Refine
Document Type
- Doctoral thesis (2)
- Scientific article (1)
Has Fulltext
- yes (3)
Is part of the Bibliography
- no (3)
Language
- English (3)
Keywords
Institute
In this thesis, we develop a mathematical framework to estimate the compartment sizes of an epidemic model with some unobserved compartments (i.e., a model with partial information) and formulate a social planner's stochastic optimal control problem for such an epidemic. Including asymptomatic and unreported infectious individuals in the model allows a more realistic description of the course of an epidemic and avoids misleading estimation.
In order to estimate the unobservable states, a suitable description of the dynamics of the epidemic progression by a diffusion approximation is achieved using counting processes. The estimation of the unobservable states is attributed to a stochastic filter problem. To solve this problem, we use the extended Kalman filter, which provides approximate solutions for such non-standard filter problems. This approach is applied to the state estimation of a model with incomplete information for a disease with lifelong immunity after recovery or vaccination. Simulations are carried out to demonstrate the effectiveness of this approach in practice.
The main part of this work addresses a social planner's stochastic optimal control problem for an epidemic model with a partially observed state process. It decomposes the population into susceptible, detected and non-detected infected, detected and non-detected recovered, and hospitalized individuals. Control measures include social distancing, testing, and vaccination. We apply a filtering argument to transform the partially observable stochastic optimal control problem into a completely observable control problem by replacing the hidden state process with the associated Kalman filter processes. This transformed problem with an eight-dimensional state process is treated as a Markov decision process after a time discretization. The associated Bellman equation is solved numerically using a backward recursion algorithm combined with state discretization and quantization techniques to mitigate the curse of dimensionality. In particular, two approaches are implemented. The first involves state discretization and linear interpolation of the value function between grid points, while the second utilizes state discretization and parameterization of the value function with educated ansatz functions. Extensive numerical experiments are presented to demonstrate the effectiveness of both approximate optimal controls in containing the epidemic.
Finally, we generalize the results of Picard (1991) on the efficiency of the continuous-time extended Kalman filter with small noise scaled by a parameter ε>0. First, we show that when the observation drift coefficient is strongly injective and the signal and observation drift become linear for ε → 0, the estimation error is of order √ε. Subsequently, we establish conditions under which the error in the initial filter estimate decays exponentially fast.
This thesis presents a novel stochastic modeling framework for epidemic dynamics, focusing on the challenges posed by non-observable states and the gradual decline of immunity, as revealed by the Covid-19 pandemic. Our research develops an extended compartmental model that incorporates sequences of cascade states to accurately represent the progressive loss of immunity after vaccination or recovery. This model, rooted in a non-homogeneous continuous-time Markov chain and approximated by a diffusion process, facilitates efficient computation. The resulting dynamics exhibit non-linearities in both the drift and diffusion coefficients, making the process of hidden state estimation particularly challenging. These complexities highlight the need for advanced techniques to accurately capture and predict the system's behavior under partial observability. A novel approach using cascade states is introduced to tackle the issue of partially hidden compartments where either inflow or outflow is observed, but not both. By leveraging information from observed transitions, this method significantly enhances model accuracy. The framework also enables the application of advanced filtering techniques, particularly the Extended Kalman Filter (EKF), for estimating both hidden epidemic states and time-varying parameters. This allows for calibrating based on real data, ensuring robust and accurate predictions. Furthermore, this thesis extends the methodology to multi-group (multi-patch) models, capturing the complexities of heterogeneous population dynamics. This includes factors such as mobility, super-spreader events, and accounting for partial information. By incorporating these elements, the model provides a more comprehensive understanding of epidemic spread across diverse and interconnected populations. To address the curse of dimensionality often encountered in high-dimensional filtering problems, a modified EKF based on a reduced-order EKF is introduced, leveraging a low-rank approximation of the covariance matrix. This reduced-order EKF allows for stable and efficient estimation of hidden states even with limited computational resources. Finally, the work explores parameter estimation within the partial information context, employing both likelihood-based inference and state augmentation techniques. Numerical experiments demonstrate the model's ability to accurately reproduce observed Covid-19 dynamics in Germany, highlighting its potential for informing public health strategies and managing future outbreaks. This work offers a robust and adaptable framework for understanding and predicting the spread of infectious diseases.
The optimal control of sustainable energy supply systems, including renewable energies and energy storage, takes a central role in the decarbonization of industrial systems. However, the use of fluctuating renewable energies leads to fluctuations in energy generation and requires a suitable control strategy for the complex systems in order to ensure energy supply. In this paper, we consider an electrified power-to-heat system which is designed to supply heat in the form of superheated steam for industrial processes. The system consists of a high-temperature heat pump for heat supply, a wind turbine for power generation, a sensible thermal energy storage for storing excess heat, and a steam generator for providing steam. If the system’s energy demand cannot be covered by electricity from the wind turbine, additional electricity must be purchased from the power grid. For this system, we investigate the cost-optimal operation, aiming to minimize the electricity cost from the grid by a suitable system control depending on the available wind power and the amount of stored thermal energy. This is a decision-making problem under uncertainty regarding the future prices for electricity from the grid and the future generation of wind power. The resulting stochastic optimal control problem is treated as finite-horizon Markov decision process for a multi-dimensional controlled state process. We first consider the classical backward recursion technique for solving the associated dynamic programming equation for the value function and compute the optimal decision rule. Since that approach suffers from the curse of dimensionality, we also apply reinforcement learning techniques, namely Q-learning, that are able to provide a good approximate solution to the optimization problem within reasonable time.
