Application of Machine Learning in Thermodynamic System Modeling and Optimization

Li Wang, Chang Liu, Ran Huo, Xin Lü, Yue YangInternational Journal of Heat and Technology 2024 · DOI

Abstract

With the growing demand for energy and increasing environmental concerns, accurate modeling and efficient optimization of thermodynamic systems have become key research areas in engineering. In recent years, the application of machine learning techniques in thermodynamic systems has demonstrated significant potential, greatly improving modeling precision and optimization efficiency. However, existing methods face challenges in handling temperature delay phenomena, limiting the accuracy and effectiveness of system modeling and control. This paper introduces a temperature delay identification algorithm to enhance the accuracy of delay representation. Additionally, it develops a predictive control model for thermodynamic systems that incorporates temperature delay. These two components aim to improve the precision and efficiency of thermodynamic system modeling, providing new insights and methods for the intelligent management of complex thermodynamic systems.

Introduction: The Challenge of Temperature Delay

In thermodynamic systems, such as industrial heat exchangers or building energy management systems, there is a significant time lag between energy input and the resulting temperature change. This "temperature delay" is caused by factors like pipe length, fluid flow rates, and thermal inertia. Traditional control algorithms often fail to account for these variable lags, leading to energy inefficiency and system instability. If a model does not account for this lag, it may "over-correct," resulting in poor regulation and wasted energy.

Schematic of a typical heat exchange station showing the primary and secondary networks where delays occur.

Methodology

The researchers proposed a two-stage machine learning framework to address these challenges, focusing on both identification and prediction.

1. Temperature Delay Identification

The first stage involves a Pearson correlation-based algorithm. By using a sliding time window and constructing a "Return Water Temperature Lag Matrix," the system shifts output data incrementally. The optimal delay time is identified as the specific time shift that yields the highest correlation between the input (supply water) and output (return water) sequences. This approach is computationally efficient compared to complex physics-based equations.

2. Hybrid HHO-PSO Predictive Model

For system optimization, a hybrid model combining Harris Hawks Optimization (HHO) and Particle Swarm Optimization (PSO) was developed. This model introduces a "Leader" role among the particles to bridge global exploration (HHO) and local exploitation (PSO). This innovation helps the algorithm avoid local optima and ensures more robust convergence through an adaptive weight strategy, where the system starts with large adjustments and narrows down to fine-tune the control.

The proposed machine learning framework, integrating delay identification and predictive modeling.

Experimental Setup and Results

The framework was validated using three distinct datasets: Industrial Process Control, Meteorological Monitoring, and Building Energy Management Systems (BEMS).

Key findings include:

  • Delay Characteristics: Analysis showed that in most thermodynamic systems, delays are concentrated between 1,000 and 2,000 seconds, though they can reach up to 6,000 seconds.
  • Predictive Accuracy: The model demonstrated high precision, particularly in industrial settings. Under a standard 1,500-second delay, the Root Mean Square Error (RMSE) remained low, starting at 0.884 and rising only slightly to 1.147 after 72 iterations.
  • Stability: Median prediction values remained highly consistent across different environmental conditions, proving the model's robustness and stability.

Conclusion

By accurately identifying and predicting temperature delays, this hybrid machine learning approach allows controllers to transition from reactive to proactive management. While the model faces challenges with extremely long delays (exceeding 2,500 seconds), the integration of HHO-PSO and correlation-based lag identification represents a significant advancement in the intelligent management of complex thermodynamic systems. Future work will focus on improving the model's adaptability to a wider variety of thermodynamic environments and real-time hardware integration.

Summary updated

Abstract

With the growing demand for energy and increasing environmental concerns, accurate modeling and efficient optimization of thermodynamic systems have become key research areas in engineering.In recent years, the application of machine learning techniques in thermodynamic systems has demonstrated significant potential, greatly improving modeling precision and optimization efficiency.However, existing methods face challenges in handling temperature delay phenomena, limiting the accuracy and effectiveness of system modeling and control.This paper introduces a temperature delay identification algorithm to enhance the accuracy of delay representation.Additionally, it develops a predictive control model for thermodynamic systems that incorporates temperature delay.These two components aim to improve the precision and efficiency of thermodynamic system modeling, providing new insights and methods for the intelligent management of complex thermodynamic systems.
BibTeX
@article{wang2024application,
  author = {Li Wang and Chang Liu and Ran Huo and Xin Lü and Yue Yang},
  title = {{Application of Machine Learning in Thermodynamic System Modeling and Optimization}},
  journal = {International Journal of Heat and Technology},
  year = {2024},
  doi = {10.18280/ijht.420515},
}
Figure