SolarSizer: An Accessible Web Application for Off-Grid Solar Planning

Written by Cassidy Quigley and edited by Trent Dillon

Imagine you want to install a solar energy system at your home. You’d need to answer some basic questions first, right? How many panels would you want? What about batteries? Inverters? How many appliances could it power? And, perhaps most importantly, how much would it all cost?

How would you answer these questions?

You might spend a few evenings surfing the web — but you wouldn’t find much specific to your exact location and energy needs. You could dial up that friend or family member who works in energy — you’d need a personal connection. You could try using a software like HOMER — but this requires (1) money for a subscription and (2) a technical background. You could chat with an energy advisor at Tesla Panels — but this directs you to their products and their quotas, which are intended for wealthy residences, not to mention simply out of reach in many areas of the world.

In fact, it’s not just outsourcing to Tesla that could be costly. Lots of the pathways for learning more about harvesting solar energy demand money, time and privilege. Imagine trying to make decisions about solar energy if you’re working multiple jobs to pay rent? Or if you don’t speak english? Or don’t have reliable access to an internet connection?

It’s not hard to see that there are systemic barriers in who can and can’t (and therefore will!) consider off-grid solar energy and the benefits of cost-savings, energy sovereignty and pollution reduction. To ensure that solar power and its benefits are equitably realized, we need to make it easier to answer the complex and interesting questions about what an installation might look like and require. That’s where our in-house software, SolarSizer, could be impactful in the broader struggle for a just energy transition. Here, we provide some essential background on SolarSizer, including context on the development of solar energy, comparable softwares, and the trajectory of SolarSizer, including the project’s past, present and future.

Why Solar Power?

Due to improvements in the manufacturing process, solar panels have significantly dropped in price over the last twenty years. At the utility scale, the cost of solar energy is currently comparable to natural gas systems and, like other renewables such as wind energy, is carbon free in electricity conversion.

Figure 1. Full cost of a residential PV system since 2010 [1]

Figure 2. LCOE for utility scale energy systems in the United States [2]

But it isn’t just the environmental and cost benefits that make solar energy compelling. Solar energy works incredibly well off the grid. One reason for this is due to the durability and simplicity of solar panels, which, unlike wind turbines, don’t have moving parts and are more secure in extreme weather events. Another reason is due to the widespread availability of solar irradiance. With direct access to sunlight, end users don’t need to import fuels to produce electricity or run long transmission lines, helping communities with financial planning and reducing maintenance requirements. A system that burns fuel like natural gas or diesel will also have volatility in fuel prices that could further complicate financial planning.

To draw a real-world example, these are some of the very same factors that motivated our frequent collaborator, Francis Abugbilla, to consider using solar energy to power facilities and residences in his hometown, Kpantarigu, Ghana. According to Francis:

Ghana, as a tropical country, is ideal for solar energy because of the abundance of sunshine. In fact, solar is well suited for my community because it is located in the northeastern part of the country, where we have sunshine throughout the year. In addition, I see solar as one that democratizes access to electricity because one can start from a small scale such as the provision of solar lanterns and solar home systems, to micro-grids; unlike hydropower or wind power that are capital intensive to afford.

Francis is a great example of a person who often seeks to answer the basic questions mentioned at the beginning of this article.

Importance of Sizing Software

Sizing a solar energy system is not trivial. Both solar irradiance and the electrical load it is being converted to meet (for example, a residence’s energy consumption) vary in time, following daily, weekly and annual trends. A solar energy system will utilize batteries to store energy and meet the load when irradiance is low, but how does one know how many batteries are needed? Furthermore, how many panels are needed? Of course, we want to keep costs and therefore the size of the system as small as possible, but how do we know what’s too big or too small? How do we know whether to save costs on panels or batteries?

If we think about this like a multi-objective optimization problem, where, one objective is to reduce the cost of energy generation (or number of solar panels) and the second objective is to reduce the cost of energy storage (or number of batteries), we can see that a ‘Pareto Frontier’ of optimal solutions emerges. In the schematic below, an ‘infeasible’ solution would be an energy system that provides insufficient power over time and a ‘dominated’ solution would be an energy system that produces superfluous power. Meanwhile, the Pareto Frontier represents a collection of solutions that can be considered optimal given the objectives we are aiming to minimize, ranging from solutions that are heavy on panels and low on batteries to solutions that are low on panels and heavy on batteries.

Figure 3: Pareto Frontier Schematic (source: Xuhan Liu [3])

This is, of course, an oversimplification of a complex and even higher-dimensional optimization. However, it helps to illustrate the fundamental role of solar sizing softwares. Solar sizing softwares run time-based simulations that include energy generation (solar panels), energy storage (batteries) and energy consumption (load profiles) to identify ‘optimal’ solar energy solutions for a specific application. This simulation and optimization is important work that must be done accurately, otherwise a community or individual may not fully understand the cost of the solar energy system they are trying to install, may invest in a system that does not meet all of their needs, or waste money on a system that is larger than what is needed.

For example, take Francis’ case of needing to design a rural computer library with 50 computers, used for schooling during weekdays and occasional use in evenings and on weekends. A sizing software helps to bridge the gap between the purpose of the solar energy system and its design. From the library’s intended use pattern, we can assume hour-to-hour power demand profiles for computers, lighting, and other devices. Software can then, using time-resolved irradiance data at the location of the library, be used to find the power availability of different sized solar and battery storage systems until an arrangement that optimally matches the power profiles is identified, typically with the intention of minimizing cost or wasted energy. This can give the designer or financial planner a better idea of what the solar energy system will look like and cost without the risk of trial-and-error.

Existing Softwares

There are many solar sizing softwares available, but not all of them are adequately accessible.

Homer

Homer, or Hybrid Optimization of Multiple Energy Resources is a powerful software initially developed by the National Renewable Energy Laboratory (NREL) and then became the private company Homer Energy before being bought by UL. Homer can model almost any utility or micro-grid scale energy system with or without energy storage. It can be used for modeling off-grid solar, but if that is your only use for Homer, you are leaving a lot of its value untapped. Its base cost is $187.5 per month, which is significant. While very dynamic in what it can assess, it is overbuilt for off-grid PV systems and is better suited for grid-tied utility scale energy production or optimizing the economics of energy systems.

Homer also requires experience with energy systems and software to create simulations, barring those with no technical background. GRID’s lead officer, Stone Fennel became familiar with Homer and described the intended users as “micro-grid installation professionals, and it excels in that role”. Stone found that Homer’s support team was very useful and eased the process of setting up simulations, but that having a free version for preliminary simulations would be useful and helpful financially.

SAM/PYSAM

SAM, or System Advisor Model is a software for modeling renewable energy systems and energy storage developed by NREL. It is free to download and use and has many online resources that go over modeling PV systems. It includes system components databases for solar panels and inverters so a user can compare energy outputs for different sized inverters and types of solar panels. It also has financial simulations, making it the most complete free software for renewable energy modeling. There are some downsides, SAM will not design a system based on a load profile so a user has to iterate to find a system with the ideal energy output and SAM requires a technical background for the user to be easily set up. SolarSizer uses PySAM, the python interface for models in SAM’s simulation core, to simulate solar energy systems.

PVGIS

PVGIS or photovoltaic geographical information system is a free online tool developed by the European Commision. PVGIS’s spatial coverage is excellent, featuring a nearly complete global dataset, with information on solar irradiance spanning all continents except some parts of Russia, Greenland, and northern Canada. The user can choose between grid connected, tracking, and off-grid solar power systems and selects the system’s location on a global map. PVGIS prompts the user for system information like installed peak power, battery storage, and the software outputs information on the performance of the PV system, including battery state of charge statistics, and monthly graphs of energy put to use and energy not captured (or discarded). Additionally, the user can upload a CSV file with the energy usage per day broken down by the hour. Because it is free and easy to use, an end user could manually iterate through PVGIS, testing various system sizes to find a PV system that meets their needs. However, the search space for this is quite large and an automated software would be much easier to use. Additionally, PVGIS does not detail the system design such as number of panels and inverters or financial estimates, which is a primary concern for most end users.

Review of Existing Technologies

HOMER, SAM and PVGIS illustrate a significant need for free and easy-to-use solar sizing resources. Many sizing softwares require previous knowledge of energy systems, which restricts who is able to use the software and therefore consider and execute solar energy installations. Furthermore, most sizing softwares include multiple renewable energy options including wind, biomass, and solar, and have different modules for grid tied systems and energy storage. Their intent is more-so for utility or multi-generation micro-grid use cases than someone intending to electrify their home or a community facility. On the flip side, softwares specifically designed for nontechnical end uses considering residential and off-grid solar energy systems could go a long way in terms of increasing trust, awareness and equitable adoption of solar energy technologies.

SolarSizer

The history of GRID’s work on SolarSizer dates back to when we first began collaborating with Francis in 2020. Since then, the concept and state of the software have undergone a lot of development.

Past Work

When Francis first approached GRID, he asked for support with sizing the power system for his concept, a solar-powered computer library in Ghana. Our technical team, including Sarthak Jariwala, Ian Murphy, Wesley Tatum, Sophia Votava and John Gannon, began developing the first version of the sizing code. The first iteration was a jupyter notebook that used the PySAM package to model PV systems. The user would manually change the number of panels in a system to find the solution that worked best from the perspective of the user. The goal was to help with the sizing process for Francis’ project and the intended users were engineering students who wanted free sizing software. Since then, we’ve realized that this tool could benefit a lot of people, not just students, and have begun to think about how we can make the software more accessible to a variety of end users, leading to the software’s current form.

What SolarSizer Currently Does

The current code consists of a graphical user interface (GUI) written with DASH, processing user inputs using python scripts, and modeling solar output using the PySAM package that was developed by GRID previously.

Here’s how the code works:

1. The user inputs the latitude and longitude coordinates for their location, which must be within the United States, and a daily load profile in a CVS format.
2. The user’s latitude and longitude are used with the National Solar Resource Database API, linking a location in the United States with an irradiance file.
3. The code processes the load profile into useable data for PySAM
4. PVModel iterates through multiple PV systems of different sizes, simulating each one to evaluate the performance of the energy system.
5. The code computes the uptime percent of each system. This is found by taking the number of hours where the PV system energy output meets the energy required by the load profile and then dividing by the total number of hours.
6. Finally, the code displays an output table in the GUI that includes uptime percent, number of panels, and number of inverters for all of the PV systems being evaluated.

The main upgrade in comparison to GRID’s original version is a GUI, enabling the user to input their latitude, longitude, and load profile. Previously this was hardcoded in a jupyter notebook. Our hope was that a GUI would make it easier for the user to navigate the system without much technical background. This required us to write a handful of functions to process user inputs from the GUI for use in the PySAM model. Another upgrade was using NREL’s National Solar Radiation Database API to automatically find and link SolarSizer with the irradiance file for a given latitude and longitude, which simplifies user inputs. Most of these efforts occurred from the hard work of Cassidy Quigley, Ning Wang, Clayton Sasaki and Lindsey Taylor, who took on developing SolarSizer as a course project for CSE 583, a software development course. Here’s the github repository for the project.

Lessons Learned

The recent updates made to GRID’s sizing code was a challenging process that taught us many lessons about data science and software. The most important lesson was to do extensive research on existing datasets that may be crucial to the software. For SolarSizer that is the irradiance database. Databases for these files are often either easy to use but have limited coverage, or have global coverage but require data cleaning and preprocessing. This made it difficult to select a database that met our needs and was easy to use. Another lesson learned was on the complexity of GUI’s. Our GUI is simple but the code to create it is quite involved, as it is not as standardized as the modeling and data processing codes used for the project. The future of this project may benefit from including students with experience in UI design.

Next Steps for SolarSizer

Although we’ve advanced the code through the addition of a GUI, there is still a lot of room for improvement. For example, SolarSizer cannot be used globally due to its use of a United States dataset, does not consider energy storage which is of course a significant element of real-world solar energy systems, and does not output financial estimates. These considerations are vital for an accessible and equitable solar sizing software. However, the current version of SolarSizer sets a strong foundation toward this vision that we look forward to continuing to build upon.

Here’s a list of changes we want to make soon.

1. Change the irradiance database from NSRDB to PVGIS for global coverage
2. Add energy storage and vary energy storage to optimize the system.
3. Implement component costs to provide financial estimates for the user.
4. Make it possible for the user to create a load profile in the GUI rather than input a CSV file.
5. Improve the simulation process to reduce unnecessary iterations and add resolution to the sizes evaluated.
6. Enable creating a load profile from an energy use survey.
7. Model panel and battery degradation.

These changes will require writing code in python, writing code in dash, tweaking PySAM models, writing tests for code, and stylization checks.

Conclusion

Currently, available sizing softwares do not meet the needs of individuals with limited time, resources, and experience with energy technologies. An individual can use PVGIS to understand how much energy is created based on watts, watt-hours, and battery storage, but it doesn’t quantify those numbers with solar panels, batteries and inverters. An individual can download SAM, but for off-grid solar there is no clear simulation choice from the menu. What SolarSizer strives to do is to take a location, energy load profile either through numbers from the user or a non-technical survey, and then clearly state the number of panels, batteries, inverters, how much it will all cost, and how much area is needed so that they can know if solar is a viable option. In doing so, the software supports the broader efforts for a just energy transition in the midst of the climate crisis and rapid adaptation and carbon reduction measures that, without conscientious effort, will reinforce existing inequalities, most notably along class and global lines.

References

1. Solar installed system cost analysis. NREL.gov. (n.d.). Retrieved April 26, 2022, from https://www.nrel.gov/solar/market-research-analysis/solar-installed-system-cost.html
2. Electricity Generation — Energy Information Administration. (n.d.). Retrieved April 21, 2022, from https://www.eia.gov/outlooks/aeo/pdf/electricity_generation.pdf
3. Liu, X., IJzerman, A. P., Westen, G. J. P. van, Renz, P., Rompaey, D. V., Wegner, J. K., Klambauer, G., Wee, J., Xia, K., Das, S., Scholes, H. M., Sen, N., & Orengo, C. (2021, March 31). Fig. 2 pareto frontier in multiobjective optimization. take Two… ResearchGate. Retrieved May 17, 2022, from https://www.researchgate.net/figure/Pareto-frontier-in-multiobjective-optimization-Take-two-objectives-as-an-example_fig1_344000353