish electricity grid, the refinery now has the opportunity to purchase and sell electricity from and to the grid.
The general methodology of the Tüpraş experimentation was composed of five phases:
Knowledge and data transfer from Tüpraş to Artelys
- Workshop organized between Artelys and Tüpraş experts
- Description of the detailed characteristics of the energy assets
Calibration of asset models based on historical data
- Generation of a formal representation of the asset physical behavior and operational constraints
Replay of historical strategy (data reconciliation)
- Generation of a modified historical management strategy that is consistent with the calibrated mathematical representation
- Generation of an evaluation of historical costs
Optimization of management strategies / asset portfolios
- Generation of optimized and practical management strategies
- Generation of an evaluation of the optimal costs
- Generation of KPIs of various investments plans
- Implementation of the new strategies
- Evaluation of actual gains obtained with the new strategies
This study was supported by the development of the CitInES tool, which leverages state-of-the-art optimization technologies to provide the most advanced functionalities for detailed analysis of the dynamics of complex industrial systems, such as Izmit’s power plant.
In particular, the CitInES tool:
Models large-scale industrial systems with complex technical/operational constraints:
- 9 sources of energy: natural gas, refinery gas, fuel oil, electricity, water, LP steam, MP steam, HP steam, VHP steam
- more than 10 different units including boilers, steam turbines, gas turbines, and letdown stations
- 1-year time horizon, with 1 hour granularity
- Use of linear and piecewise-linear yield models
- Minimum/maximum capacities (varying over time and depending on the type of fuel used)
- Minimum/maximum duration in ON/OFF states
- Safety reserve constraints
Replays historical data while ensuring that the energy material balances and asset yield models are properly enforced (data reconciliation)
Optimizes management strategies / asset portfolio
- Generation of optimal strategies
- Generation of practical strategies defined as script describing the operator’s decision sequence
All data acquisition, modeling, and software development tasks have been performed, and the project is now focused on the study of four different questions asked by Tüpraş:
- When and how should VHP steam or HP steam be produced to improve global energy efficiency?
- When and how should fuel gas and fuel oil be used as boiler feedstock to improve the trade-off between costs and pollutant emissions?
- When and how should import/export from/to the electricity grid be used to improve utilization of existing assets and reduce overall costs?
- Which investment plans (gas turbines or renewables) will better help optimize CO2 emissions saving and RoI taking into count volatile energy prices and demand?
Simulation results showed an approximate 7% decrease in cost and polluting emissions may be obtained with no need for expensive investment projects or use of the electricity grid markets. Such gains are solely obtained by the design of new practical asset management strategies that remain very simple to implement in the field.
Further simulation results showed that savings of the same order of magnitude can be obtained thanks to interactions with the electricity markets.
After studying the four scenarios, the experimentation of the proposed management strategies began at the Izmit refinery by the end of 2013.