This project has developed an Agent-Based Model (ABM) of solar panel adoption in Darwin using the results of a Darwin-wide survey to parameterise key adoption variables. Agent-based modelling has increasingly been used to explore the diffusion of technology and ideas in communities as it is well-suited to explore the stochastic heterogeneity and dynamic interactions between individuals or households. An ABM allows the emergent phenomena in a system to be better understood by modelling a realistic decision-making process of “agents” considering their social interactions and the environments changing over time and space.
A spatially explicit ABM model was developed through this project for the Northern Suburbs of Darwin with solar panel installation mapped using high-resolution imagery. Houses without solar panels were allocated preference to adopted attributes. One of these attributes includes a neighbourhood influence. Within the model, households react to neighbouring influence creating hot-spots of installation. This is also affected by feed-in-tariff and other adoption variables.
ABM models are rarely considered predictive rather they are designed to explore patterns, process and interactions between variables. This is the case with this model as a small scale trial to enable the dynamic visualisation of solar panel adoption and known influences. There are known data deficits in the model as it is.
How it works
The model responds to three household/resident attributes:
- Likelihood to install
- Likelihood to be influenced by neighbors
- Time to install once a decision has been made.
These are set within the model with the values derived from the survey results.
Neighbourhood influence is a result of the pre-set 'likelihood to be influenced by neighbours' variable and is modified by the number of surrounding solar installations. Surrounding installation influence is calculated at three levels of proximity with decreasing influence with distance as follows:
- Next-door neighbours
- Close neighbours
- Nearby neighbours
The likelihood to install is also influenced by the installation cost. This was found, through the household surveys, to be directly related to the time it takes to pay off the installation, which is in turn related to how much solar power is actually used on-site. Solar power directly used by the house results in reduced electricity consumption from the grid and therefore a decrease in repayment time. The default value is set to 50% of electricity being used on-site. This value can be modified via a slider.
In addition a proportion of dwellings that are either rental or units are unlikely to have solar installed. By default 20% of houses, without solar already installed, are assigned as rental and will not get solar installed. This value can be modified via a slider.
The flow chart below shows the variables and processes moderating the uptake model.
How to use it.
Before you run the model set the household and global variables:
Global Variables
- Increasing the Feed-in Tarif increases the Likelihood of installation
- Decreasing the installation cost increases the Likelihood of installation
House Variables
- Increasing the average proportion of solar used in on-site increases the Likelihood of installation
- Decreasing the installation cost increases the Likelihood of installation
Press the set-up button to initiate the model.
When you first load the model it will display dwellings attributed as follows:
- Blue is rental
- Grey houses no install
- White houses are soon to install
- Red houses have solar
- Orange - new install
To run the model press the 'run' button.
Modeling Software
These models were created in Netlogo modeling and simulation software. This is free open-source software that can be downloaded and installed on any computer to run the models described.
THINGS TO NOTICE
Note the amount of time under different variable settings that it takes for the number of houses with solar to be more than those without.
THINGS TO TRY
View the Solar concentration. This displays a heat map of solar panel uptake. Areas of already high uptake should continue to grow under the neighbour influence model.
EXTENDING THE MODEL
The model could be developed to include:
- More accurate lot-level attribution of dwelling type to differentiate those lots that, such as multi-level units that are unlikely to install solar.
- Include suburb level estimation in the proportion of houses in the rental market or government houses that are unlikely to install solar.
- Incorporate the influence of battery installation.
- Incorporate the influence of public education campaigns.
- Include a larger proportion of Darwin.
CREDITS AND REFERENCES
Developed by Rohan Fisher (rohan.fisher@cdu.edu.au) and Dr Kersitn Zander (Kerstin.Zander@cdu.edu.au) with funding from the Northern Territory government.