Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 12

Non-Intrusive Load

Monitoring
Abiola Ogundeji and Ogooluwa Ajayi

26/09/2023
Paper Review: Development of a low-cost IoT system based on LoRaWAN for monitoring variables related to
electrical energy consumption in low voltage networks.

The authors designed an hardware for measuring electrical measurement consumption

1. YHDC SCT016S for current measurement

2. Atmel, M90E3AS chip, a polyphase measurement IC for performing calculation off main electrical variables

3. ESP32

4. LoRa Module

Implementation is possible but time of completion may affect the Design challenge timeline as we are
only given 43 days after our proposal has been accepted.
Paper Review: Enerduino-pro: Smart meter led probe
using Arduino

The authors designed a system to measure Active and reactive power.


• They use a photoresistor to detect light pulses of smart power
meters
•From reading the light pulses, active and reactive power can be
measured with accuracy provided by the power meter every 1Wh or
1Wh
SAMPLING

• Low frequency sampling


1. Hardest way to distinguish between appliances

2. Low cost

3. Sampling frequency <= 1Hz

• Medium frequency sampling


• can capture harmonics

• sampling frequency 120Hz – 1kHz


SAMPLING cont’d

• High frequency

1. Can capture startup power spikes of appliance

2. Appliance can easily be characterize by their startup spikes

3. Expensive
LOW FREQUENCY SAMPLING

• Due to specific hardware needed for sampling at high frequency

and its high cost, low frequency sampling is preferable.

• To overcome the limitations of low frequency sampling, low

frequency dataset must have these characteristics


LOW FREQUENCY
SAMPLING cont’d
• Multiple aggregated readings needed

such as

•Active Power

•Reactive Power

•Apparent Power

As not all activations may be registered SOURCE: NILM lecture

in each readings
LOW FREQUENCY
SAMPLING cont’d

Same appliances with

similar power consumption

may have different cycles

due to usage patterns


SOURCE: NILM lecture
LOW FREQUENCY
SAMPLING cont’d

Usage of appliance may be

different with months. Heat

pumps may work more in the

SOURCE: NILM lecture


cold season and work less in the

hot season
LOW FREQUENCY
SAMPLING cont’d

Some appliance just have

two state, while some have

multiple state

SOURCE: NILM lecture


DATA TRANSMISSION MEDIUM

GSM MODULE WIFI


• Wide coverage • Limited coverage
• Slower • Faster
• Low power consumption • Consume more power
• More secure • Secure if properly configured
• Easily deployed • Require Wi-Fi infrastructure like
Mifi module
It appears that the GSM module will be
MACHINE LEARNING ALGORITHM, DISCRETE
WAVELET TRANSFORM and APP INTERFACE
• We are still researching the machine learning algorithm on how to
implement and deploy it.
• The discrete wavelet transform is needed to know when the
appliances are switched ON and OFF. We are still researching this
concept too.
• The app interface is

You might also like