Bill's Weather Station - Cloud-based Instance

Current Obs Last 24h Last 48h Week to now Last 2 weeks Climate Table Climate Graph Web Cam
Weather Underground PWS KNJLITTL12


This is a primitive weather station presently recording only indoor and outdoor temperatures. It consists of a Raspberry Pi and several DS18B20 1-wire temperature sensors. The station software is written in Python, temperature data is logged into a sqlite3 database, reported to Wunderground and this instance of the station software running in the cloud, i.e. on a virtual private server. The sensors include:

Station Data

Current Observations


Static Temperature Graphs

These graphs show data current as of the past midnight and load much faster than the dynamic ones. The current week is generated daily just before midnight. The Current week static graph together with the Last 24 hours dynamic graph covers the current week to now.

Dynamic Temperature Graphs

These graphs show data up the the present. They require a bit of patience proportional to the length of the graph and can take several tens of seconds. The RPi is a great machine but for generating graphs with matplotlib it is a bit slow.

Climatological Data

The climatological data includes temperature averages, lows, highs and degree-days by year, month and day in tabular form and averages, lows, and highs for every day in the year by year in graphical form.


Gas Consumption (Experimental)

I am interested in recording natural gas consumption with the thought that the data would help identify thermal parameters of my home, areas to improve thermal behavior, and provide feedback for changes. I hope to correlate gas usage with meteorological parameters including temperature and wind. Many years ago and in a different home I made daily recordings by hand of the gas meter during one heating season. That data showed clearly the benefits of keeping the door between the basement and the first floor closed. I'm no longer up for manual recording and daily readings lack sufficient resolution for the usage I want to study and so I looked for an automatic solution.

It turns out that my gas company reads the meter from a vehicle on the street. If they can read it I though I could as well. The technology of Automatic Meter Reading (AMR) is well described online and others have already developed both open-source and proprietery products for this using software-defined-radio concepts. It was a simple matter to glue a few existing parts together to receive the broadcasts from the meter and then write a little software to record and display it.

My gas meter, an Itron 100G DLT, periodically broadcasts Standard Consumption Messages (SCM) in bubble-up mode, a peculiar term that means it broadcasts spontaneously, not on demand from a reader. The messages are received with a RTL-SDR dongle (an inexpensive DVB-T receiver containing a Realtek RTL2832U demodulator and R820T tuner) connected to a Raspberry Pi 2 B+. The RPi runs a program known as RTLAMR (written by Douglas Hall and generously contributed to the open-source community) to decode the meter signal and convert it to JSON text. A Python program reads the output from RTLAMR and send it to the cloud server for storage and presentation. The processor load on the RPi 2 B+ is 100%. It appears to drop some packets but works well enough to provide the meter readings at an adequate sampling rate, typically once every few minutes. The capture process is stochastic in nature because of intereference from other meters in the neighborhood and the frequency-hopping nature of the transmissions.

The graphs linked in the following table explores the option space of graph period, histogram bucket size, and measurement units. This is experimental to help understand the most useful display formats. The % Max column is specific to my Weil-McLain ECO 155 boiler with a maximum input of 155,000 BTUs/hour. It shows the actual usage as a percentage of the maximum input to the boiler. The cost column includes only components proportional to usage (gas and delivery), excludes the fixed Residental Customer Charge. Note that all of the columns display identical data. The graphs differ only in the units and scale factor of the Y-axis.

Period Cubic Feet BTU % Max Cost
Last 7 days by hour Show Show Show Show
Last 14 days by hour Show Show Show Show
Last 48 hours Show Show Show Show

Related Factoids

Future Work

I hope to add more sensors later including barometric pressure, wind speed an direction, and possibly more, depending on availability and cost.