1 of 13

P3 SOLAR MINI-GRID PROJECT

David Moy | Bardia Zadeh | Andres Chamorro

Ashrith Bandla | Toshihiko Sato

2 of 13

Project Chronology

Problem Statement

The design of mini-grid powerline distribution networks presents a significant opportunity for optimization. In these networks, the locations of the generation plant and all customers are known, along with regulatory constraints such as the maximum allowable distance between powerline poles and the maximum distance from a pole to a customer's home or business.

The challenge is to design the most cost-effective network that connects all customers to the generation plant while adhering to these constraints. Currently, this process is performed manually, with no effective tools available for network optimization. Developing such a tool provide immense value to the industry.

3 of 13

Project Chronology

Course Principles

Sustainability has been the key principle embedded throughout our project reflecting:

Our deliverables are based on tools accessible by anyone without having to pay any license fees (proposed solution: Google Colab)
Our deliverable are based on methodologies that are widely known to ensure reproductivity

4 of 13

Project Chronology�Initial Group Steps

The members of the team had indicated they were interested in this project. The Professor assigned the members to the project. The members met and discussed their strengths and which parts of the project they desired to work on. Based on this, Bardia and Andres were assigned to work on UI/UX, David and Ashrith on backend/algorithms, and Toshi on project management. At this time possible solutions and algorithms were researched that could solve the given problem description at the time.

5 of 13

Project Chronology�First Meeting With Partner

The next step of the project was meeting Mr. Nicholas Selby to discuss the project details. He described the problem in more detail than what was provided in the text form during group formation. He also described a voltage drop calculation problem, but the team decided to focus on the network optimization problem first, then address the voltage drop problem with time permitting. He sent a copy of a thesis where a student tried to solve the problem for further study. It was agreed upon the final deliverable would be a Notebook (Google Collab or Jupyter).

6 of 13

Project Chronology�Initial Problem Investigation and Team Website Creation

Backend team started to investigate what algorithms could help solve this problem. Some algorithms explored were Prim’s Minimum Spanning Tree, Kruskal’s Minimum Spanning Tree, and Steiner Tree (handles insertion of new nodes). These algorithms were tested with randomly generated coordinates. Parallely, the UI/UX focus team members started working on the team website.

Shown here is the initial prototyping with MST and random points and the team website.

7 of 13

Project Chronology�Second Meeting With Partner

We met with Mr. Nicholas Selby a second time to show progress (see previous slide). He gave feedback that the focus of the initial solution should be the insertion of new points (poles) to cover the existing points (homes). He also gave the actual constraints of maximum of 50 meters between poles and maximum of 20 meters from pole to home. He also provided a real world location (Koyamo Island, Kenya) where this was done. Looking at the location in Google Maps, we could see where the actual home locations were. Clicking on the roofs gave the latitude and longitude coordinates and these were used at a further point in the project.

8 of 13

Project Chronology�Further Algorithm Investigation

Given the new constraints, further algorithms were tested to attempt to place new points to provide coverage to existing points. Work was also done to enforce a maximum distance of 50 to long connections by repeatedly splitting them until they were shorter than 50.

9 of 13

Project Chronology�Real Coordinates and Further Algorithm Investigation

Previously, random coordinates were being generated. We switched to coordinates obtained from Koyamo Island to provide real world test of the algorithms. Testing was started for different algorithm to find coverage, the one that was selected was KD Tree (K-Dimensional Tree).

10 of 13

Project Chronology�Final Meeting With Partner

In the final meeting with Mr. Nicholas Selby, he gave the feedback that the KD-Tree algorithm for placement optimization wasn’t sufficient. He suggested seeding the entire area with candidate points and selecting the best ones and removing the worst ones. He also said the MST algorithm for connections did not find shortest distances when there is a triangle that could be formed.

Example of problem with MST.

The red is shorter distance.

Example of problem with KD Tree Algorithm.

There is too much overlap among red circles.

11 of 13

Project Chronology�Final Algorithm Investigation

With the feedback from Mr. Selby, investigation was done on how to use candidate points to find the optimal pole placement. An optimization library from Gurobi was discovered but could not be tested on the real world data due to rate limit from free license.

Example of Gurobi optimization library.

Note that coverage was not 100%

The triangle problem was attempted to

be solved using networkx library but no

final optimized solution was obtained.