Details
| Original language | German |
|---|---|
| Qualification | Doctor of Engineering |
| Awarding Institution | |
| Date of Award | 1 Oct 2024 |
| Place of Publication | Hannover |
| Print ISBNs | 978-3-948492-05-8 |
| Publication status | Published - 2024 |
Abstract
measurement-based control strategies, is limited due to strong temporal variations in N2O emissions and a lack of data regarding the influencing parameters.
To address these issues, a novel AI-based process optimization method for minimizing N2O emissions was developed in this work. This method uses a genetic algorithm to automatically identify the control settings associated with minimum N2O emissions for a specific operating situation. In addition to decreasing emissions, this approach enables a holistic control of biological wastewater treatment by incorporating additional individually definable operating targets such as sufficient nitrogen elimination. The genetic algorithm relies on a prediction model
to evaluate the effect of individual control parameter sets on N2O emissions and other operating targets. For this purpose, neural networks were trained in this work. Neural networks are advantageous as they offer faster calculation speed than conventional mechanistic models due to their mathematically simpler structure. The data for network training, however, was generated using a mechanistic model. This approach is beneficial in practical applications as prediction networks could be successfully trained even if only limited data is available. The developed method is also incorporated with a classification algorithm to give the reliability of
the suggested control strategy.
The novel approach described in this thesis was successfully tested for three exemplary applications. In all examples, a considerable reduction in emissions was achieved. The main causes of N2O formation were consistently identified by the algorithm despite a limited number of input parameters, and targeted reduction measures were recommended. With its proven applicability, the proposed method for intelligent optimization is ready to support the plant operating staff. In the long term, intelligent algorithms will enable further automation and digitalization of wastewater treatment plants leading to sustainable, low-emission operations.
Keywords
- greenhouse gas emissions, artificial intelligence, system optimization
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
Hannover, 2024. 170 p.
Research output: Thesis › Doctoral thesis
}
TY - BOOK
T1 - N₂O-Emissionsreduzierung durch KI-unterstützte Regelung der biologischen Abwasserreinigung
AU - Freyschmidt, Arne Holger
PY - 2024
Y1 - 2024
N2 - Current efforts to reduce global greenhouse gas emissions also lead to a sharp increase in the implementation of targeted measures in sanitary engineering. One of the main focuses of the ongoing research and plant operating practice is minimizing N2O emissions during wastewater treatment. N2O is a greenhouse gas that can be formed as a by-product of biological nitrogen elimination under unfavorable operating conditions (e.g. high loads, nitrite accumulation, shock loads...). The potential of the available measures for achieving lower N2O emissions, namelymeasurement-based control strategies, is limited due to strong temporal variations in N2O emissions and a lack of data regarding the influencing parameters.To address these issues, a novel AI-based process optimization method for minimizing N2O emissions was developed in this work. This method uses a genetic algorithm to automatically identify the control settings associated with minimum N2O emissions for a specific operating situation. In addition to decreasing emissions, this approach enables a holistic control of biological wastewater treatment by incorporating additional individually definable operating targets such as sufficient nitrogen elimination. The genetic algorithm relies on a prediction modelto evaluate the effect of individual control parameter sets on N2O emissions and other operating targets. For this purpose, neural networks were trained in this work. Neural networks are advantageous as they offer faster calculation speed than conventional mechanistic models due to their mathematically simpler structure. The data for network training, however, was generated using a mechanistic model. This approach is beneficial in practical applications as prediction networks could be successfully trained even if only limited data is available. The developed method is also incorporated with a classification algorithm to give the reliability ofthe suggested control strategy.The novel approach described in this thesis was successfully tested for three exemplary applications. In all examples, a considerable reduction in emissions was achieved. The main causes of N2O formation were consistently identified by the algorithm despite a limited number of input parameters, and targeted reduction measures were recommended. With its proven applicability, the proposed method for intelligent optimization is ready to support the plant operating staff. In the long term, intelligent algorithms will enable further automation and digitalization of wastewater treatment plants leading to sustainable, low-emission operations.
AB - Current efforts to reduce global greenhouse gas emissions also lead to a sharp increase in the implementation of targeted measures in sanitary engineering. One of the main focuses of the ongoing research and plant operating practice is minimizing N2O emissions during wastewater treatment. N2O is a greenhouse gas that can be formed as a by-product of biological nitrogen elimination under unfavorable operating conditions (e.g. high loads, nitrite accumulation, shock loads...). The potential of the available measures for achieving lower N2O emissions, namelymeasurement-based control strategies, is limited due to strong temporal variations in N2O emissions and a lack of data regarding the influencing parameters.To address these issues, a novel AI-based process optimization method for minimizing N2O emissions was developed in this work. This method uses a genetic algorithm to automatically identify the control settings associated with minimum N2O emissions for a specific operating situation. In addition to decreasing emissions, this approach enables a holistic control of biological wastewater treatment by incorporating additional individually definable operating targets such as sufficient nitrogen elimination. The genetic algorithm relies on a prediction modelto evaluate the effect of individual control parameter sets on N2O emissions and other operating targets. For this purpose, neural networks were trained in this work. Neural networks are advantageous as they offer faster calculation speed than conventional mechanistic models due to their mathematically simpler structure. The data for network training, however, was generated using a mechanistic model. This approach is beneficial in practical applications as prediction networks could be successfully trained even if only limited data is available. The developed method is also incorporated with a classification algorithm to give the reliability ofthe suggested control strategy.The novel approach described in this thesis was successfully tested for three exemplary applications. In all examples, a considerable reduction in emissions was achieved. The main causes of N2O formation were consistently identified by the algorithm despite a limited number of input parameters, and targeted reduction measures were recommended. With its proven applicability, the proposed method for intelligent optimization is ready to support the plant operating staff. In the long term, intelligent algorithms will enable further automation and digitalization of wastewater treatment plants leading to sustainable, low-emission operations.
KW - Treibhausgasemissionen
KW - Künstliche Intelligenz
KW - Systemoptimierung
KW - greenhouse gas emissions
KW - artificial intelligence
KW - system optimization
M3 - Dissertation
SN - 978-3-948492-05-8
CY - Hannover
ER -