Turning smart-meter data into actionable energy insights

Understanding and optimizing the data. Smart meters generate an enormous volume of readings (active and reactive power, voltage, current, and harmonics) sampled every second across every household. Through statistical analysis (including correlation and principal component analysis), we identified redundant variables and validated, using regression models, that reducing the sampling frequency and dropping low-value signals preserved the essential information. The result: storage savings of up to 90% in some scenarios, with minimal impact on data quality.

From data to decisions. With a leaner, well-understood dataset in hand, we built a personalized "expected consumption" model that accounts for household size, climate zone, and time of year, letting each user be compared against a realistic baseline rather than a generic average. We then applied clustering techniques to group households into meaningful consumption patterns, with a dedicated, more granular model built for the Spanish market.

We helped the client to move from basic descriptive dashboards to a genuine predictive and prescriptive analytics layer: one capable of telling a user, in plain terms, how their consumption compares to homes just like theirs, and why.

Turning raw, high-volume sensor data into intelligent systems that create real value for both the business and its customers.