Breaking Down Analytical and Computational Barriers Across the Energy Industry Using Databricks
1. Breaking Down Analytical and
Computational Barriers in
Energy Data Analytics
Jonathan Farland
DNV GL Energy
2. • IntroductionsWho is DNV GL?
• Overview
• Data Science
Energy Analytics
• Demonstration
Statistical
Computing Pilot
• Plans
Concepts in
Development
• DiscussionQ&A
Agenda
17. Forecasting Approaches
– Similar Day Matching
– Statistically Adjusted Engineering
(SAE)
– Univariate Time Series (ARIMA)
– Multiple Linear Regression
– Econometric
– Machine / Statistical Learning
– Semiparametric Regression
– Artificial Neural Networks
– Fuzzy Logic
– Support Vector Machines
– Gradient Boosting
18. Additive Semiparametric Model
𝑦" = ℎ 𝑡𝑖𝑚𝑒 + 𝑓 𝑤𝑒𝑎𝑡ℎ𝑒𝑟 + 𝛼 𝑏𝑒ℎ𝑎𝑣𝑖𝑜𝑟 + 𝜀"
Short Term
Electricity
Demand
Time of Year
Prevailing
Atmosphere
Conditions
Recent Demand
Behavior
19. Additive Semiparametric Model
𝑦" = ℎ 𝑡𝑖𝑚𝑒 + 𝑓 𝑤𝑒𝑎𝑡ℎ𝑒𝑟 + 𝛼 𝑏𝑒ℎ𝑎𝑣𝑖𝑜𝑟 + 𝜀"
Short Term
Electricity
Demand
Time of Year
Prevailing
Atmosphere
Conditions
Recent Demand
Behavior
24. Benefits of Big Data from Advanced Metering Infrastructure
ü A deeper understanding of demand and thereforehuman
behavior (think energy efficiency)
ü Cost effective operating costs
ü Real-timenotification of power outages
ü ImprovedSystem Planning and Reliability
ü Allows for integration of disruptive technologies like
Electric Vehicles
Statistical Computing Pilot
32. Current Concepts in Development
Weather Normalization at Scale (e.g., California)
Real-time Energy Forecasting Using Statistical Learning and Spark Streaming API
Real-time Customer Sentiment Analysis
Grid Reliability Analysis
Cybercrime Protection of Electricity Grids