BACHELOR ASSIGNMENT
There is an ongoing increase in the electricity usage of the average household due to the addition of heat pumps and electric vehicles. Government incentives such as ‘net metering’ combined with the decreasing cost of PV panels have led to an increase in their installations. Both these factors now lead to a highly congested grid operating at its peak capacity in many regions of the Netherlands. For example, no new connections can be added onto the grid in places such as Amsterdam and Flevoland, to name a few. One solution is to reinforce the grid. However, making infrastructural additions to the grid can be an expensive and slow process, partly due to lack of personnel. Therefore, a second and more preferred solution is to focus on the ICT layer of the electric grid and use smart energy management algorithms on the demand side (or user side). Such demand side algorithms are designed to intelligently schedule devices and plan their consumption profiles such that power peaks may be avoided and more sustainable energy can be integrated and directly utilized. This branch of research is called Demand Side Management.
There are many algorithms developed in the field of Demand Side Management. One such algorithm developed in the Energy Group is Profile Steering [1]. Profile Steering is a decentralized hierarchical heuristic based on steepest gradient descent principles. It is used to iteratively build a day-ahead planning for devicessuch that the aggregate profile matches a pre-determined desired profile as closely as possible. The devices in general can be abstracted as child nodes that iteratively work on submitting better profiles to a coordinatornode, which manages the process of selecting profiles to be added into the aggregate and steering the aggregate profile towards the desired profile. This lends the hierarchical decentralized nature to the profile steering algorithm. A live example of the algorithm can be seen in [3].
In its current version, the coordinator makes no distinction between different devices or child nodes when asking for their profile submissions. To take an example, imagine a house, with the smart meter being the coordinator node. Currently, the smart meter (coordinator node) would execute a random optimization order with the house’s devices (child nodes). However, we believe that better and faster results could be achieved by applying some logic to the optimization order. For example, would it be better to first ask the electric vehicle to optimize and send its profile and then the washing machine? Or the other way around? An electric vehicle is known to offer higher flexibility than a regular white good. Should the optimization order then be designed as per device flexibility? Furthermore, different structural representations (e.g., plan all electric vehicles simultaneously) could potentially also increase the performance of the algorithm.
Methodology
a) Understand the working of the current profile steering algorithm implemented in our current in- houseopen-source energy management software DEMKit [2]. Use the default random optimization order to build a baseline case.
b) Build different optimization orders based on various logic (such as device characteristics or grid scenarios).
c) Evaluate the quality of the aggregate and the performance of the optimization process for each these optimization orders w.r.t the baseline case.
d) Produce a conference-paper-like report, including the methodology, results, and conclusions of the study.
Background
[1] M. E. T. Gerards, H. A. Toersche, G. Hoogsteen, T. van der Klauw, J. L. Hurink and G. J. M. Smit, "Demand side management using profile steering," 2015 IEEE Eindhoven PowerTech, Eindhoven, Netherlands, 2015, pp. 1-6, doi: 10.1109/PTC.2015.7232328.
[2] https://www.utwente.nl/en/eemcs/energy/demkit/
[3] https://wwwhome.ewi.utwente.nl/~gerardsmet/vis/ps/
Workload
Theory: 25%
Coding: 30%
Evaluation: 25%
Writing: 20%
Contact
Supervisor (CAES Group): Aditya Pappu
Coordinator (CAES Group): Gerwin Hoogsteen
