Abstract
This article presents the application of a recent neural network topology known as the deep echo state network to the prediction and modeling of strongly nonlinear systems typical of the process industry. The article analyzes the results by introducing a comparison with one of the most common and efficient topologies, the long short-term memories, in order to highlight the strengths and weaknesses of a reservoir computing approach compared to one currently considered as a standard of recurrent neural network. As benchmark application, two specific processes common in the integrated steelworks are selected, with the purpose of forecasting the future energy exchanges and transformations. The procedures of training, validation and test are based on data analysis, outlier detection and reconciliation and variable selection starting from real field industrial data. The analysis of results shows the effectiveness of deep echo state networks and their strong forecasting capabilities with respect to standard recurrent methodologies both in terms of training procedures and accuracy.
Highlights
In the last decade, process industry and, in general, all the energy-intensive sectors, are facing increasingly complex economic challenges, due to the variability of the raw materials market, the enormous variability of the demand for goods and services and the daily fluctuation of the energy market, in terms of cost and availability of electricity and fuels
The novelties presented in this work are related to the application of deep learning (DL) methodologies and in particular deep ESNs (DESN), for forecasting the energy contents of processes that are common in the steel industries, namely processes characterized by state variables that cannot be measured and poor available exogenous information
The paper proposes the application of a particular reservoir computing approach based on DESN in order to model the nonlinear dynamics typical of complex industrial processes
Summary
Process industry and, in general, all the energy-intensive sectors, are facing increasingly complex economic challenges, due to the variability of the raw materials market, the enormous variability of the demand for goods and services (let us just consider what is happening in the ongoing period characterized by the COVID19 pandemic) and the daily fluctuation of the energy market, in terms of cost and availability of electricity and fuels. In the context of process industry, Matino et al presented a work related to the forecasting of blast furnace gas through ESN techniques [13], and Dettori et al highlighted the effectiveness of AI methodologies for modeling energy transformation equipment in the industry [14]. This work proposes a comparison between deep echo state networks (DESN) and long short-term memories (LSTM) for modeling and forecasting the complex nonlinear behavior of the blast furnace (BF) process and of some related auxiliary units (i.e., hot blast stoves) as far as off-gas production and consumption are concerned. The novelties presented in this work are related to the application of DL methodologies and in particular DESN, for forecasting the energy contents of processes that are common in the steel industries, namely processes characterized by state variables that cannot be measured and poor available exogenous information. The paper is organized as follows: Sect. 2 provides some theoretical background on DESNs and LSTMs; Sect. 3 presents the considered industrial application; Sect. 4 provides details of the developed models and exploited industrial datasets; Sect. 5 focuses on the obtained numerical results, while Sect. 6 provides some concluding remarks and hints for future work
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
