Review of Soft Sensors in Anaerobic Digestion Process

Abstract

Anaerobic digestion is associated with various crucial variables, such as biogas yield, chemical oxygen demand, and volatile fatty acid concentration. Real-time monitoring of these variables can not only reflect the process of anaerobic digestion directly but also accelerate the efficiency of resource conversion and improve the stability of the reaction process. However, the current real-time monitoring equipment on the market cannot be widely used in the industrial production process due to its defects such as expensive equipment, low accuracy, and lagging analysis. Therefore, it is essential to conduct soft sensor modeling for unmeasurable variables and use auxiliary variables to realize real-time monitoring, optimization, and control of the an-aerobic digestion process. In this paper, the basic principle and process flow of anaerobic digestion are first briefly introduced. Subsequently, the development history of the traditional soft sensor is systematically reviewed, the latest development of soft sensors was detailed, and the obstacles of the soft sensor in the industrial production process are discussed. Finally, the future development trend of deep learning in soft sensors is deeply discussed, and future research directions are provided.

A Critical Review of Data Science Applications in Resource Recovery and Carbon Capture from Organic Waste

“…Data-driven modeling tools such as ML methods can often leverage real-time monitoring sensor data in modern RRCC infrastructure to improve the efficiency of treatment processes or provide additional benefits over conventional practices where process controls are informed by past performance and operator’s experience. ,− However, at the current global climate where authorities are pushing waste management infrastructures to adopt practices for reducing environmental impacts while maintaining economic prosperity, the priority for achieving process efficiency for RRCC has developed. , For example, at the end of 2022, United States Department of Energy issued a USD 23 million funding opportunity to drive innovative decarbonization strategies to reduce greenhouse gas emissions from public water resource recovery infrastructure in the USA by 25% without increasing the overall cost . Although these infrastructures have achieved high RRCC efficiency over the past decade, large amounts of energy are consumed by the underlying treatment processes that are resulting in life cycle emissions close to that of the cement industry (approximately 44 t of CO 2 equivalents). − One possible approach that can aid the research and development of such innovative decarbonization strategies could be the integration of data-driven models for process modeling with LCA and LCCA/TEA for the environmental and economic impact analyses, respectively.…”

RegGAN: A Virtual Sample Generative Network for Developing Soft Sensors with Small Data

“…Data-driven soft sensors that learn mapping relationships between auxiliary and quality variables to construct predictive models represent an effective method for overcoming the challenges of acquiring crucial quality variables . The rapid developments in distributed control systems and artificial intelligence algorithms have resulted in the widespread adoption of data-driven soft sensors . Among these, soft sensors leveraging deep learning have garnered favor among researchers owing to their robust nonlinear fitting capabilities.…”

Machine Learning Algorithm Integrated with Real-Time In Situ Sensors and Physiochemical Principle-Driven Soft Sensors toward an Anaerobic Digestion-Data Fusion Framework

“…In terms of model prediction, it is vital to ensure the interpretability of MLA for AD systems. 28 Commonly, simple models (e.g., linear regression) 30,31 and model-agnostic methods exploit feature interactions (e.g., predicting methane kinetics) 19,32,33 and internal data dependencies (e.g., local surrogates). 34,35 While the state-of-the-art MLA effectively describes the relationship between important ML features and model predictions, the lack of direct translation between statistically summarized feature importance and AD mechanisms hampers the reliability and clarity of MLA features for AD operation and adjustment.…”

“…In terms of model prediction, it is vital to ensure the interpretability of MLA for AD systems . Commonly, simple models (e.g., linear regression) , and model-agnostic methods exploit feature interactions (e.g., predicting methane kinetics) ,, and internal data dependencies (e.g., local surrogates). , While the state-of-the-art MLA effectively describes the relationship between important ML features and model predictions, the lack of direct translation between statistically summarized feature importance and AD mechanisms hampers the reliability and clarity of MLA features for AD operation and adjustment. , Our second objective is to bridge the gap between the statistical feature importance of MLA and the concrete physiochemical meanings of AD mechanisms (Figure a). Our hypothesis is that the combined data from multidimensional sensor monitoring and “white-box” biochemical models contains sufficient information to accurately predict patterns in unseen AD processes.…”