Biogas Digester State Classification Using Outlier Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Biogas plants face challenges in process control due to inadequate knowledge of operators and under-equipment regarding online measurement technology, leading to unfavorable substrate feed quantities and compositions, which can disrupt the digestion process, especially in co-substrate operations, and current methods for process state classification are either not up-to-date or require extensive training data.
Innovation Solution
A method for classifying the process state of a biogas digester using a model trained with historical data from the digester, identifying outliers in measured process variables such as hydrogen and methane concentrations, and employing algorithms to detect anomalies, thereby providing real-time stability monitoring.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual VFA sampling and HPLC analysis are used for process state classification, then measurement precision is improved, but loss of time increases significantly (1-3 days delay)
Solution Approach 1:
The patent replaces the mechanical/chemical analysis system (manual sampling and HPLC) with an electronic sensor-based measurement system. Online sensors continuously monitor process variables (pH, temperature, gas composition) and use machine learning models to classify process states, eliminating the time-consuming manual analysis while maintaining classification accuracy through pattern recognition algorithms.
Solution Approach 2:
The system performs preliminary classification of process states using online sensors and pre-trained machine learning models, enabling real-time detection of process deviations before they become critical. This allows operators to take preventive actions based on predicted process states rather than waiting for delayed HPLC results.
2Device complexity
If a one-fits-all threshold approach is used for process state classification, then device complexity is reduced, but measurement precision deteriorates due to system-specific variations
Solution Approach 1:
The patent transforms the classification approach from using fixed threshold values to using dynamic parameter ranges derived from historical data. Machine learning models analyze patterns in multiple process variables simultaneously, adapting classification criteria to each specific digester's operational characteristics while maintaining a unified system architecture that can be applied across different plants.
Solution Approach 2:
The system implements feedback mechanisms where classification results and operational outcomes are continuously fed back into the machine learning models. This allows the system to learn from past performance, refine classification accuracy, and adapt to changing operational conditions while maintaining system simplicity through automated adjustments rather than complex manual threshold tuning.
3Measurement precision
If machine learning models are trained with extensive labeled disruption events, then process state classification accuracy is improved, but loss of time increases due to data preparation and model re-training requirements
Solution Approach 1:
The patent performs preliminary training of machine learning models during system commissioning using historical data and expert-labeled disruption events. Once trained, the models are deployed for continuous real-time classification without requiring frequent re-training. The system maintains accuracy through continuous learning from new data and periodic updates rather than extensive repeated training cycles.
Solution Approach 2:
The machine learning models implement self-service through continuous learning mechanisms where the system automatically adapts to new patterns in the data. The models can identify and learn from new disruption types autonomously, reducing the need for manual data labeling and re-training while maintaining high classification accuracy over extended operational periods.
Data Source
AI summary
Method for classifying a process state of a biogas digester based on at least one process variable of a biogas plant, the method comprising the steps of: a) Measuring a set of values of at least one process variable of a biogas digester; b) Providing a model that monitors the stability of the biogas digester, wherein the model is trained with a dataset comprising historical data of the digester of the biogas plant, the historical data including the at least one process variable; c) Implementing the measured set of values of the process variable of step a) into the model of step b) with the aid of a data processing unit; d) Identifying whether an outlier occurs in the set of values of step a), after implementing the at least one process variable into the model; e) Classifying a process state indicative of the stability of the biogas digester based on the presence or absence of the identified outlier of step d) with the aid of the data processing unit.