Grid Intelligence and Optimisation
Electrical grid operators have been struggling in recent years to deploy new lines to meet demands for more energy and higher quality of service. Regardless of their success in that endeavour, every grid operator is now asked to push the use of existing lines until the limits of safety. Grid Intelligence Optimisation (GIO) is Albatroz Engineering's answer to the utilities' need to optimise the remaining lifetime of ageing grids while maintaining or improving quality of service. GIO is settled on geo-referenced databases from inspection data (from Albatroz Engineering's PLMI), third party data (such as meteorology, pedology and edaphology, environment, and forest exploration), land use, SCADA variables, grid planning and project and asset management.
1. Grid Maintenance Cycle
Defining grid maintenance as a cycle of processes that keeps the electrical grid running includes inspection, quality and condition audits and remedial actions. The proposed framework to optimise this cycle features an architecture and a tool set to aggregate data and methods from different sources into a consistent model for grid maintenance. The Figure in the centre shows one model of the cycle with twelve main tasks, organised as a clock dial, where green background represents field tasks and sand background represents office tasks.
This exploitation cycle involves inspection (from 1 to 3 o'clock on the figure), followed by in depth multi-variable analysis (from 4 to 7 o'clock) that produces maintenance guidelines to the field (from 8 to 9 o'clock). Analysis of the circuit reliability and risk minimisation at grid level (from 10 to 12 o'clock) combines previous steps to plan maintenance and refurbishment for the following years.
The most innovative tools being developed focus on tasks depicted at 4 to 7 and 12 o'clock, briefly outlined below with illustrative cases:
- Pattern matching (condition, 4 o'clock)
With a time and space database containing every issue/incident found, it is possible to search for patterns. Matches in different fields (e.g.: manufacturer, time of day, pollution, line load) can then be detected by data crunching, highlighting common factors hitherto unnoticed. The cover page illustrates this process for lightning, where one correlates lightnings in time and space detected by weather stations with grid events attributed to lightning.
- Timeline (condition, 4 o'clock; risk, 5 o'clock; remaining value/life, 10 o'clock).
For instance, vegetation growth and mechanical defects detected by thermography are monitored so that one can project evolution timelines and determine the optimal time and scope of remedial actions.
- Difference detection (inspection, 1 o'clock; condition, 4 o'clock; risk, 5 o'clock)
When a given issue has been detected and the next inspection fails to detect the issue, this example difference has to be analysed. Inspectors may have failed to use the best methods, thus inspection should be audited and enhanced. If maintenance has occurred, inspection acts as an ad-hoc audit: missing issues validate maintenance, while resilient issues denote underlying causes that were not tackled by remedy actions.
- Condition-based risk estimation (advance from 4 o'clock to 5 o'clock)
The goal is to estimate the risk associated to each asset condition issue. Extending the knowledge to a whole grid involves wrapping each asset subject to each sort of fault with a model of the environment as rich and detailed as possible.
- Probability of failure based on risk estimation (advance from 5 o'clock to 6 o'clock)
Integrating local probability of failure into the whole grid analysis is easier to express formally using probability laws but eludes analytic expressions for any but the most basic grid, calling upon numerical simulations and algorithms. Results are trustworthy only if they cope with the many factors affecting the line and if they predict the behaviours observed in the grid.
- Maintenance resource allocation (7 o'clock)
Vegetation management is a major resource-consuming activity. Computational methods are used to determine the optimal maintenance schedule as a sequence of sections requiring maintenance, depending on clearance measurements, tree species growth estimation, distances from roads, among other factors.
- Inspection scheduling (12 o'clock)
The purpose here is to replace full line inspection at fixed intervals with sections of different lines inspected at optimal intervals and connected to minimise travelling between inspections.
2. Data and Architecture
Environmental causes have a major influence on the performance of overhead lines. Vegetation, lightning and birds are on the top of concerns, thus the rise of environmental concerns transformed the way power systems are designed and exploited.
In order to reinforce continuity and improve power quality, Albatroz Engineering is working to intensify preventive maintenance work on the electrical system, since acting pre-emptively avoid inconveniences due to lack of energy on the system. Environmental data, also including forest fires, fog and pollution, together with a record of historical incidents and their causes have been integrated in a PostgreSQL database related with the grid topology.
The architecture (figure on the top) with links to geographic information systems (GIS) and asset management systems (SAP), allows developing tools for mining, time-space correlations, searches of nexus of causality, optimising inspection and maintenance plans, and so on, thanks to the integration of geo- and time-referenced relevant data of different sources, and the expertise of grid operators and of Albatroz Engineering team.
3. LIONS Project and Tools
Such architecture has been implemented in R&D joint project LIONS (Lines Inspection OptimisatioN System) with REN, the Portuguese Transmission System Operator. In this context, different tools have been developed, ranging from automatic detection of stork nests using inspection videos, to a comprehensive probability-based risk model for the power-lines, aggregating a number of possible causes for fault occurrence. Each such cause has a specialised treatment due to specific inherent characteristics, as well as to the kind, precision, granularity, and amount of its data available. Nevertheless, all causes are handled in a unified well-founded risk model, which combines probability of failure with the severity of failure for each line in different time periods. The probability of failure for each line due to a given cause may be calculated by aggregation of probabilities of failure on a finer granularity such as the span, when available data makes it possible.
With risk indices available at span level, it is possible, for instance, to optimise maintenance plans, based on how low one wants to keep the global probability of failure. The location and extension of vegetation to clear are then calculated as shown in the figure on the left, based on a graph analysis of the grid, taking possible path redundancies into account when transmitting the necessary power from source to sinks satisfying all line capacity constraints. Introducing time analysis, it should be possible to schedule future right-of-way inspections and predict future maintenance efforts.