Skip to content

Input Sheet Documentation

Jump to bottom
Trevor Stanley edited this page Aug 17, 2020 · 2 revisions

High-Level Overview:

The input sheet is the primary way the model is configured. It maps the selected drop down options to various tables in the codebase as well as the database that are used in the model run. Additional inputs that are relatively sensitive to change are also included in the input sheet. These include the Solar Bass Diffusion Parameters, ITC rate incentives, and Net Metering. These are all described in more detail below.

Description by Sheet:

Main - Scenario Options:

  • Scenario Name: whatever you want to name the scenario you are going to run
  • Technology: The Beta version of the model only supports Solar + Storage (named as "Solar Only") as of now. Wind will be added at a later date.
  • Agent File: The user defined file must be the same name of the agent file you want to run without the file extension (i.e. '.pkl').
  • Region to Analyze: Drop down options are either "United States" or the name of an individual state.
  • Markets: Drop down options are either "Only Residential" or "Only Commercial".
  • Rate Structure: Drop down options are either "Complex Rates", "Flat (Annual Average)", or "Flat (User-Defined)". "Flat (User-Defined)" is not supported at this time.
  • Load Growth Scenario: Several options are available and link to tables within the "difusion_shared" schema in the database. Available options come from the 2019 Annual Energy Outlook. There is also a "User Defined" option available if the user wants to use their own load growth data/assumptions. This must be formatted in the same way as the AEO scenarios.
  • Retail Electricity Price Escalation Scenario: Several options are available and link to tables within the "difusion_shared" schema in the database. Available options come from the 2019 Annual Energy Outlook. There is also a "User Defined" option available if the user wants to use their own load growth data/assumptions. This must be formatted in the same way as the AEO scenarios.
  • Wholesale Electricity Price Scenario: This option is not supported for the Beta version. Data used is set as default to named table in the database.
  • PV Price Scenario: Drop down options come from the Annual Technology Baseline 2019 Report. A User Defined field is also available. This must be formatted in the same way as the ATB scenarios.
  • PV Technical Performance Scenario: This option is not supported for the Beta version. Data used is set as default to named table in the database.
  • Storage Cost Scenario: This option is not supported for the Beta version. Data used is set as default to named table in the database.
  • Storage Technical Performance Scenario: This option is not supported for the Beta version. Data used is set as default to named table in the database.
  • Financing Scenario: The user can specify the file name (without an extension) of a financing scenario corresponding to those listed in the "diffusion_shared" schema of the database.
  • Depreciation Scenario: The user can specify the file name (without an extension) of a depreciation schedule corresponding to those listed in the "diffusion_shared" schema of the database.
  • Carbon Intensity Scenario: This option is not supported for the Beta version. Data used is set as default to named table in the database.
  • Res/Com/Ind Max Market Curves: Drop down options are "Nems", "Navigant", "RW Beck", and "NREL". Detailed information on and references to these curves can be found on page 23 of this report:https://www.nrel.gov/docs/fy16osti/65231.pdf

The dGen model is calibrated by fitting two parameters of the Bass Diffusion model, p and q. These parameters correspond respectively to the probability of an “innovator” and an “imitator” adopting in a given year. The market share parameter (M) is exogenously stated based on customer willingness-to-pay surveys. The time period variable (t) is not calibrated but solved for as the number of years implied by the Bass model since adoption began. For shorthand this variable is referred to as the time-equivalent, or teq. Future calibration efforts could explore alternate ways to calibrate teq to improve goodness-of-fit.

Calibration of the model was performed at the state-sector level by fitting a backcasted version of the model from 2014 – 2018. A reduced form version of the dGen model is recursively run over different parameter combinations. The grid search is performed using a genetic algorithm that evolves to minimize the difference between the modeled and empirical adoption for a given state-sector.

Based on a literature review of the adoption of comparable consumer technologies, constraints are placed on the p and q parameter bounds of p ϵ [0.0001, 0.002] and q ϵ [0.2, 0.4]. These result in a minimum time to peak adoption of 13.2 years and a maximum of 38.0 years. Without these constraints, the estimated parameters resulted in unrealistic diffusion timescales.

When calculating the backcasted model the market saturation parameter (M) is not uniquely calculated for each backcasted year. Instead the M value for 2018 is used for all of the backcasted periods. In theory using unique M-values for each historic year should yield an improved model fit because it better reflects changing economic conditions. However, in practice the time period-varying M values lead to worse fit, presumably because the M-values are misestimated. Due to concerns of statistical degrees of freedom in the number of backcasted years available, a validation exercise was not conducted but is planned for future work. A table of the calibrated parameters is given below and is also in the input sheet.

Clone this wiki locally