The Blog of operational excellence and energy management for industry

Browse our articles, expert advices and clients testimonials: our experience save your time!

Big Data: Quantity is nice but quality is better.

In this era of Industry 4.0, industrial operators know that they can count on Big Data or Artificial Intelligence technologies to enhance the energy efficiency of their factories. Yet they are not always aware that the quality of available data prevails over their quantity in order to achieve optimum energy management outcomes. Explanations below.


What data to use? Are they available?

Whether operational data and parameters for machines, utilities or production programs, weather data or other applications, a myriad of data are used or generated by industrial operations. But are these data always relevant to the energy efficiency improvement process? Or more practically, is the power consumption for each parameter to be optimized properly measured? Are the factors impacting power consumption known accurately enough? “From far away, everything looks fine”, says Business Consultant Yves Bergeron. “The client has dozens of meters and several thousand dozens of data. But it is only when defining precisely the information that the future Energy Information System must deliver that you sometimes realize that the data available are not the right data. Actually, the Big Data approach to energy management is something new and may require specific data. And your project may end up in a deadlock until the right data are made available. Whenever that proves impossible, then the business case must be redefined, or may even have to be dropped entirely.”


Upgrading your data to enable proper analysis and energy optimization

The data directly relevant to the energy efficiency project should be identified. Thus, for each major energy issue, the information to be delivered by the Energy Information System for operators to improve the energy efficiency must be clearly defined; and the data enabling such information to be generated must be identified: for instance, for demand management during a production shutdown (baseload power demand), you need to know the “off” and “on” status of the main power-consuming systems as well as the power consumption in the various areas of the shop-floor. If such data are incomplete, unreliable or non-existent, then the data must be corrected, or if impossible, the energy improvement project needs to be redefined.

But that is not all, and it leads to the next question: Even if the data are reliable, how often are they cascaded back up? For instance, many meters work in “pulse” mode and, in this case, the sampling frequency must be checked: there should be at least twice as many measurements as the duration of the event. Therefore, based on hourly variations, measurements should be taken every 30 minutes.


A step-by-step method to improve data quality

The success of an energy optimization project is therefore largely dependent on the quality of the data. Data expert Mickael Ngo recommends a specific methodology to ensure such quality:

  1. Inventory all existing sources of data in the factory.
  2. Retrieve only the data truly useful for the use case under study.
  3. Qualify such data based on:
    • Precise location,
    • Explicit name and description,
    • Appropriate time step (or sampling frequency), in line with the reality of the factory,
    • Predefined measurement unit (e.g. kWh/m3 of blown air).
  4. Check the proper calibration of sensors.


Necessity for data governance

“We are frequently faced with a difficulty to qualify the data correctly”, explains Mickael Ngo. “Because industrial operators focus on their core business, they outsource their data acquisition to service contractors. Hence, they lose knowledge over their data that they may not have taken the time to codify or document either. Sometimes, the data they thought they had, may not be the right data.” One solution would be to set up a data governance system to ensure data integrity, to be managed by a new player whose role (that remains to be invented) would be to guarantee the quality and proper use of the treasure trove of data that are plentiful in the factories. “Today we talk about manufacturing secrets in the industry. Tomorrow, it will be about data secrecy”, concludes Yves Bergeron.