Get access

The Blog

Good and bad weather stations

Hugo Birkelund - 24 October 2017 13:55

Instruments for climate analysis Solar-powered weather station at Fort Caroline National Memorial along the St. Johns River in Jacksonville, Florida, USA.jpeg

EQ research shows: Removing “bad” weather stations substantially improves the quality and stability of power consumption forecast systems. 

In general, you need to find the correct weather stations if you want to succeed in modelling the weather-driven parts of the power market. Using a sophisticated model, but skimping on included stations, could even make matters worse. Even if you want to report something as simple as the average temperature in an area, this is a golden rule. If the weather stations are ill-chosen, the spatial-averaged temperature will be far off  whatever you want it to represent. And the inherent error will be passed on to subsequent models that use temperature, or any weather parameters, as input.

Be it consumption, hydrology, hydro production, wind power SPV, or any of the models mentioned in the list beneath.

But how to choose the correct stations? And why not use all available stations?

Some findings from EQ researches on our new consumption system may shed some light on this.

Consumption and needed input

As a rule of thumb, modelling power consumption is all about capturing reoccurring patterns. A bit oversimplified, this modelling task may be parted in capturing two effects: One part of consumption is determined by the social patterns. The other is dependent on how humans adapt their power demand to the climate in general, and current weather in particular.   

Social pattern: Humans' adaptation in terms of power consumption to e.g. time of the day, work pattern over the week, vacations and moving holidays.

Weather/climate: Created by earth orbiting the sun, and how individuals have adapted to local weather conditions for e.g. temperature, overcast/daylight, humidity, wind-chill, heat-radiation, etc.

Capturing the climate and weather-induced part of the power demand, rely on being able to forecast the weather and affecting factors with sufficiently high spatial representation and accuracy. And this is where weather stations come into the question.

The spatial part is solved by adding what we term a sufficient number of weather stations. However, each weather station added incurs costs in terms of time and money spent on maintenance. So conventionally the business standard has been to limit the number of stations included. One might call it the strategy of good-enough.

The EQ approach

EQ acknowledges that the time for this strategy may belong to yesterday. Increased request for higher accuracy, more focus on understanding and forecasting local conditions, and the requirement for higher time resolution, all points to this.

So, setting up our new consumption model system EQ adopted the strategy of using all available weather stations that fulfil quality criteria. In effect, we have removed what we have termed “bad” weather stations. “Bad in this context is one station, either lacking or showing a poor update of actuals, or simply one located far away from where power consumption is decided.  

Following this modelling strategy has:

  • Substantially improved the quality and stability of the new EQ power consumption system
  • Shortened the lag time for actual data, meaning time from now and until last updated data from a station shows on our web
  • Led to a massive increase in included stations, compared to what we believe is conventional for the business 
  • Still, about 30% of all available stations in Europe were removed

What do you think about this?  Mail us your feedback

that critically rely on choosing correct weather stations.

Temperatures, Wind speed, Wind direction, Precipitation, Moist, Evaporation, Air pressure, Cloud coverage, Solar radiation (array of types), Snow

Power demand, Wind power production, Photovoltaic production, Precipitating energy, Wind-Chill index, Heating index, Cooling ondex, Overcast index, Cloud covers, Run of river production, Reservoir production, River temperatures


Topics: Weather stations- Consumption modelling- Wind Power- Photovoltaic production- Wind-Chill index- Heating index- Cooling index- Overcast index- Reservoir production- River temperatures- Run of river production

Hugo Birkelund

Hugo Birkelund

Cofounder and CEO at Energy Quantified. He has since 1999 held various leading positions within development, marketing and sales of analysis platforms for NTE/PointCarbon and Markedskraft/MKonline. Before that he worked 10 years with research at ENRI and Statistics Norway. By education he holds a degree at the University of Oslo specializing in econometrics and mathematical economy. Hugo is the go-to person for questions related to EQ’s development, forecast models and data feed.

Previous Post

What is Energy Quantified?

Next Post

Make 3 wishes for a new power analysis platform


Energy Quantified

Energy Quantified was founded in September 2017 with the aim of creating the most innovative data and analysis platform for the power markets. Here we will post updates on the progress of the services and descriptions of our models and features as they develop.

Recent Posts