Predicting NOx emissions in Biochar Production Plants using Machine Learning

TL;DR

The paper tackles NOx emission control in scaling biochar production for Biochar Carbon Removal (BCR). It develops a Random Forest regression surrogate to map pyrolysis machine states to sensor outputs, enabling real-time NOx prediction and optimization without continuous NOx sensors, demonstrated on two reactors with IoT integration. The approach achieves strong predictive performance (e.g., $R^2=0.97$ for PRYEG and $R^2=0.84$ for ARTi) and shows a viable optimization path under emission constraints, highlighting practical applicability for scalable, regulation-compliant BCR deployment. This work provides a tangible route to reducing emissions while maintaining production, leveraging ML surrogates and IoT-driven workflows across heterogeneous pyrolysis equipment.

Abstract

The global Biochar Industry has witnessed a surge in biochar production, with a total of 350k mt/year production in 2023. With the pressing climate goals set and the potential of Biochar Carbon Removal (BCR) as a climate-relevant technology, scaling up the number of new plants to over 1000 facilities per year by 2030 becomes imperative. However, such a massive scale-up presents not only technical challenges but also control and regulation issues, ensuring maximal output of plants while conforming to regulatory requirements. In this paper, we present a novel method of optimizing the process of a biochar plant based on machine learning methods. We show how a standard Random Forest Regressor can be used to model the states of the pyrolysis machine, the physics of which remains highly complex. This model then serves as a surrogate of the machine -- reproducing several key outcomes of the machine -- in a numerical optimization. This, in turn, could enable us to reduce NOx emissions -- a key regulatory goal in that industry -- while achieving maximal output still. In a preliminary test our approach shows remarkable results, proves to be applicable on two different machines from different manufacturers, and can be implemented on standard Internet of Things (IoT) devices more generally.

Figures (3)

• Figure 1: Sketch of a general continues pyrolysis reactor. The screws (e.g. valve position or ventilator) are the "machine states" (input features), while the sensor symbols (e.g. temperature or NOx) are the target which the Random Forest is predicting given the machine states. Note that the ARTi reactor has multiple champers so this image only shows a very basic design. For a more detailed representation we encounter the reader to visit the websites pyreg and arti.
• Figure 2: Figure \ref{['fig:pipeline']}: Overview of the used machine learning pipeline. The grey boxes represent the parts that are employed at the machine, while the brown boxes show the pipeline to pre-train the model on a server. Figure \ref{['fig:subfigA']}: First test of predicting the reactor temperature which is a key feature of how much NOx is produced.
• Figure 3: Figure \ref{['fig:subfigB']} and \ref{['fig:subfigD']}: Show the prediction of NOx value using historical data from the PYREG and the ARTi reactor respectively. Figure \ref{['fig:opti']} shows the optimization path for the ARTi data to minimize NOx and have CO2 and O2 (as two example values) constrained between CO2 0-10 % and O2 0-20 %.