Periodic Reporting for period 4 - HetScaleNet (Analysis and control of large scale heterogeneous networks: scalability, robustness and fundamental limits)
Reporting period: 2021-02-01 to 2022-07-31
The project is associated with the analysis and control of large scale networks, and involves the development of corresponding theoretical tools as well as the study of related applications associated with engineering and biological networks.
In particular, for the case of engineering networks our research has focused on the design of decentralized control policies with a plug-and-play capability, i.e. stability guarantees are provided for the network when heterogeneous subsystems are added or removed from the network. This is a major requirement in many important applications such as smart grids and power networks, where the increasing presence for distributed generation, fluctuations from renewable energy sources and their low inertia, and the introduction of load-side participation schemes impose severe challenges associated with the ability to ensure a stable network with good performance. Throughout the project various results have been developed that lead to appropriate decentralized design protocols for such networks. These improve their efficiency and reliability, and facilitate the implementation of such highly distributed control schemes.
For the case of biological networks our research has focused on the analysis of the effects of noise in biochemical reaction networks at the molecular level. Such molecular fluctuations can drive metabolites away from desired concentrations or be sometimes advantageous contributing to diversity and evolution. Furthermore, such stochastic models are also relevant in epidemiological models and can be used to understand the role of feedback when mitigation strategies are introduced. Throughout the project novel tools have been developed for analyzing the stochasticity in classes of biochemical reactions, and tools for constructing optimal feedback policies in stochastic epidemiological models have been developed where various generic properties of those have been characterized.
In particular, for the case of engineering networks our research has focused on the design of decentralized control policies with a plug-and-play capability, i.e. stability guarantees are provided for the network when heterogeneous subsystems are added or removed from the network. This is a major requirement in many important applications such as smart grids and power networks, where the increasing presence for distributed generation, fluctuations from renewable energy sources and their low inertia, and the introduction of load-side participation schemes impose severe challenges associated with the ability to ensure a stable network with good performance. Throughout the project various results have been developed that lead to appropriate decentralized design protocols for such networks. These improve their efficiency and reliability, and facilitate the implementation of such highly distributed control schemes.
For the case of biological networks our research has focused on the analysis of the effects of noise in biochemical reaction networks at the molecular level. Such molecular fluctuations can drive metabolites away from desired concentrations or be sometimes advantageous contributing to diversity and evolution. Furthermore, such stochastic models are also relevant in epidemiological models and can be used to understand the role of feedback when mitigation strategies are introduced. Throughout the project novel tools have been developed for analyzing the stochasticity in classes of biochemical reactions, and tools for constructing optimal feedback policies in stochastic epidemiological models have been developed where various generic properties of those have been characterized.
Throughout the project systematic methods have been developed for the design of decentralized control policies with a plug-and-play capability in large scale networks. Main application areas that have been addressed are associated with decentralized control schemes in power networks. In particular, these include decentralized design protocols for converter based grids such that stability and power sharing can be ensured, which is an important problem associated with the large-scale integration of renewable energy sources. Furthermore, related issues have been addressed for the control of hybrid AC/DC networks, and schemes where load-side participation contributes to the real time balancing of supply and demand. The effect of delays has also been studied in such control policies which is an important aspect due to the additional communication layer present in their implementation. Results in the area of distributed optimization have also been established; these include the characterization of the asymptotic behaviour of a broad class of dynamics in this area, which can facilitate the construction of new decentralized algorithms with improved performance.
For the case of biological networks tools have been developed for quantifying the stochasticity in classes of biochemical reactions with nonlinear reaction rates. Furthermore, the problem of optimal control for classes of stochastic epidemiological models has been studied, and systematic tools for constructing optimal feedback policies have been developed. Various generic properties in optimal feedback policies have also been derived, which reveal limitations of feedback control in providing an improved performance.
For the case of biological networks tools have been developed for quantifying the stochasticity in classes of biochemical reactions with nonlinear reaction rates. Furthermore, the problem of optimal control for classes of stochastic epidemiological models has been studied, and systematic tools for constructing optimal feedback policies have been developed. Various generic properties in optimal feedback policies have also been derived, which reveal limitations of feedback control in providing an improved performance.
Issues of plug-and-play capability are currently addressed in heuristic ways in practical implementations often resulting in conservative designs due to the lack of stability and performance guarantees. Therefore the results developed throughout the project facilitate the implementation of highly distributed control schemes in large scale networks, providing stability guarantees and improved performance. In particular, the application of these results to power networks and converter based grids will contribute to the development of appropriate grid codes for such networks and thus facilitate the large scale integration of renewable energy sources. For biological networks the approaches for analysing stochasticity that have been developed and the problem of optimal control of stochastic epidemic models that has been addressed will facilitate the development of mitigation strategies during the outbreak of infectious diseases.