Resilience in the Gas Grid Infrastructure: Making Smart Decisions in Critical Situations
Germany's electricity and gas transmission and distribution infrastructure is exceptionally robust. Most disruptive events can be handled without causing significant supply disruptions. But the global coronavirus pandemic, devastating floods and increasing cyber attacks show us our vulnerability. Despite high levels of precautionary investment, unforeseen events and threats unsettle us. But how can operators of critical infrastructures react quickly and safely in exceptional situations? What do control systems need to do today so that utilities can confidently manage crises in the future?
To ensure the necessary resilience of the power system, data is increasingly automated and provided in a decision-oriented manner. Malfunctions can thus often already be detected before they hit.
Control Systems Are the Technical Heart of Energy Systems
Disruptions to the energy infrastructure, for example due to terrorist attacks on storage facilities and networks or even line disconnections, as well as leaks due to landslides during natural disasters, are often unpredictable. Control systems, as the technical heart of power systems, enable the development and integration of crisis response functions as needed. Crisis response functions should provide a high degree of safety in critical situations. To achieve this, they must be easy and safe to operate - especially under stress.
Comparable to the distress call on ships: the red distress button can be used to request help in an emergency and to initiate the rescue operation via a proven response chain. In the event of a crisis, the focus is thus on automated alerting and information provision as well as fast and secure decision support.
Digitized Energy Networks Enable High-Quality Situational Awareness
Our energy systems are increasingly becoming highly complex, distributed technical systems: large-scale power plants are being replaced by a variety of renewable energy sources. In order to ensure a stable and affordable energy supply even in the face of volatile feed-ins and seasonal fluctuations, the interaction between electricity and gas networks is continuously increasing.
A prerequisite for this is the digitization of energy networks and continuous monitoring of as many operating resources as possible. Malfunctions can thus often be detected already before they happen.
However, due to the complexity, disruptions cannot be completely prevented. Therefore, the energy system must be designed in such a way that any disruptions that occur do not have serious consequences. Depending on the scale of the crisis or the severity of the emergency, there may still be room to maneuver, and experienced dispatchers may be able to take advantage. However, the more widespread a disruption becomes, the more difficult it becomes to keep track of potential collateral damage from individual decisions. In order to be able to bring the power system quickly and safely into a stable state in an exceptional situation, a complete situation picture of the
- current network status,
- impending bottlenecks and
- possible options for action is required.
PSI stands for powerful support in network management, network simulation, forecasting and transport management. PSI applications provide comprehensive and precise information for the smooth and permanent operation of the gas infrastructure for network operators. This includes network condition and process data as well as forecast data for different scenarios.
Decision Support With Artificial Intelligence
We are constantly working to improve the resilience of critical gas network infrastructure and to adapt it to increasing complexity. The existing product portfolio is to be expanded to include AI-based security solutions. The aim is to provide powerful and intelligent tools for
- forecasting and
- decision support
in real-time operations to further strengthen the resilience of the gas infrastructure.
Complex network state situations are analyzed and evaluated using powerful AI-based algorithms. A first result is the software solution PSIgasguide, which combines simulation and compressor optimization with established AI-based models for decision-making. A special feature here is that the focus is on decision support with regard to several target variables is optimized with Fuzzylogic. Dispatchers thus obtain a more comprehensive picture of possible courses of action.
The simulation is used for highly precise simulations of flows in gas transport networks and gas distribution networks. This offers users the possibility to plan gas network conditions in advance. It is also successfully used for leak detection and leak location.
In the event of a crisis or disaster, crisis management is supported by the provision of a reliable status assessment and possible options for action to maintain or quickly restore supply security. Response times in crisis situations are thus to be significantly shortened and cascade effects minimized.
Continuous Condition Assessment and Reliable Anomaly Detection
The basis for action recommendations is provided by the data collected by the control system. This data is transferred into a network condition model in which each network condition is qualitatively evaluated and deviations from a safe network condition are reliably detected. Measures to eliminate the deviations are mapped in an action model in the form of switching and setpoint specifications as well as the dispatcher's empirical knowledge.
The knowledge of experienced dispatchers is standardized and processed for the evaluation of exceptionally critical situations (network condition model) and control measures based on this (action model). These form the ground truth data for the AI. The network state and the actions performed are thus coupled: the situation and the actions performed are continuously evaluated.
Powerful Training Systems for Dispatchers
In the development of the AI-based security solution, PSI focuses on strengthening trust in the AI algorithms and in the traceability of the procedures and learning effects used. To this end, simple controls are provided through which users can influence the prioritization of different criteria.
In addition, a goal-relationship matrix is disclosed, through which mutual effects can be identified. On this basis, powerful training systems for dispatchers can be built and resilience KPIs for attack strategies can be obtained and provided from adversarial learning.
The teaching of the security solution by experienced users is an integral part of the solution approach. The transferability of the solutions to other critical infrastructures for electricity, water and heat and their connection to a uniform overall system is an architectural principle of the selected solution.
Reacting Quickly and Safely in Critical Situations
Operators of critical infrastructures must react quickly and safely, especially in exceptional situations. A comprehensive picture of the situation across all affected energy and structural sectors is just as important as fast and secure access to experiential knowledge.
The expansion of the PSI product suite with the AI-based security solution supports our customers in further strengthening the resilience of the gas infrastructure.
What is your opinion on this topic?
Definitions at a Glance
is the ability of a system to maintain its ability to function under stress or to restore it in the short term. Resilience goes beyond the property of robustness. A resilient energy system remains functional even in the event of disruptions, and disruptions are quickly remedied.
provides an up-to-date picture of the energy system in real time. Data on connected resources in all affected energy and structural sectors and on the network status are analyzed automatically. Situation images for the energy supply provide reliable information
- for the respective sectors of electricity, gas, water
- for energetic and structural sector coupling, i.e. the connection between the networks
- for structural sector coupling, i.e. the connection between the energy sectors with the consumption sectors for household, trade, industry and transport
are the learning data or also the “basic knowledge”. In our case, they include the knowledge of the dispatchers and the complex dependencies between fluid mechanical and thermodynamic parameters, mathematical modeling and legal requirements for controlling the power system mapped in the control system software. The quality of the learning data is crucial for the success of the AI algorithms used.
is a deep learning technique that uses two neural networks for this purpose. One neural network is used as a “generator” and supports the generation of new data instances. The other neural network acts as a “discriminator” and evaluates the data for authenticity. Both networks learn with each other in a double feedback loop.