1. EU PVSEC 2017 Conference 5DO.5.6 Increasing the Efficiency of Photovoltaic (PV) Batteries through Non-Intrusive Load Monitoring Amsterdam
Presented by Patrick Baumann from the Zurich University of Applied Science in Winterthur

2. Patrick Baumann, ZHAW School of Engineering, Mail: baup@zhaw.ch
Increasing the Efficiency of Photovoltaic (PV) Batteries through Non-Intrusive Load Monitoring
Authors:
Patrick Baumann, ZHAW School of Engineering, Mail:
Pirmin Held, HFU Furtwangen University, Mail:
Prof. Dr. Andreas Heinzelmann, ZHAW School of Engineering, Mail:
Prof. Dr. Dirk Benyoucef, HFU Furtwangen University, Mail:
Prof. Dr. Franz Baumgartner, ZHAW School of Engineering, Mail:

3. Agenda Abstract Background Methods Measurements Results Conclusions
Increasing the Efficiency of Photovoltaic (PV) Batteries through Non-Intrusive Load Monitoring
Agenda
Abstract
Background
Methods
Measurements
Results
Conclusions

4. Increasing the Efficiency of Photovoltaic (PV) Batteries through Non-Intrusive Load Monitoring
Abstract
Main goal: Analyze and optimize the energy usage of a battery-storage and a PV plant.
Get an idea on how to deploy the valuable energy in an expensive battery for private-household in the best way
Developing and using of a new measuring tool to identify load current in a simple and economical way

5. High losses of valuable and expensive energy
Increasing the Efficiency of Photovoltaic (PV) Batteries through Non-Intrusive Load Monitoring
Background
The efficiency of an inverter is much better at high load (close to max. Power, η > 90%)
Same counts for the whole battery storage system
About 60% of the energy used is rated below 1kW power value
η in this power range is around 60 to 85%
This means, much energy provided by a battery storage is used to heat the electronics
High losses of valuable and expensive energy

6. Increasing the Efficiency of Photovoltaic (PV) Batteries through Non-Intrusive Load Monitoring
Distribution curve of the power
to be compensated by the
battery (above) and the
distribution of the load (both
below, linear & logarithmic).
Amount / %
Most power < 1kW
Batterypower / kW
Amount / %
Load (kW)
Efficiency quite bad
Efficiency of the path BAT2AC depending of the discharge power compared to the load distribution over a synthectic year
Efficiency / %
Amount / %
Source: Li-Ionen Heimspeichersysteme: Performance auf dem Prüfstand
Munzke, N.; Schwarz, B.; Büchle, F.; Barry, J.
Power / kW

7. Deploy the battery energy only for high power devices
Increasing the Efficiency of Photovoltaic (PV) Batteries through Non-Intrusive Load Monitoring
Method / Idea
Deploy the battery energy only for high power devices
Thus drive the inverter in a high efficiency range
Need to know when high power is activated
Recognize this as fast as possible to minimize the losses
Need of a fast measurement unit for “high” power
Simple and inexpensive to install
Fast algorithms to detect high power devices in the load current

8. Measurement: Non-Intrusive Load Monitoring, the «new way»
Increasing the Efficiency of Photovoltaic (PV) Batteries through Non-Intrusive Load Monitoring
Measurement: Non-Intrusive Load Monitoring, the «new way»
(Goal: Recognize powerful devices as fast as possible to supply it with the battery)
Predict the power usage shortly after the device was identified; NOT the allocation of electricity consumption to a certain consumer
Devices and their usage must be known / automatically learned

9. Measurement: Non-Intrusive Load Monitoring, the «new way» Innovation:
Increasing the Efficiency of Photovoltaic (PV) Batteries through Non-Intrusive Load Monitoring
Measurement: Non-Intrusive Load Monitoring, the «new way»
Innovation:
No averaging of a period, thus more detailed informations
FIT-PS: Frequency invariant transformation of periodic signals
Interpolation and reshaping to a «perfect» period results in better representation of the signal, thus less information losses
Devices and their usage must be known / automatically learned

10. Signal analysis using FIT-PS
Separate the voltage signal into its individual periods.
Find a significant point in the voltage signal e.g. the zero crossing.
Mains frequency varies
Deviations sample rate of the measuring unit
Interpolation within one period
Same number of samples in every period
Source: HFU Furtwangen

11. Signal analysis using FIT-PS
Using the voltage to trigger the current.
Reshape the signal
The number of dimensions of the new features resulting from the number of measurement points within one period.
Source: HFU Furtwangen

12. High sampled signals (12 kHz)
Increasing the Efficiency of Photovoltaic (PV) Batteries through Non-Intrusive Load Monitoring
High sampled signals (12 kHz)
lower amplitude
no symmetry
higher amplitude
phase shift
Phasenverschiebung am nulldurchgang, weitere information die nur so zum tragen kommt, da bei Q P verfahren nicht auffällt
Bei 4200 wird ein gerät eingeschaltet, stromsignal ändert sich, bzw nulldurchgang verschiebt sich, maxima bleiben
Source: HFU Furtwangen

13. High sampled signals (12 kHz), in detail:
Increasing the Efficiency of Photovoltaic (PV) Batteries through Non-Intrusive Load Monitoring
High sampled signals (12 kHz), in detail:
device is switched on
transient state
phase shift
Phasenverschiebung am nulldurchgang, weitere information die nur so zum tragen kommt, da bei Q P verfahren nicht auffällt
Bei 4200 wird ein gerät eingeschaltet, stromsignal ändert sich, bzw nulldurchgang verschiebt sich, maxima bleiben
Source: HFU Furtwangen

14. Simulation in Matlab / Simulink with real parameters
Increasing the Efficiency of Photovoltaic (PV) Batteries through Non-Intrusive Load Monitoring
Simulation in Matlab / Simulink with real parameters
Simulation used to find best strategy of depolying energy from battery storage
Direct modelling (and deploying?) of different algorithms
Effects like aging of the battery, non-linearities in relation to current load etc. also regarded
η of Battery to AC conversion,
modelled in Simulink

15. Prediction in Matlab / Simulink (work in progress)
Increasing the Efficiency of Photovoltaic (PV) Batteries through Non-Intrusive Load Monitoring
Prediction in Matlab / Simulink (work in progress)
Energy consumption highly depends on human behaviour / activities
Prediction / modelling of energy consumption in near future to further optimize the usage of the energy from the battery
Derived from real household data
Device usage
Prediction
Oven
PC-Workstation
Blender
Stove
Heater
Time

16. New measurement unit, with very high rate (~98%) of event detection
Increasing the Efficiency of Photovoltaic (PV) Batteries through Non-Intrusive Load Monitoring
Results: Measurement
New measurement unit, with very high rate (~98%) of event detection
In-depth simulations which allows a deep insight an optimization of different «energy strategies»
Source: HFU Furtwangen

17. Results: Simulation of different strategies
Increasing the Efficiency of Photovoltaic (PV) Batteries through Non-Intrusive Load Monitoring
Results: Simulation of different strategies

18. There is a potential to improve the usage of battery storage
Increasing the Efficiency of Photovoltaic (PV) Batteries through Non-Intrusive Load Monitoring
Conclusions
There is a potential to improve the usage of battery storage
Estimations pending (depending on load between 5-10% more energy aviable)
Also potential in other application areas
Since activities of daily living are highly linked to energy
A «new control device» must operate with existing batteries for investment reasons
Looking for industrial partners for further development

19. THE END Thank you Questions?
Patrick Baumann, ZHAW School of Engineering, Mail:
Pirmin Held, HFU Furtwangen University, Mail:
Prof. Dr. Andreas Heinzelmann, ZHAW School of Engineering, Mail:
Prof. Dr. Dirk Benyoucef, HFU Furtwangen University, Mail:
Prof. Dr. Franz Baumgartner, ZHAW School of Engineering, Mail: