Holistic Analytics to Meet Business Objectives
In cases when refineries, plants, or business units are utilizing an EIDI in their operating plants, it may make sense to analyze problems holistically, to determine if improvement is possible across the entire refinery, a fleet of refineries, or for the business unit. To do that, companies often combine contextualized production and operations data with business data to identify roadblocks that keep the facility from operating more effectively. Big data analytics enable management to effectively plan where and when to allocate funding for debottlenecking purposes, expand physical capacity, or when to optimally schedule maintenance outages.
In addition, artificial intelligence and predictive analytics provide estimates on how long assets will operate effectively or when they need more production capacity. For example, during the oil and gas industry downturn of the mid 2010s, the larger oil and gas operating companies performed large-scale analyses to identify controllable losses, production obstacles, and correctable equipment inefficiencies to optimize their upstream, midstream, and downstream operations. Data analysis of this scale usually requires a large quantity of high-fidelity time-series data combined with other business and equipment information to feed process and equipment models, big data analytics, and predictive analytics.
When companies undertake such a large data analysis project, specific IT and data science skills are required. As such, when using plant and engineering information, the ОТ world converges with the IT world. To implement a project of this magnitude successfully, they must align and work together, bringing their individual disciplines and skill sets to the table. An example of an analytic integrating multiple types of data is shown in Figure 10.2.
Data types integrated for big data analytics.
Plant and business operations personnel bring extensive experience and knowledge of how plants and processes operate. They are generally aware what types of problems exist, but may not know exactly where to find them. They are very familiar with time-series data and know how to utilize it. More than likely, they can determine which data values are accurate and which values are outliers that should be discarded and not be used by the analytics. They may even know how to incorporate cost data into operations and production data, but likely do not know where these costs reside or know how to access this data in the IT world.
IT workers and data scientists possess skills that complement the business people and engineers. While they may not have the technical knowledge of how product is made and which issues limit plant effectiveness, they can determine which types of data best facilitate big data analytics and how best to utilize them. Some examples are financial databases, equipment databases, and geospatial databases.
Relational databases are more transactional than real-time databases and designed to easily access relationships between data. If used, geospatial databases are optimized for spatial interactions among data. Similarly, operations data needs to be organized and formatted for IT analytics. Fortunately, this operations data can be extracted and converted to a format that the business analytic system or geographic system can easily ingest to link process data to non-time-based systems or databases to solve business issues, such as the following:
- • Should we add capacity at a particular location?
- • Where are the hidden production losses?
- • When will a specific asset be expected to fail?
- • What are the costs of maintaining existing equipment versus purchasing new assets?
- • Where are the best performing assets located?
- • What would be the energy costs if the heating, ventilation, and air- conditioning (HVAC) infrastructure was modernized? Would it be cost-effective in the long run to do so?