Automated facilities management system
OwnerAnker Berg-Sonne, David M. Huselid +2inventors
InventorsAnker Berg-Sonne, David M. Huselid, Howard A. Nunes, Sumanth Rayancha
Life of this patent
Estimating the term…
Abstract
An automated facilities management system has the ability to predict occupant behavior by identifying recurring patterns in the way that people use buildings and comparing them with environmental characteristics. This technology is not limited to human behavior patterns, but extends to any mechanical systems or data points that tend to vary in recurring patterns. The data processing is carried out by rules engines triggered by relational database modifications.
Claims17
A method for predicting future behavior, comprising:
- collecting historical sensor data related to a behavior;
- processing the historical sensor data into a plurality of intervals related to the behavior, each interval comprising a plurality of time slices;
- generating a plurality of distinct interval pairs, each interval pair comprising two intervals included in the plurality of intervals;
- for each respective interval pair included in the plurality of interval pairs, comparing time slices between intervals in the respective interval pair to yield a similarity measurement;
- generating one or more clusters of intervals based on the plurality of interval pairs and their respective similarity measurements;
- identifying a plurality of patterns based on the one or more clusters of intervals;
- comparing current data related to the behavior with the plurality of patterns to determine a best fit pattern that is most similar to the current data;
- and
- applying data in the best fit pattern to predict future data related to the behavior.
A system for predicting future behavior, comprising:
- a database configured to store a collection of historical data related to a behavior, the historical data organized into intervals related to the behavior with each interval comprising a plurality of time slices;
- a processor configured to:
- generate a plurality of distinct interval pairs, each interval pair comprising two intervals included in the plurality of intervals,
- for each respective interval pair included in the plurality of interval pairs, compare time slices between intervals in the respective interval pair to yield a similarity measurement,
- generate one or more clusters of intervals based on the plurality of interval pairs and their respective similarity measurements;
- identify a plurality of patterns based on the one or more clusters of intervals,
- determine a best fit pattern to current data related to the behavior, the best fit pattern being the most similar to the current data;
- and
- a local energy optimizer component configured to:
- predict future data related to the behavior derived from the best fit pattern.
A computer-implemented method for predicting future behavior, comprising:
- receiving, by a computer system, sensor data related to a behavior from one or more sensors;
- dividing, by the computer system, the sensor data into a plurality of time intervals;
- creating, by the computer system, a list of pairings of the plurality time intervals;
- determining, by the computer system, a similarity value for each of pairing included in the list;
- ranking, by the computer system, the pairings in the list in increasing order according to their respective similarity values;
- performing, by the computer system, an iterative clustering process to assign each time interval to one of a plurality of clusters, the iterative clustering process comprising:
- determining, by the computer system, a pattern fitting current data related to the behavior based on the plurality of clusters;
- and
- predicting, by the computer system, future data related to the behavior based on the pattern.
Dependencies are inferred from the claim text.