He is one of the handful of researchers I have met who manages to conduct research that has both academic merit and significant social impact. However, the most I learned from Mani is how to be a good person. I spent days banging my head against the wall on how to solve the problem, but I inevitably left his office at the end of the week with a clear sense of the next direction.
I would also especially like to thank my main collaborator and partner-in-crime, Desmond Cai. It was one of the most intellectually stimulating experiences I've had, listening to everyone talk about their work in our group lunches. The rapid increase in residential photovoltaic (PV) use in the last half decade has created a need in the electricity industry for a broadly accessible model that estimates PV adoption based on a combination of various business and policy decisions.
The app allows users to experiment with different customer demographics, rate structures and subsidies, allowing them to tailor the app to the geographic region they are studying. This study then demonstrates the different types of analysis possible with the application by examining the relative impact of different policies related to rate structure, fixed costs, and PV pricing on PV adoption.
Introduction
- Data on residential customers
- Customers’ savings from solar PV
- Methodology for variance tests
- Results of variance tests
- Diffusion model for adoption
- Fitting diffusion parameters
Our model is based on an extension of the well-established Bass diffusion model for technology adoption. To our knowledge, there has been no previous work on the impact of financial savings through PV on adoption rates. There is also relatively little work studying the impact of financial savings through PV adoption on the utility "death spiral" [5, 17].
We extend a previous study of the utility "death spiral" [5] by analyzing historical data and explicitly factoring in financial savings when studying this feedback effect. Detailed descriptions of all the functions and the sixty-six PRIZM clusters can be found in [12]. We then calculate the expected future utility bills based on the electricity prices at the time of installation.1.
We estimate the NPV of the purchase costs using the size and installation date of the PV systems. In particular, we obtain PV prices (in $/kW) paid by residential customers at SCE between 2007 and 2012 from the California Solar Initiative (CSI) database [9] and fit a linear model to the data for estimated PV prices based on installation date. However, there is no information on the size and costs of the PV systems that these households would install.
Furthermore, this sample variance is maximized when exactly half of the customers in that bin approve (Mk = 1/2). To provide a benchmark, we also calculate the adoption variance of the original data without any segmentation. This is because each subsequent feature is a weaker predictor of adoption (based on the results in Table 2.2.2.
PCA confirmed the results of variance testing and indicated that economic savings and income were the most important factors. Studies have shown that the prevalence of a new technology has a significant impact on the level of adoption. G is the population size (adopters and non-adopters), and pi and qi are constant parameters.
The constant pi can be interpreted as the coefficient of innovators: customers who adopt PV regardless of current penetration. Konstantaqi is the coefficient of imitation: customers who adopt PV based on the proportion of customers who have already adopted.
Simulation Runs
- Description of different rates and policies tested
- Results of simulations
- Strengths of Model
- Weaknesses of Model
- Applications
- Future steps
The meanings of the columns are as follows: Levels: Model input: Number of levels in the tariff; Tier 1 Final/Highest Tier Final: Price of the lowest-priced tariff/highest-priced tariff in the final year of the model; Fixed Fee. The second row called "PV Drop" shows the impact of a significant decrease in the cost of PV with prices falling at 10% per year, down from 5%. The third row, "4T Fixed" shows the impact of the introduction of a fixed connection charge of $10 per month in 2015 for all customers, independent of the kWh consumed.
The sixth row "2T Ratio" uses a different rate structure than the rest of the series with a fixed ratio between the 2 rates instead of a fixed price differential. The seventh and eighth rows, “Q halved” and “P halved,” show the sensitivity of the model to two tuning parameters adjusted from historical acceptance data by halving the q and p parameters, respectively. The ninth row, "No Revenue Escalation," shows the impact on adoption if we ignore the increase in revenue escalation typically imposed by utilities.
Fifth, it appears that the impact of the -10% annual decline in PV prices is as large as the impact of expected rate changes in California, although the effects are in opposite directions. The impact of a 10% annual decrease in PV prices (compared to the 5% baseline) increases the number of PV users by 50% in October 2018. Regarding the sensitivity of the p,q fits, we see that halving the p-value (reducing the share of innovators by half) has a negligible impact on the adoption rate.
The base diffusion model variant used in the model, which has an embedded savings equation, has several advantages:. This is very important in a state like California, where the California Public Utility Commission's (CPUC) initiative to reevaluate residential rate structures [2] has led to a transition period where new rates are being experimented with without a firm understanding of the effects of those changes. By changing the required inputs, the impact of adoption can be modeled at the service territory, state, or national level.
The advantage is that once the system area is appropriately defined (eg number of customers, average sunlight received, etc.), the model can be run several times very quickly with different tariff structures and PV prices. In addition, it would be ideal to add. The model also lacks other features that are likely to become important going forward: (1) commercial customers, (2) electric vehicles and (3) energy storage for a more comprehensive overview of the dynamics involved in the grid going forward. Modeling the impact of power storage and commercial customers is less clear due to the lack of clarity in the future efficiency of power storage devices and the diversity in the types of customers respectively.
The impact of rate design and net metering on bill savings from distributed pv to residential customers in California. Customer Economics of Residential Photovoltaic Systems (Part 1): The Impact of High Renewable Energy Penetrations on Net Metering Electricity Bill Savings.
