Frequently Asked Questions
We’d like to use Bellweather’s technology. What is the resource intensity to support this on our side?
Very low. At the outset, we will cooperate with you to design a process of collecting your data (using APIs or a file-drop process) and creating outputs that will meet your specific needs. Once the process is designed, it takes no work on your end to receive outputs from our system. Everyone is already accounting for weather in some way, so by relying on Bellweather for weather analysis, your overall workload will go down.
We are currently using a few weather stations and averaging them. How is Bellweather’s approach to weather data different?
We begin by finding the relevant geography—a service territory map or an electric market territory. We then choose a distribution of weather stations; in the past we've used approximately twice the number of station clients used previously because our automated process for analyzing data removes the constraints of using many stations. Once we have selected stations, we use population maps and other factors to assign dynamic weights for the stations.
We don’t collect hourly energy data. Can Bellweather’s technology still benefit us?
Absolutely. Data restrictions are a part of the job and there are a variety ways we can work around those restrictions. Some clients have hourly data samples that we can extrapolate to learn about the whole data set. Bellweather also has generic hourly profiles that can be adapted for use in situations that no hourly data is available.
What type of techniques is the model using?
The model uses a non-parametric modeling technique in conjunction with a control algorithm. The algorithm creates a dynamic, adaptive model that optimizes for both common and extreme weather conditions.
Is there anyone else using Bellweather’s modeling approach for weather?
No, but there is a wide variety of academic literature advocating for non-linear hourly based modeling. Tao Hong from UNC is the best-known advocate for this practice.