Data Analytics for Future Energy Systems Dr Stuart Galloway & Dr Bruce Stephen Advanced Electrical Systems Group Institute for Energy and Environment Department of Electronic and Electrical Engineering University of Strathclyde Glasgow G1 1XW United Kingdom {stuart.galloway|bruce.stephen}@strath.ac.uk
Introduction • Smart Grid is essentially about data • More informed power system: sensing at an extent and rate previously unknown • Q: How to get business value from this investment in monitoring? – New services? – Augment existing services? • A: Understand current operation & constraints…
Characteristics of Demand WHAT DRIVES THE NETWORK?
Demand • Smart Meters/BEMS – 30 minute reads • Understand demand to manage it – DSM • When used/when not • Realign peak demand with renewable generation • All data driven!
•Loads on the LV network were always assumed to have a high degree of variability to them due to the nature of domestic routine •AMI & BEMS deployment allows much greater insight into the behaviour of these loads •Patterns are difficult to capture given the complex nature of the data Patterns repeat in load •High dimensional – how do they all profiles both across days and relate? customers… •Non-stationary (many sub-behaviours observed e.g. weekends, holidays) •Tools have been developed for our past projects that: •Formulate classes of energy use and automatically categorise residents •Quantify behavioural consistency in terms of movers and stayers Load •Forecast aggregated residential load characteristics can be •Model appliance usage and its attributed to variability appliance •Wet appliances usage patterns, which can also •Heating (space and water) be quantified… •EV Charging?
Power System Operation and Control HOW DOES THE NETWORK BEHAVE?
Network Operation • Load flows – Power flow directions in the network – Voltage excursions along feeders • State estimation – Infer unmeasured quantities from measured ones • Data model drives physical models – Real loads are not homogenous • Homogenous assumption may result in different outcomes…
28 44 37 46 5 17 37 33 44 47 28 46 26 1 9 48 5 1 7 22 9 19 26 48 47 9 33 13 24 6 12 18 41 21 21 1 2 34 1 3 22 7 15 81 81 2 11 27 4 1 1 8 7 1 5 45 11 kV sub- 41 43 0.4 kV 29 35 42 24 station 43 10 2 42 11kV 30 50 27 50 1 1 25 6 40 49 25 49 31 4 32 1 6 30 29 11kV 32 31 3 1 8 35 34 40 39 23 Topology is OK but the 45 1 4 16 39 14 23 8 Single laid 3 real interest comes Double laid …but what is realistic? 36 Triple laid 36 38 from what and where 1 0 38 Realism is important 20 (how far) loads are 20 from the DNO attached from the perspective as the feeder end. Solution is Feeder/Microgrid Model following examples to superimpose the illustrate… Use to evaluate power system health metrics (load flows, voltages network model on an etc.) and inform. Load models plug into point loads and resulting existing geographic aggregate can be evaluated (minus losses) at a desired point e.g. grid infeed. housing layout…
…one infeed Rural/Suburban transformer – possibly specified for fewer houses than were Heterogeneous eventually built. properties of a similar vintage (1973), generous but uneven spacing, increasing PV penetration… Stephen, B., Mutanen, A., Galloway, S., Burt, G. & Jarventausta, P. (2013) Enhanced load profiling for residential network customers. IEEE Transactions on Power Delivery. ISSN 0885-8977 (In Press) Stephen, B., Isleifsson, F., Galloway, S., Burt, G. & Bindner, H. (2013) Online AMR domestic load profile characteristic change monitor to support ancillary demand services. IEEE Transactions on Smart Grid. ISSN 1949-3053 (In Press)
Condition Monitoring HOW DOES THE NETWORK AGE?
Network Monitoring • Condition monitoring – Could be SCADA, could be more… • Whats ‘right’, whats ‘wrong’? – Huge volumes of data – What does fault(s) look like? • How to predict fault onset?
Fault Detection/Diagnosis • Offshore wind Generation – Look at the way the power curve behaves – Plant wear and degradation • How do changes manifest? • Transmission Power Transformers – Partial Discharge detection – Different frequency compositions • What types of faults do they represent? • Distribution Switchgear – Trip coil testing – Mechanism and control system faults • What does the shape of the test record indicate? Stephen, B., Galloway, S.J., McMillan, D., Hill, D.C. & Infield, D.G. (2011) A copula model of wind turbine performance. IEEE Transactions on Power Systems, 26 (2). pp. 965-966. ISSN 0885-8950 Baker, P., Stephen, B. & Judd, M. (2013) Compositional modelling of partial discharge pulse spectral characteristics. IEEE Transactions on Instrumentation and Measurement, 62 (7). 1909 - 1916. ISSN 0018-9456 Stephen, B., Strachan, S.M., McArthur, S.D.J., McDonald, J.R. & Hamilton, K. (2007) Design of trip current monitoring system for circuit breaker condition assessment. IET Generation Transmission and Distribution, 1 (1). pp. 89-95. ISSN 1751-8687
Asset Management and Replacement HOW CAN THE BUSINESS MAKE INFORMED DECISIONS?
Asset Management • Track fault occurrences/onsets on individual plant • Look across whole fleet – Inter and intra plant faults • Manage/model lifecycle of fleet – Spares inventory management • Identify subsystems failures – Repair prioritisation
Intra-Plant Condition Who is most similar � who will fail next?
• Markov chain type models • How do assets move through their lifecycles over time? • Fault tree type models • How do systems of components fail? Wilson, Graeme and McMillan, David (2013) Modeling the effects of seasonal weather and site conditions on wind turbine failure modes.In: ESREL 2013, 2013-09-30, Amsterdam.
How to go forward? • Power Systems can benefit from Big Data • Hence Smart Grid - Operational robustness & business value through informed operation • Good physical models and domain knowledge exist • ‘Black box’ type models may fail to capitalise on this • Other Smart Grid stakeholders could see benefit from these outcomes. How can we all work together? • IT Service Providers • Community Energy Groups • Distribution Network Operators • Work with Power Systems engineers to develop models that leverage big data insight with well understood domain knowledge
Recommend
More recommend
