In this presentation, Martti Kontula discusses EnerKey’s strategy for reducing energy consumption, how using a time series database enhances EnerKey’s competitive advantage, and their approach to using machine learning to help their customers forecast and optimize operations.
2. Agenda
• EnerKey company overview
• Solution overview
• Old system rewrite & InfluxDB
selection process
• Data collection architecture
• Super Clever EnerKey – adding ML
to the mix
• Q&A
4. Save Energy Save MoneyIncrease
productivity
Sustainabilityand Savings
Ourmission, yourvalue.
Fulfilling
stakeholder
requirements for
sustainability.
Reducing energy
consumption and
environmental
footprint.
Certified ISO 50001
and ISO 14001
support.
5. 100 000+ 15 000+ 1st SaaS
METERING POINTS PROPERTIES COMBINE SUSTAINABILITY
AND ENERGY MANAGEMENT
1 000+ 80+ UP TO30%
CUSTOMERS INTEGRATIONS AND
INTERFACES
VAAKA BUYOUT FUND
AND MANAGEMENT
6M€ 60+
REVENUE 2019 PROFESSIONALS
OWNED BY
COST SAVINGS
EnerKey facts
8. •More than 80 out-of-the-shelf
integrations to energy
companies, building automation,
measuring data systems, IoT
devices, Solar PV systems, etc.
•EnerKey supports seamless
integrations to connected
systems and services through
state of the art API
Connectivity
If wedon’t have it, wewill build it.
9. • Energy company or Property Management System
provider, modernize your existing energy
reporting services with no additional
investment and project-related risks.
• Offer a superb customer experience and boost
your competitiveness.
• Expand your business by offering new services
to generate added value for your customers.
• Give your customers access to services for
managing and splitting energy bills among
tenants.
• Virtual Energy Manager – expert services for
managing energy and improving energy
efficiency.
Poweredby EnerKey
Your logo, your brand.
11. References-Retail
“Long co-operation with EnerKey has generated annual
savings of 5 million euros.”
“Energy saving is one of the key actions to combat climate
change. Kesko is among the frontrunners in energy saving.
We are well on track in meeting the objectives.”
Matti Kalervo, Director of Corporate Responsibility, Kesko PLC
12. References – Powered by EnerKey
“Among our corporate customers, there is a growing need to
monitor, report and meter energy consumption and
environmental impacts with the help of data.”
“At Helen, we aim to provide our customers with the most
powerful tools.”
Jyrki Eurén, Head of B2B Business, Helen
13. Customer problemoverview
• Real estate owners & managing
companies have dozens of
facilities distributed
geologically
• Different energy companies in
various regions provide these
facilities with water,
district heating, electricity
and 90+ more energy quantities
• Data resides in various energy
company portals Energy
Management by Excel
14. Solutionoverview
• EnerKey integrates to over 80+
different building automation
systems and data sources
• We collect the data via real
time and scheduled integrations
and harmonize the data to
common energy consumption
format at one hour resolution
different quantities becomes
comparable
• InfluxDB is used to store the
raw data as well as harmonized
data. Metadata about facilities
and buildings is stored in
Azure SQL
15. Timeseriesproblem
• Old EnerKey product used MS SQL server
• As single database capacity was peaked, a manual
process created next-in-sequence database
• Readings1, Readings2,…,Readings6
• New EnerKey development started slightly
on wrong foot
• Continued use of MS SQL for timeseries
data
16. EnerKeyproductand InfluxDBtimeline
12.6.2020 16
1995
Energiakolmio
company
founded
2014
Rewrite of
EnerKey
begins
2016 H1
New
development
faces perf
problems with
MS SQL
storage
2016/10
I started
working
with the
company
as EA
2016/12
First talks
with
InfluxData
2017 H1
Performance
and functional
testing side by
side other
development
2017/12
Decision to
buy
InfluxDB
Enterprise
2018
Very fast
paced
development
Hybrid
deployment
with MS SQL
and InfluxDB.
2019 New year
First failed
attempt to
move all data to
InfluxDB.
Rollback.
2019 Q3
All data
from legacy
platform
migrated to
InfluxDB
17. InfluxDBdecisionfacts
• PRO
• High ingestion rate
• High output rate
• Group by time
• Irregular intervals does not matter,
we still get the sharp ANY supported
resolution data
• Natural upsert
• Also has some caveats
• CON
• Lack of natural month
• Quite easily mitigated by
aggregation from days
• At the time of selection, on-prem
was the only feasible
alternative, moving to cloud
could be easier
• Started in late 2016
• VM based Microsoft SQL Server
Active/Passive cluster expensive and
slow. Performance limits exceeded.
• Open source at first, testing
alongside other alternatives
• Comprehensive testing during 2017
and GO decision made at end of year
• Alternatives:
• PostgreSQL with table manipulation
• MongoDB with timeseries oriented schema
• Cassandra with timeseries oriented schema
• Native time handling biggest single
decisive factor
18. Why InfluxDB Enterprise?
• Business requirement for reliable storage
• Some additional services includes billing based on gathered
data
• Data is not simple metrics business value
• Support
• Lots of testing and analysing before buy decision
• Realized the need for first class support for new technology
• Influx enterprise support has been VERY valuable
• Performance case from early 2019
• Single rogue query caused 40-50% of CPU load
• Enterprise support spotted this from our logs
• Unbounded low limit when searching backwards for latest
datapoint
• Changed to exponentially widening sequential search:
1day, 2 days, 1 week, 1 month, 3 months, 12 months
New data
now
1
2
1 week
1 month
3 months
Most likely hit
19. Dataacquisitionarchitecture
Hangfire Scheduler Data Sources
Public internet
VPN tunnels
Azure Service Bus
Schedule
Pull any format
Convert to common
data format
Post to service bus
Auto QA functions
Detect faults, auto-fix
Raw data
storage
Reading functions
Raw data API
Calculation functions
Measurement API
Reporting data
storage
Azure
VNET
- Normalize with temperatures
- Aggregate to natural months
DLQ
• Scheduled tasks for pull functions
• Listening functions for pushed data
• Once “on the bus”, data is safe
• Raw vs. reporting data allows manipulation without
losing original values
• Automatic quality checks fill in the blanks
• Natural calendar aggregations performed outside
InfluxDB
• Normalization calculations with location &
temperature allows comparison regardless of
location
Push any format
20. TICK stack
• Telegraf is unfortunately ruled out mostly
because we don’t have access to data
sources at this level.
• At EnerKey, we mostly pull the data from
customer’s systems rather than push it to
InfluxDB with Telegraf.
• We do support also push type of
integrations but in these cases the use of
Telegraf has not been plausible.
• Chronograf replaced with Grafana for
extensive use for monitoring the
platform as well as querying raw
business data mixed with metadata from
Azure SQL
C
T
• InfluxDB widely used for business data
storage, calculations and aggretations
• InfluxDB also used for platform metrics
I
• Kapacitor initially planned for automatic
data quality assurance, but later
replaced by service bus-based solution.
Not used.
K
21.
22.
23.
24. Adding intelligence
• Baseline
• EnerKey is a very good platform for Sustainability and
Energy Management and Reporting
• Data is secure, fast, and robust
• Integrations in an out in place
• Challenge
• Competitors exist, but they’re mostly facility and real
estate management systems with some Sustainability and
Energy Management features
• We needed to offer something that the others could not
• Solution
• Add Machine Learning models to pin point energy consumption
profiles that are misbehaving
25. MachineLearningbasics
• Harmonized consumption data for both main
metering points and sub-metering points
- Electricity consumption
- Heating energy consumption
• Good quality environment (weather) data
- Temperature
- Wind
- Sun radiation
• Good quality metadata
- Gross area
- Gross volume
- Geolocation
- Building year
- Building type
- Opening hours
26.
27.
28.
29.
30. Cooling energy in Grocery Stores
12.6.2020 30
• A regression line fitted to
analyze the increase of
energy consumption during
warm days
𝑅𝑎𝑡𝑖𝑜 =
𝐴𝑣𝑒𝑟𝑎𝑔𝑒 ℎ𝑜𝑢𝑟𝑙𝑦 𝑐𝑜𝑛𝑠𝑢𝑚𝑝𝑡𝑖𝑜𝑛 𝑤ℎ𝑒𝑛 𝑜𝑢𝑡𝑠𝑖𝑑𝑒 𝑡𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒 > 12
𝐴𝑣𝑒𝑟𝑎𝑔𝑒 ℎ𝑜𝑢𝑟𝑙𝑦 𝑐𝑜𝑛𝑠𝑢𝑚𝑝𝑡𝑖𝑜𝑛 𝑤ℎ𝑒𝑛 𝑜𝑢𝑡𝑠𝑖𝑑𝑒 𝑡𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒 < 12
• A bar represents one facility
• The color of the bar represents average electricity use per square
meter
• The slope represents the
average increase of energy
consumption when outside
temperature rises one degree
• A bar represents one facility
• The correlation coefficient
tells how good the estimate
is.
• A point represents one
facility and it is colored red if
the slope > 0.7
31. Coolingenergyresults
• One medium sized grocery store stood out in the results compared to stores
of similar size and location
• The slope increase was substantiallysteeper when temperatures exceeded
20+°C
• Note! Warmweather inFinland
• After investigation byprofessionals on site, undersized condensing
equipment was discovered and replaced by adequately sized ones
• Excesscooling energy consumption did not occur any more
32. Thanks!
Follow our story atLinkedInand our website:www.enerkey.com!
Martti Kontula, CTO
+358440160579
martti.kontula@enerkey.com
33. We look forward to bringing together our community of
developers in this new format to learn, interact, and share
tips and use cases.
8-9 June, 2020
Hands-On Flux Training
www.influxdays.com
23-24 June, 2020
Virtual Experience