Over the last 40 years many different control methodologies for substrate feed control of anaerobic digestion processes have been proposed in order to increase plant efficiency and sustainable long-term energy production. This review shows that although sophisticated controllers exist, full-scale biogas plants are mostly still operated without a closed-loop feed control. No matter which application, such control always has to find a compromise between maximizing economic yield, minimizing the ecological footprint and minimizing the risk of process failure. For anaerobic wastewater treatment, control systems which come close to this ideal, exist, but for agricultural as well as industrial biogas plants such control has not yet been developed and neither been successfully implemented and validated at full-scale. Main challenges are a lack of robust and reliable process monitoring using online instrumentation as well as a conservative industry which is reluctant towards the implementation of fully automated process control strategies.

In-situ characterisation of thermocouple sensors is a challenging problem. Recently the authors presented a blind characterisation technique based on the cross-relation method of blind identification. The method allows in-situ identification of two thermocouple probes, each with a different dynamic response, using only sampled sensor measurement data. While the technique offers certain advantages over alternative methods, including low estimation variance and the ability to compensate for noise induced bias, the robustness of the method is limited by the multimodal nature of the cost function. In this paper, a normalisation term is proposed which improves the convexity of the cost function. Further, a normalisation and bias compensation hybrid approach is presented that exploits the advantages of both normalisation and bias compensation. It is found that the optimum of the hybrid cost function is less biased and more stable than when only normalisation is applied. All results were verified by simulation.

The measurement of fast changing temperature fluctuations is a challenging problem due to the inherent limited bandwidth of temperature sensors. This results in a measured signal that is a lagged and attenuated version of the input. Compensation can be performed provided an accurate, parameterised sensor model is available. However, to account for the influence of the measurement environment and changing conditions such as gas velocity, the model must be estimated in-situ. The cross-relation method of blind deconvolution is one approach for in-situ characterisation of sensors. However, a drawback with the method is that it becomes positively biased and unstable at high noise levels. In this paper, the cross-relation method is cast in the discrete-time domain and a bias compensation approach is developed. It is shown that the proposed compensation scheme is robust and yields unbiased estimates with lower estimation variance than the uncompensated version. All results are verified using Monte-Carlo simulations.

An essential task for operation and planning of biogas plants is the optimization of substrate feed mixtures. Optimizing the monetary gain requires the determination of the exact amounts of maize, manure, grass silage, and other substrates. For this purpose, accurate simulation models are mandatory, because the underlying biochemical processes are very slow. The simulation models may be time-consuming to evaluate, hence we show how to use surrogate-model-based approaches to optimize biogas plants efficiently. In detail, a Kriging surrogate is employed. To improve model quality of this surrogate, we integrate cheaply available data into the optimization process. To this end, multi-fidelity modeling methods like Co-Kriging are applied. Furthermore, a two-layered modeling approach is used to avoid deterioration of model quality due to discontinuities in the search space. At the same time, the cheaply available data is shown to be very useful for initialization of the employed optimization algorithms. Overall, we show how biogas plants can be efficiently modeled using data-driven methods, avoiding discontinuities as well as including cheaply available data. The application of the derived surrogate models to an optimization process is only partly successful. Given the same budget of function evaluations, the multi-fidelity approach outperforms the alternatives. However, due to considerable computational requirements, this advantage may not translate into a success with regards to overall computation time.

This paper presents and analyses an engineered expected hypervolume improvement (EHVI) algorithm for solving the problem of PID parameter tuning and the optimization problem of controlling the substrate feed of a biogas plant. The EHVI is the expected value of the increment of the hypervolume indicator given a Pareto front approximation and a predictive multivariate Gaussian distribution of a new point. To solve this problem, S-metric selection-based efficient global optimization (SMS-EGO), EHVI based efficient global optimization (EHVIEGO) and SMS-EMOA are used and compared in both the PID parameter tuning problem and for biogas plant feed optimization. The results of the experiments show that surrogate model based algorithms perform better than SMS-EMOA, and the performance of EHVI-EGO is slightly better than SMS-EGO.

In anaerobic co-digestion plants a mix of organic materials is converted to biogas using the anaerobic digestion process. These organic materials, called substrates, can be crops, sludge, manure, organic wastes and many more. They are fed on a daily basis and significantly affect the biogas production process. In this thesis dynamic real-time optimization of the substrate feed for anaerobic co-digestion plants is developed. In dynamic real-time optimization a dynamic simulation model is used to predict the future performance of the controlled plant. Therefore, a complex simulation model for biogas plants is developed, which uses the famous Anaerobic Digestion Model No. 1 (ADM1). With this model the future economics as well as stability can be calculated resulting in a multi-objective performance criterion. Using multi-objective nonlinear model predictive control (NMPC) the model predictions are used to find the optimal substrate feed for the biogas plant. Therefore, NMPC solves an optimization problem over a moving horizon and applies the optimal substrate feed to the plant for a short while before recalculating the new optimal solution. The multi-objective optimization problem is solved using state-of-the-art methods such as SMS-EMOA and SMS-EGO. The performance of the proposed approach is validated in a detailed simulation study

To produce biogas on demand the substrate feed at biogas plants has to be optimally controlled. For codigestion plants this results in a multi-objective, dynamic optimization problem which has to be solved in real-time. In this dissertation a potential solution candidate is developed and is evaluated in simulation studies.

Zusammenfassung Um Biogas bedarfsgerecht zu produzieren, ist eine optimale Regelung der Substratzufuhr auf Biogasanlagen notwendig. Für Co-Vergärungsanlagen ergibt sich ein multikriterielles, dynamisches Optimierungsproblem, welches in Echtzeit gelöst werden muss. In dieser Dissertation wird ein potenzieller Lösungskandidat entwickelt und in Simulationsstudien evaluiert.

To enhance energy production from methane or resource recovery from digestate, anaerobic digestion processes require advanced instrumentation and control tools. Over the years, research on these topics has evolved and followed the main fields of application of anaerobic digestion processes: from municipal sewage sludge to liquid-mainly industrial-then municipal organic fraction of solid waste and agricultural residues. Time constants of the processes have also changed with respect to the treated waste from minutes or hours to weeks or months. Since fast closed loop control is needed for short time constant processes, human operator is now included in the loop when taking decisions to optimize anaerobic digestion plants dealing with complex solid waste over a long retention time. Control objectives have also moved from the regulation of key variables-measured online-to the prediction of overall process performance-based on global off-line measurements-to optimize the feeding of the processes. Additionally, the need for more accurate prediction of methane production and organic matter biodegradation has impacted the complexity of instrumentation and should include a more detailed characterization of the waste (e.g., biochemical fractions like proteins, lipids and carbohydrates) and their bioaccessibility and biodegradability characteristics. However, even if in the literature several methodologies have been developed to determine biodegradability based on organic matter characterization, only a few papers deal with bioaccessibility assessment. In this review, we emphasize the high potential of some promising techniques, such as spectral analysis, and we discuss issues that could appear in the near future concerning control of AD processes.

The in-line measurement of COD and NH4-N in the WWTP inflow is crucial for the timely monitoring of biological wastewater treatment processes and for the development of advanced control strategies for optimized WWTP operation. As a direct measurement of COD and NH4-N requires expensive and high maintenance in-line probes or analyzers, an approach estimating COD and NH4-N based on standard and spectroscopic in-line inflow measurement systems using Machine Learning Techniques is presented in this paper. The results show that COD estimation using Radom Forest Regression with a normalized MSE of 0.3, which is sufficiently accurate for practical applications, can be achieved using only standard in-line measurements. In the case of NH4-N, a good estimation using Partial Least Squares Regression with a normalized MSE of 0.16 is only possible based on a combination of standard and spectroscopic in-line measurements. Furthermore, the comparison of regression and classification methods shows that both methods perform equally well in most cases.

Background: Environmental systems often have a very complex structure. Methods from computational intelligence (CI) that are often inspired by nature can help to improve these systems. On the one hand, CI methods can be used for optimization; on the other hand, they can be used to extract information out of time series recorded from environmental systems.Methods: Methods from different fields of computational intelligence are investigated. Among them are supervised and unsupervised machine learning methods used for classification and cluster analysis, respectively. Furthermore, methods from evolutionary computation and multi-agent systems are used to develop control and optimization solutions for environmental processes.Results: In this paper, five applications in the fields of anaerobic digestion, pellet-heating, and wastewater management are studied. Using CI methods, e.g., biogas plant operation or a pellet-heating process can be optimized. Furthermore, important process variables can be obtained from huge measurement datasets that otherwise would be unanalyzed and therefore data cemeteries.Conclusions: The results reveal that using CI methods environmental processes can be improved in a favorable cost-benefit fashion.

In this paper a closed-loop substrate feed control for agricultural biogas plants is proposed. In this case, multi-objective nonlinear model predictive control is used to control composition and amount of substrate feed to optimise the economic feasibility of a biogas plant whilst assuring process stability. The control algorithm relies on a detailed biogas plant simulation model using the Anaerobic Digestion Model No. 1. The optimal control problem is solved using the state-of-the-art multi-objective optimization method SMS-EGO. Control performance is evaluated by means of a set point tracking problem in a noisy environment. Results show, that the proposed control scheme is able to keep the produced electrical energy close to a set point with an RMSE of 0.9 %, thus maintaining optimal biogas plant operation.

The concentration of organic acids in anaerobic digesters is one of the most critical parameters for monitoring and advanced control of anaerobic digestion processes. Thus, a reliable online-measurement system is absolutely necessary. A novel approach to obtaining these measurements indirectly and online using UV/vis spectroscopic probes, in conjunction with powerful pattern recognition methods, is presented in this paper. An UV/vis spectroscopic probe from S::CAN is used in combination with a custom-built dilution system to monitor the absorption of fully fermented sludge at a spectrum from 200 to 750 nm. Advanced pattern recognition methods are then used to map the non-linear relationship between measured absorption spectra to laboratory measurements of organic acid concentrations. Linear discriminant analysis, generalized discriminant analysis (GerDA), support vector machines (SVM), relevance vector machines, random forest and neural networks are investigated for this purpose and their performance compared. To validate the approach, online measurements have been taken at a full-scale 1.3-MW industrial biogas plant. Results show that whereas some of the methods considered do not yield satisfactory results, accurate prediction of organic acid concentration ranges can be obtained with both GerDA and SVM-based classifiers, with classification rates in excess of 87% achieved on test data.

Optimal control of biogas plants is a complex and challenging task due to the nonlinearity of the anaerobic digestion process involved in the conversion of biodegradable input material to biogas (a mixture of the energy carrier methane and carbon dioxide). In this paper a nonlinear model predictive control (NMPC) algorithm is developed to optimally control the substrate feed of the anaerobic digestion process on biogas plants. The implemented algorithm is investigated in a simulation study using a validated simulation model of a full-scale biogas plant with an electrical power of 750 kW, where the control objective is to achieve high biogas production and quality while maintaining stable plant operation. Results are presented demonstrating the feasibility of the proposed approach. The optimal operating state identified by the controller provides an additional return of investment of 650 €/day compared to a nominal operating state. Using the proposed algorithm it will be possible in the near future to optimize full-scale biogas plants using nonlinear model predictive control and therefore to advance the use of anaerobic digestion for eco-friendly energy production.

The concentration of organic acids in anaerobic digesters is one of the most critical parameters for monitoring and advanced control of anaerobic digestion processes. Thus, a reliable online-measurement system is absolutely necessary. A novel approach to obtaining these measurements indirectly and online using UV/vis spectroscopic probes, in conjunction with powerful pattern recognition methods, is presented in this paper. An UV/vis spectroscopic probe from S::CAN is used in combination with a custom-built dilution system to monitor the absorption of fully fermented sludge at a spectrum from 200 to 750 nm. Advanced pattern recognition methods are then used to map the non-linear relationship between measured absorption spectra to laboratory measurements of organic acid concentrations. Linear discriminant analysis, generalized discriminant analysis (GerDA), support vector machines (SVM), relevance vector machines, random forest and neural networks are investigated for this purpose and their performance compared. To validate the approach, online measurements have been taken at a full-scale 1.3-MW industrial biogas plant. Results show that whereas some of the methods considered do not yield satisfactory results, accurate prediction of organic acid concentration ranges can be obtained with both GerDA and SVM-based classifiers, with classification rates in excess of 87% achieved on test data.