Skip to content

Home

OpenDSM OpenDSM

OpenDSM: An open-source python package to develop and implement standard methods for predicting metered energy usage.

PyPi Version Supported Python versions License Code Style

OpenDSM (formerly OpenEEmeter) is an open-source library (source code) used to measure the impacts of demand-side programs by using historical data to fit models and then create predictions (counterfactuals) to compare to post-intervention, observed energy usage.

Key Features include:🔗

  • Open Source: All code is open source, making OpenDSM an excellent choice for measuring energy savings when results must be transparent and reproducible.
  • Fast: A key tenet of OpenDSM modules are to be highly efficient, critical when there are millions of meters to be modeled.
  • Easy and Intuitive: The API interface is inspired by scikit-learn, a well-known data science package for building models. Just fit and predict.
  • Pandas DataFrame Support: Input data and predictions use pandas DataFrames, a well-known format for data scientists and engineers.
  • Model Serialization and Deserialization: Models can be serialized into dictionaries or json objects, saved, and deserialized later.

Core Modules🔗

OpenEEmeter🔗

Create long-term models fit on historical data to generate predictions of energy consumption

  • Models to Fit Your Data: Billing, daily, and hourly models, depending on the time-series resolution
  • Input Data Formatting: Meter usage and temperature time series are input to models through data classes to ensure standardization and avoid common pitfalls
  • Data and Model Sufficiency: Native sufficiency checking to verify measurement compliance with the approved methodology

DRmeter🔗

Create short-term models fit on historical data to generate predictions of energy consumption

  • Hourly Model: All demand response models require at least hourly resolution data

Comparison Groups🔗

Assign comparison groups (CGs) to correct OpenEEmeter models using non-participant population through one of the following methods

  • Comparison Group Clustering: Cluster on model error profiles to select unique CG for each treatment meter
  • Individual Meter Matching: Create population-level corrections by choosing the nearest n meters using Euclidean distance
  • Stratified Sampling: Select meters for CG based upon shared characteristics (outdated methdology)

* Planned feature in future OpenDSM versions

EEweather🔗

Get the most appropriate weather data for a location

  • Match Location: Select a weather station using latitude/longitude or ZIP code (ZCTA)
  • Climate Zone Sensitive: Ensures that the selected weather station is within the same climate zone
  • Reliable Data Sources: Utilizes US Census Bureau, NOAA NCDC, and NREL as primary data sources

* Planned integration in future OpenDSM versions