Advanced design and operation of Energy Hub for forest industry using reliability assessment

• Published in: Applied thermal engineering 2023-07, Vol.230, p.120751, Article 120751
• Main Authors: Talebjedi, Behnam, Laukkanen, Timo, Holmberg, Henrik, Syri, Sanna
• Format: Article
• Language: English
• Subjects: Forest Industry; Long-short-term memory (LSTM); Markov chain; Reliability & Availability analysis; Thermo-mechanical pulping
• ISSN: 1359-4311
• DOI: 10.1016/j.applthermaleng.2023.120751

•Reliability assessment results in more efficient design and operation of the Energy Hub.•Results show a robust design that isn't impacted by unsupplied demand penalty rates.•Integrating thermoeconomic and reliability assessment reduce the total cost by 14.5–28%.•Long-short-term memory models are powerful tools for industrial heat prediction.•Bidirectional Long-short-term memory appears to perform better than other models. A large part of the refining heat production in the thermomechanical pulp mill can be recovered to supply the paper machine heat demand. This study introduces a novel approach for the heat integration of a thermomechanical pulp mill and paper machine using Energy Hub. An Energy Hub consisting of a steam generator heat pump and the electric boiler is integrated with the thermomechanical pulp mill to provide the heating demand of the paper machine. The advanced cost-efficient design and operation of the Energy Hub are investigated in this research by integrating thermo-economic analysis, reliability & availability assessment, and load profile prediction. The thermo-economic analysis combines economics and thermodynamics, which is necessary for energy system unit commitments. Reliability assessment will lead to more accurate modeling of real-life system operating conditions since system components' availability is considered in the design process. Load profile prediction estimates the Energy Hub load for the next hour, which helps with the optimal operation of the Energy Hub. Different state-of-the-art long-short-term memory (LSTM) neural network models have been developed to achieve the best time series model for refining heat prediction in the thermomechanical pulp mill. Results show that all the time series models are effective for refining heat prediction, while Bidirectional LSTM appears to perform better than others with the correlation coefficient and root mean square error of 0.9 and 0.15, respectively. In addition, the proposed Energy Hub design approach is compared with the conventional design method. The proposed design method offers a robust design that isn't impacted by unsupplied demand penalty rates. Depending on the penalty rates, the total system cost could decrease by 14%-28% utilizing the proposed design method.