4 steps road to a Data-centric Utility
Despite the relevance of the utilities' digitalization, it could be undermined in the event of lack of data governance and well-structured processes. The utility's successful digitalization depends of developing a comprehensive data and analytics journey, by addressing: 1) data architecture; 2) governance; 3) data centralization; and 4) people competences.
1. Data architecture
The development of a data-centric utility benefits from having an utter vision of the different process layers, from the grid to the decision-making models. In Fig. 1, an architecture is proposed for building a data-centric utility, encompassing the next layers:
Grid: physical infrastructure subject to energy flow, O&M and engineering, procurement and construction (EPC);
Condition monitoring and sensing: online and offline data collection, through sensors and local protection, automation and control (PAC) systems, as well as via in-field asset inspection;
Data integration and storage: the enterprise asset management (EAM) system provides static data of the assets (e.g. age, installation, manufacturer, technology type) and maintenance information (e.g. failures, maintenance orders, O&M plans, inspections results), whereas SCADA stores historical records of incidents/outages, events and operations (e.g. asset maneuvers);
Analytics: this layer is accountable for developing deterministic and probabilistic risk models, focused in the short (less than one year), medium (between 1 and 3 years) and long-term (more than 3 years) analyses;
Simulation and decision-aid: investment and maintenance plans based on financial constraints, regulatory model, workforce and risk indices.
Fig. 1: Architecture for a data-centric utility
Besides the traditional structured data for asset management (failures, incidents, maintenance orders, etc.), the utility should also prepare the data infrastructure for handling unstructured data, such as asset real-time attributes, weather variables, corrosion indices, vegetation growth, lightning, etc.. The structured data should be collected through extraction, transformation and loading (ETL) tools, and stored in a data warehouse. Regarding the unstructured data, a data ingestion tool should be used together with a data lake, since flexibility and scalability are key for these types of variables.
Taking into account the wide variety of data sources and internal users, the robustness of the decision aid models requires the adoption and implementation of a holistic data governance. A centralized Data and Analytics Office (DAO) should be set up and led by the Chief Data Officer (CDO). The CDO is responsible for drawing the mission and rules for the utility’s data governance, as well as ensuring periodic meetings for alignment.
This governance model should also include the Data Management Council (DMC), to address the next topics:
Global coordination and working group members;
Owners, contributors, users and access rights of each software application;
Metadata policy and data processes;
Data models and dictionaries (e.g. unique data ID);
Communication and decision rules whenever an application upgrade is needed;
Procedures and frequency of data collection and sharing;
Reporting and key performance indicators.
3. Data Centralization
Typically, in the laptop of each engineer one finds a data warehouse. This stereotype is a reality in several utilities, undermining the accuracy, transparency and accessibility of business data. Data centralization is (at least) key for asset management, grid planning and system operation.
It starts from a context of countless data warehouses with different models and IDs, and aims at building a centralized data process. This data centralization journey requires the empowerment of the Data and Analytics Office (acting as a “single point of truth”) and the engagement of the information systems division.
The first tier of this endeavor consists of mapping all the dynamic and static data across the utility’s software applications, as exemplified in Fig. 2 and Fig. 3.
Fig. 2: Asset data model
Fig. 3: Data mapping
The data centralization also involves identifying the asset ID in each application (illustrated in Fig. 4), which will enables building up data dictionaries and analytical models.
Fig. 4: Asset ID in the different software applications
Finally, data dictionaries must be written in order to standardize the asset information across the company’s applications and ensure data quality processes.
4. People and Competences
Besides the process and technology transformation towards a data-centric organization, the workforce should be reskilled. In a digital-driven utility, traditional engineering roles – such as design, grid planning and system operation – will turn out commodities, due to the advent of artificial intelligence, robotics and automated processes.
Consequently, utilities ought to carry out a strategic workforce plan, by matching the business trend (i.e. asset management transformation) with people’s skills.
The utility of the future will be surely sustained by engineers with strong digital competences, including programming. Designing overhead lines or planning grid expansion will require the use of machine learning and optimization methods.
If the general utility’s workforce will have to become rather digital-oriented, a dedicated data and analytics office should be set up, gathering way different profiles. This team should be led by a manager (CDO) with utility-based experience, though holding a strong analytical background and computer science proficiency.
The data and analytics office, on its hand, will benefit of blending domain experts (e.g. former grid planners and maintenance engineers) with data scientists. While the former could be found internally, the latter will most probably be hired externally.
Data scientists are both critical for the asset management transformation and hard to find in the market. Indeed, this role compromises an odd combination of skills, i.e.: business know-how; statistics and mathematics; and data engineering and programming.