Project Background

This project is a look into the behavior of distributed grids which there are both aggregators, producers and consumers. Today, the grid is primarily uni-directional - meaning the electricity is produced by producers, and consumed by consumers. If managed properly, enabling production by grid nodes which are historically consumers could lead to more accessible, resilient and efficient electrical grid system.

For this analysis, a Monte Carlo simulation is run with the following rules:

1. There are two aggregators which sell/buy electricity
2. The aggregators buy/sell to all nodes directly connected to them and may also buy/sell to other aggregators they are connected to
3. Prosumer nodes (resembling houses with solar or factories with generation capabilities) sell/buy only from an aggregator they are connected to
4. Prosumer nodes are connected to only one aggregator
5. A transmission capacity is set between the aggregators

What was explored

Following the rules above, we observed simulated data which models the behavior of a de-centralized grid system. Specifically, a node may be a buyer and/or seller, but in this case we build in a bias to the two different ‘communities’ in terms of proportion of buyers and sellers. One of the aggregator nodes has mostly buyers and one has mostly sellers. So, when a node requires energy, the node messages their aggregator and requests a given amount of energy based on their individual price elasticity. When the price is high enough, more nodes will send a message to their aggregator declaring they are willing to sell electricity, and how much they can sell. In a world where all messages are delivered with no fraudulent or mistaken information, the price and quantity equilibria should remain stable, and the transmission of intra- and inter-community should never experience a transmission overload. However, we don’t live in a perfect world, and attached to the messages in this simulation is a random probability that the message is lost, simulating a denial of service attack.

What was found

Comparing arbitrary percentages of messages lost at random, we can see the possible implications of poor cyber-security protection. Interconnected IoT devices which can automatically buy/sell energy to the grid may be helpful conceptually, but increase the potential attack surface for grid-wide cyber threats. We can see that even small percentages of messages lost at random can result in huge gaps in market price and demand information, which can lead to grid-wide overloads.

As seen below, on a dual-aggregation grid with a transmission capacity of 1800 MW, when 10% of messages are lost at random, nearly half of the total transmission simulations lead to an overload.

Though we may simply be able to increase transmission capacity, even this solution cannot stop the cascading effects of both an artificial decrease in price coupled with an artificial increase in demand or vice versa.

Link to GitHub Repo