Energy efficiency evaluation and prediction of large-scale chemical plants using partial least squares analysis integrated with Gaussian process models

In the large-scale chemical plant, energy efficiency evaluation and management plays a crucial role in the sustainable development. However, most data-driven approaches to calculating energy efficiency still focus on the deterministic evaluation regardless of the effects resulting from the data quality. Besides, because of the multi-dimensional, stochastic and uncertain characteristics of the collected industrial data, it is more difficult to accurately and reliably evaluate and predict the energy usage of the complex chemical production process. To solve these limitations, a novel energy efficiency evaluation and prediction method integrated with the Gaussian process (GP) and the partial least squares (PIS) analysis, named GP-PLS method, is proposed in this paper. This integrated method can select the appropriate sampling data, construct the corresponding nonlinear model and indicate the reliability of the model. Based on the constructed models, not only are the accurate energy efficiency evaluation results calculated, but the energy efficiency tendency is also predicted. In addition, according to the reliability of the established model, the informative data are intelligently selected from the historical database to enhance the quality of the model. The effectiveness and practicality of the proposed method are demonstrated through a numerical example and a practical ethylene production. The evaluation results and the prediction values are acquired to decrease the energy consumption, guide the production process and improve the energy efficiency level in the large-scale chemical plant.