Forecasting and Energy information

CITIES 2nd General Consortium Meeting 25-26 May, 2015, DTU, Lyngby
ENFOR
Forecasting and Energy Information
Services
Henrik Aalborg Nielsen
ENFOR A/S
han@enfor.dk
Henrik Aalborg Nielsen, ENFOR A/S
1
CITIES 2nd General Consortium Meeting 25-26 May, 2015, DTU, Lyngby
ENFOR
Outline
• Cloud service solution.
• Consumption and production forecasts.
• Monitoring signals for systematic changes.
• Smart meter readings.
Henrik Aalborg Nielsen, ENFOR A/S
2
ENFOR
Energy forecast solutions
Cloud service solution
• FTP / SFTP file up- and down-loads
with retries and book-keeping.
• Client-specific (SOAP, XML, …)
• Highly configurable HTTPS based graphical
user interface, supporting access level
restrictions.
• Security of supply (NWP providers, data
centres, …)
• Data security (firewalls, encryption, …)
Henrik Aalborg Nielsen, ENFOR A/S
3
CITIES 2nd General Consortium Meeting 25-26 May, 2015, DTU, Lyngby
ENFOR
Consumption and production forecasts
• Heat load forecasts and temperature
optimization for district heating systems –
usually for a horizons up to one week.
• Power load forecasts – for horizons up to
one week, could be longer.
• Wind power production forecasts – usually
day ahead, but up to one week is feasible.
• Solar PV power production forecasts –
usually day ahead, but up to one week is
feasible.
Henrik Aalborg Nielsen, ENFOR A/S
4
CITIES 2nd General Consortium Meeting 25-26 May, 2015, DTU, Lyngby
ENFOR
General forecast principle
(Multiple)
Numerical
Weather
Forecast
Forecast
Engine
Forecast
Value(s)
Actual
Value
System
• Calibrated against actual production /
consumption available on-line or off-line in
batches.
• If on-line data is available the autocorrelation is
used in order to improve short-term
performance.
Henrik Aalborg Nielsen, ENFOR A/S
5
CITIES 2nd General Consortium Meeting 25-26 May, 2015, DTU, Lyngby
ENFOR
Example:
SolarFor – combined phys./stat. model
• Core of model based on physical principles
(direct/diffuse irradiation, panel efficiency, …).
• Model characteristics estimated from NWP data
and actual PV power production.
• An number of secondary NWP variables (e.g.
atmospheric water content) used in order to
detect systematic NWP errors via data mining /
machine learning methods.
• Continuous re-calibration of models.
Henrik Aalborg Nielsen, ENFOR A/S
8
CITIES 2nd General Consortium Meeting 25-26 May, 2015, DTU, Lyngby
ENFOR
Comparison of models using ECMWF 12z
Henrik Aalborg Nielsen, ENFOR A/S
10
CITIES 2nd General Consortium Meeting 25-26 May, 2015, DTU, Lyngby
ENFOR
SolarFor using GFS-HD 18z / ECMWF 12z
Henrik Aalborg Nielsen, ENFOR A/S
11
CITIES 2nd General Consortium Meeting 25-26 May, 2015, DTU, Lyngby
ENFOR
Automatic monitoring signals for changes
• Issue alarms if signal properties (e.g. mean and
variance) change systematically.
• Setup by specifying how fast a given change
should be detected, together with an acceptable
false alarm rate.
• Monitor for fixed level or tracks the level.
Henrik Aalborg Nielsen, ENFOR A/S
12
CITIES 2nd General Consortium Meeting 25-26 May, 2015, DTU, Lyngby
ENFOR
Example of monitored signals
• Wind farm / turbine performance indices (with
DONG Energy).
• Household electricity standby power
consumption (previously with Elsparefonden).
• Monitoring of data quality, e.g. the fraction of
missing, out of range, and “frozen” values.
• Monitoring of excess energy consumption not
explainable via climate measurements.
• …
Henrik Aalborg Nielsen, ENFOR A/S
13
CITIES 2nd General Consortium Meeting 25-26 May, 2015, DTU, Lyngby
ENFOR
Use of smart meter readings
(frequent energy measurements from houses /
households, e.g. hourly values)
• Cost-efficient district heating temperature
optimization using closed-loop control (on-line
measurements).
• Screening for houses / households with poor
energy performance (off-line measurements).
• …
Henrik Aalborg Nielsen, ENFOR A/S
14
CITIES 2nd General Consortium Meeting 25-26 May, 2015, DTU, Lyngby
ENFOR
PRESS-TO for DH systems
• Temperature Optimisation.
• Keep the network temperature as low as
possible, while meeting consumer energy and
temperature demands.
• Closed-loop control using on-line measurements
of network temperature.
• New development?: Use smart meter
measurements of building supply temperature
as a replacement for network measurements.
Henrik Aalborg Nielsen, ENFOR A/S
15
CITIES 2nd General Consortium Meeting 25-26 May, 2015, DTU, Lyngby
Målinger
ENFOR
Screening - koncept
Estimation for individuelle
huse / husholdninger
Baggrundsdata, hus N:
…
Baggrundsdata, hus 2:
…
Baggrundsdata,
hus 1:
•
Opvarmningstype
•
Grundareal
•
…
Forbruger (hus) N:
…
Forbruger (hus) 2:
…
Forbruger
(hus) 1:
•
•
•
Følsomhed for udetemp., sol, vind.
Varmtvandsforbrug.
Fyrringssæsson.
Sammenligning på tværs
af huse / husholdninger
Liste over huse / husholdninger med
en usædvanlig høj klimafølsomhed,
højt varmtvandsforbrug, …
Henrik Aalborg Nielsen, ENFOR A/S
16
CITIES 2nd General Consortium Meeting 25-26 May, 2015, DTU, Lyngby
Thank you for your attention!
Henrik Aalborg Nielsen, ENFOR A/S
17
ENFOR