Energy demand and Solar production for all 6 buildings and 6 solar panels are forecasted in order to get the data used to find the optimal battery and activity schedule. They are calculated using the historical 15-minutely energy demand of buildings and power production of solar panels from Monash University, Melbourne, Australia and the Melbourne weather data.
Models that we decided to use are lasso and random forest. Out of all these models, we found out that the best model for this dataset is random forest. The MASE calculated of both the energy production and solar production for our model is 0.80 which puts our team in 5th position of the IEEE-CIS competition.
This is a display of a single instance being scheduled, which includes batteries and activities (lectures) for the month of November 2020. This is calculated by taking the instance text file and electricity price data from AEMO, and minimising the cost of the schedule (based on energy usage).
To find solution, TinyGrid uses Mixed Integer Programming and the package Ortools by Google.
Description
What is TinyGrid?
TinyGrid is a project to develop an optimal battery schedule and an optimal lecture schedule in Monash University Clayton Campus, based on predictions of future values of energy demand and production. This problem was presented in 2021 by IEEE as a competition, and in 2022 a team of four students from Monash University took on the challenge retrospectively; TinyGrid is our solution to the competition.
Why is this needed?
One of the most important measures to deal with climate change is through the use of renewable energy such as wind and solar. However, renewable energy cannot be produced on command, which means energy storage is required. At the same time, storing energy is very expensive and may cause a loss of energy stored. Therefore, we need to forecast the energy demand and production from renewables, in order to optimally schedule energy storage such as batteries. This enables an optimal charging and discharging schedule, based on the required demand.
This is where our project TinyGrid steps in. TinyGrid at its core has a prediction module that predicts solar energy production and building energy usage from the Monash Microgrid system, and an optimization module which schedules building activities and batteries to minimise energy wastage. This is presented as a website dashboard that lets users interact with figures displaying forecasting and optimization solutions.
Weather data
The daily minimum temperature (C), maximum temperature (C), rainfall (mm) and solar exposure (MJm^2) of three weather stations near Melbourne: Olympic Park, Moorabbin Airport and Oakleigh (Metropolitan Golf Club) from BOM. Each weather series starts from 1st of January 2016 and contains corresponding values until November 2022). This is needed to forecast the solar power production.
Electricity price data
Australian electricity price data from AEMO to achieve the lowest energy cost when scheduling the activities.
Historical energy demand and power production
Historical 15-minutely energy demand of 6 buildings and 15-minutely power production of 6 solar panels from Monash University, Melbourne, Australia. Given as a time series format to forecast the energy demand and power production in October and November 2020.
Problem instance
Instances of buildings, solar panels, batteries and activities that will be optimised. This is given in a text format with the first line being ppoi (predict-plus-optimize instance) which is the summary of the problem instance. It also has Buildings with attributes id, number of large rooms and number of small rooms; Solar panels with attributes solar id and building id; Batteries with attributes building id, capacity (kWh), max power kW and efficiency; Reccuring activities with an activity object and precendeces; Once-off activities with an activity object, value($), penalty($) and precedences; where activity object has attributes id, number of rooms, room size, load (kW) and duration.
Forecasting
Optimisation
Adjust slider to view scheduled activities & grid load for different days.