Exposing Building Behaviour Using BAS Discovering groups and structures in the data that are similar without knowing the  structures of the data.

Nirosha Munasinghe MBusIT BSc BE (Hons) (Melb) Product Development Manager, Open General  Contributing Editor

“Each DDC in the system shall contain a trend that will store samples of all the of the data points” “The logged data must be sampled every 15 minutes and archived over 2 years”. “Provide trend logging, event logging and hours run on selected points with selectable time intervals”.  These are typical phrases from specifications. What do these mean?  Lots of data!  The reduction in the price of computer memory has allowed BAS systems to store unlimited amounts of data from the system. What do we do with data?  At present a BAS system presents the data in the reporting tool in graphical format for the user to interpret comparisons and report what has happened in the past and what is happening at present. This is not enough. We need to use the data to predict the future. This article examines typical methods to use data to predict the future.

A large collection of data is useful to the daily operations of a building. A temperature graph over a few days can illustrate room temperature behaviours to a plant technician. The data assists the facility department operating the building on a daily basis. Therefore it is only operational data. To further enhance the decision making process for the facility managers, the data needs to be fed into models to discover further intelligence.  Intelligence is the ability to learn or understand from experience and use reasoning to solve problems. Therefore past data can be used to learn and solve future problems. It involves acquiring data and information from a variety of sources and modelling them for decision making. The modelled data from the BAS system can then be a decision support system for the facility department.

Figure 1:  Decision Support System Architecture

The decision support system can be used for tactical and strategic decisions for the facility department. Figure 2 illustrates the data hierarchy and types decisions that can be made with the respective data types.

                             

Figure 2:  Data Hierarchy

[an error occurred while processing this directive]How do we implement such a system for BAS? The raw BAS data needs to be fed into a model. There are various proven data models available from a simple decision tree to more complex neural network networks modelling. A more recent development in data modelling is the concept of data mining. Data mining is a branch of computer science and artificial intelligence, where it extracts patterns from the dataset to transform data to intelligence to support the decision making process of a business. The model analyses large pools of data to find patterns and rules that can be used to guide decision making and predict future behaviour.  To successfully obtain interesting patterns and behaviours requires a data warehouse. A data warehouse is a physical repository where relational data is specially organized. In a modern BAS system, there are numerous data sources with web services and open web standards and it is simple to merge large collections of data into a warehouse.

The data from the warehouse is the input to the data mining engine. The data is modelled to find patterns and correlations between fields. Some common techniques used in data mining follow:

• Classification: is a technique used to predict group membership for data instances. For example, it might attempt to predict the energy cost of running the building for the day.
• Association: searches for relationships between variables. For example the relationship between temperature and energy cost.
  
• Clustering: is a technique of discovering groups and structures in the data that are similar without knowing the  structures of the data. In a large facility such as a university or a hotel chain it can group energy usage patterns between buildings.
• Apriori: is an algorithm for learning associations between fields to create rules with a given confidence. For example, it can predict that the building will use 6 kWh of energy today with a probability of 85%. 
  

To illustrate how data mining can assist in BAS system, let’s examine the following dataset that consists of temperature, the operation of a chiller, and how much energy is used in the building.  

Feeding the above dataset into a data mining model can deduce following results:

The example illustrates simple association rules that can be used by a facility department to assist the decision making process. The example association rules just touches the surface of data mining. It will be a powerful tool for the BAS market in the evolution of smart grid as we attempt to make decisions on the fly to interconnect and make networks intelligent.  Also, the predictive nature can assist in reducing energy in a building as the models can predict how much energy a building will use for any given weather condition or day.

It can be seen from the examples that data mining can change raw data to information and knowledge. The raw data will run the business on a daily basis but for the business to obtain a competitive advantage it requires information and knowledge. Therefore BAS systems need to incorporate concepts such as data mining to assist better operation of buildings.
          
        
 
   

[an error occurred while processing this directive]
[Click Banner To Learn More]

[Home Page]  [The Automator]  [About]  [Subscribe ]  [Contact Us]