-
Notifications
You must be signed in to change notification settings - Fork 294
User Guide: Forecasting
Note
This wiki page has been migrated to the following documentation page:
https://docs.openremote.io/docs/user-guide/rules-and-forecasting/forecasting
Please update your bookmarks.
OpenRemote includes a time series forecasting method for attributes, based on weighted exponential averaging, which is explained here.
Additional energy domain specific forecasting for power generation of solar panels and wind turbines are explained in the Tutorial: Create your Energy Management System.
Examples of time series forecasting which allows recognising patterns with a repetitive character, e.g per 24 hour or also per day of the week
In many cases you have data with a repeating daily or weekly pattern. Examples are the temperature in a room (similar every day at the same time) or number of occupied office parking places (similar every weekday at the same time). For those types of time series data, you can apply forecasting for attributes.
To add this forecast and store it, you can add the two configuration items 'Has predicted data points' and 'Forecast'. For the 'Forecast' you will need to add five parameters in a JSON format:
- Type: the method applied, currently only weighted exponential averaging is available ("wea")
- Past period: the length of the period(s) in the past of which values are used to calculate the forecast ("P1D" for 24 hours or "P7D" for 7 days)
- Past count": the number past periods taken into account (can be an integer of 1 or higher, 1 means it takes the value of one 'past period' back)
- Forecast period: the periodic intervals at which the forecasted values are calculated ("PT1H" calculates a data point for every hour in the future)
- Forecast count: the number of data points calculated (can be an integer of 1 or higher, 24 means 24 hours if the 'forecast period' is 1)
An example to be filled in for the 'Forecast' configuration:
{
"type": "wea",
"pastPeriod": "P1D",
"pastCount": 4,
"forecastPeriod": "PT1H",
"forecastCount": 24
}
Note that the definitions for time use ISO8601 standards:
For "pastPeriod" the representation for
- Y is the year designator that follows the value for the number of calendar years.
- M is the month designator that follows the value for the number of calendar months.
- W is the week designator that follows the value for the number of weeks.
- D is the day designator that follows the value for the number of calendar days.
For "forecastPeriod" the representation for
- H is the hour designator that follows the value for the number of hours.
- M is the minute designator that follows the value for the number of minutes.
- S is the second designator that follows the value for the number of seconds.
Source · Documentation · Forum · Issues · Docker Images · OpenRemote Inc.
- Overview
- Bluetooth Mesh
- HTTP
- KNX
- LoRa
- MQTT
- Simulator
- SNMP
- Serial
- TCP
- UDP
- Velbus
- Websocket
- Z-Wave
- Partner Integrations
- OpenRemote as Edge Gateway
- Auto provisioning devices and users
- Connect ESP32 or ESP8266 using MQTT
- Firmware updating with Hawkbit
- Partner Integrations
- Weather data using HTTP Agent
- Connect Your MQTT Client
- ChirpStack LoRaWAN Integration
- Celcius to Fahrenheit with Flow
- Configure mobile app behaviour
- Simulating data in attribute
- Preparing the environment
- Installing and using Docker
- Setting up an IDE
- Building and testing
- Working on UI and apps
- Working on maps
- Docker compose profiles
- Creating a custom project
- Maintaining an installation
- Adding Widgets on Insights
- Working on the mobile consoles
- Connecting Protocol adaptors with Agents
- Edge device firmware updating
- Licensing guidelines for contributors
- Useful commands and queries
- Data migration
- Gateway tunnelling setup