Project

ThreatSim: Decision Support Threat Simulator (DSTS)

Goal: We realized from our work in oil and gas that the most valuable type of systems to provide decision support were systems that provided the arch from knowledge of symptoms based on real-time situational awareness to optimized actions to address the symptoms. There is a type of machine learning good for that type of optimization called reinforcement learning (aka approximate dynamic programming). In the original TreatSim concept, we used a (power flow) simulation of the electrical grid infrastructure coupled with reinforcement learning to learn how to avoid or mitigate threats to the grid, natural or manmade. TheatSim can be used for various other infrastructures such as water, transportation, pipelines, etc.

Updates
0 new
5
Recommendations
0 new
0
Followers
0 new
0
Reads
0 new
86

Project log

Roger N. Anderson
added a research item
Even the best oil companies do not optimize their operations for maximum efficiency, let alone for maximum profit. People, processes and manufacturing must be wired together to produce such efficiencies and profitability. The image below represents our vision of a next generation of 4D Business Control System (BCS) for the upstream oil industry. Downstream and Power segments of the energy industry have well developed BCS's, but new technologies reflected in these have not yet made it into widespread use in the upstream. Computations necessary for risking and re-risking, real options, management-by-exception, portfolio and task management, and balanced score carding (the "upper arch of a business hierarchy) are calculated in most business units, but are generally not integrated in real-time and pushed to field sites (the lower arch) so that the efficiency and profitability can be optimized. vPatch Technology, Inc. has produced this next-generation 4D BCS that integrates SCADA data, analogs, models, and projections to produce dynamic cross-system optimization while anticipating equipment failures and other operational problems making use of data mining and visualization tools. The 4D BCS of the future that we envisage automatically ranks, guides, and warns by the use of "hotlists" customized to incentives for people with critical roles in the organization. Hotlist content is packaged on the web so actions are one click away. Hotlists are based on light weight, peer-to-peer web services, and they utilize task management, rule-based decision-making, decision trees, neural network agents and "credit-mapping" based on the Suitability Matrix (sm) to provide the action generator for the 4D BCS. In order to accomplish the computational tasks required by the next-generation 4D BCS, several inputs and outputs must be provided. Firstly, all important business processes must be brought into real-time two-way communications so that feedback control can work. This involves proper SCADA data feeds from internal and external generation facilities, combined with data such as strip prices, data-basing involving data historians and data warehousing, software application integration based on middleware and information dissemination based on the web, mobile devices, and information "push" based on lightweight peer-to-peer computing. Secondly, proper use must be made in real
Roger N. Anderson
added a research item
MY VITA as of June 1, 2020, with newest Patent Numbers issued for 5 new Continuation Patents.
Roger N. Anderson
added a research item
A Power Control System (PCS) of unprecedented efficiency and security will be required to redistribute electricity quickly and without interruption within North America as the system comes under more and more stress in the future. Terrorism is posing new threats to this most vital of economic infrastructures-the delivery of cheap electricity to the nation at all times. In the aftermath of 9/11, we at Columbia asked what tools and technologies we have that might help the national cause. We have previously developed automation and integration systems for two historically more "threatened" industries-international oil and gas production on the one hand and Internet network failure detection on the other. A new generations of American engineers and managers must be trained in electricity production and distribution under duress. We propose to create a PCS Simulator that will train in the complexities introduced by terrorism, combined with the coincident convergence of supply and demand across the electricity grid of North America. Such threats, if sustained over a period of time, will drive electricity PCS's to more and more non-linearity, causing breakdowns that have not been foreseen from previous experiences with the linear systems of today. A future workforce must be trained to cope with this uncertain future. In this proposal, we are concerned with the creation of a PCS Simulator to train the new management and engineers who will transform that next generation of decision-making. Put quantitatively, automated failure detection, combined with "make-it-so" diagnostics-to-solutions credit assignment, is a non-linear inverse problem that we seek to build a simulator to teach operators how to solve. The consideration of the integration of technologies required for this cross-system optimization problem will require an unprecedented degree of interdisciplinary collaboration in and of itself. "Electric power systems are prototypical socio-technical systems, meaning that their technical, social, economic, political, and cultural elements are tightly interwoven and impinge directly and forcefully upon each other. The national goal is to build cross-interdisciplinary partnerships, which allow more unified, coherent research to ensure reliable, secure, and efficient electric power networks" …..NSF/ONR solicitation NSF-02-241 In contrast to the Grand Challenge posed by the above National Science Foundation/Office of Naval Research collaborative research program, the current Power Control System (PCS) of your local electric utility and regional ISO (Independent Services Operator) is frighteningly simple. Since consumption is not known second-to-second, a PCS computer merely balances the spin of power generation turbines under its
Albert Boulanger
added an update
Kartikeya Upasani'sThreatSim Final Report Fall 2016
Anant Sharma's Final Report on DeepFlow (power flow using deep learning) Fall 2017
 
Albert Boulanger
added an update
Here are some links to ThreatSim:
ThreatSim Concept: https://bit.ly/2JR0UO9
ThreatSim was described in this Smart Grid Whitepaper
 
Albert Boulanger
added a project goal
We realized from our work in oil and gas that the most valuable type of systems to provide decision support were systems that provided the arch from knowledge of symptoms based on real-time situational awareness to optimized actions to address the symptoms. There is a type of machine learning good for that type of optimization called reinforcement learning (aka approximate dynamic programming). In the original TreatSim concept, we used a (power flow) simulation of the electrical grid infrastructure coupled with reinforcement learning to learn how to avoid or mitigate threats to the grid, natural or manmade. TheatSim can be used for various other infrastructures such as water, transportation, pipelines, etc.