Solar Power Plant Monitoring, Prediction of Weather Conditions and Power Production with Machine Learning Techniques

Votes: 2
Views: 597
Consumer Products
Previous Next

This project is about solar power plant monitoring, prediction of weather conditions and power production with machine learning techniques.

We propose an integrated and low cost approach to weather forecasting, solar energy prediction and photovoltaic source simulation. For these tasks, different approaches would have to make use of SCADA for energy production, a different method for weather forecasting as well as an Internet of Things (IoT) approach, leading to higher cost for the consumers.

I. Hardware

⁎ Temperature sensors (LM35 to measure panel temperature and DHT22 for environmental temperature and humidity).
⁎ Current sensor. ⁎ Voltage divider as a sensor.
⁎ Machine Learning (ML) for solar irradiance prediction (Pyranometer SP Lite 2 for validation).
⁎ ESP8266 for data gathering ►ESP32.
⁎ Data transmission via Wi-Fi and a Raspberry Pi3 model b as gateway.

II. Monitoring

♦ Data that are being sent to a website and they are being collected to be big data for prediction with ML.
♦ User Interface (UI) displaying the real-time energy production and weather conditions.

III. Software / Weather Forecasting

▫ Comparison of our data with local weather stations and prediction of future energy production.
▫For the prediction task Grey Models and other ML techniques will be implemented.

● Solar photovoltaic systems are renewable energy sources that are widely used around the globe. This idea would be applied in solar/hybrid parks.

● This idea can monitor panel efficiency, production and detect faulty panels from a remote location with low installation cost via Node-RED. In addition, it can predict future energy production and weather conditions using Grey Models and other ML techniques.

● It can be produced with few hardware components such as a voltage divider, a current sensor, solar radiation sensor or it can be approximately found using ML (ANFIS etc.), temperature and humidity sensor, esp8266 or esp32, some ADCs and also a raspberry pi3 model B. The cost is significant lower than the already implemented ideas. Our final idea is to design a PCB that accommodates the required electronics for single inverter up to four string measurements and have a mesh network connecting the multiple nodes to the gateway.

Illustration 1: Node RED graph of the master device that requests and displays data in the UI.

Illustration 2: Schematic for one panel node to develop the multi-panel node. There is a power management that breaks the circuit when the voltage is lower than 7.5 V to protect the components, a current sensor, a voltage sensor, an ADC and the panel temperature sensor, the solar irradiance sensor and his signal amplifier, the ambient temperature and humidity sensor and the ESP8266.


Voting is closed!


  • Name:
    Petros Maragkos
  • Type of entry:
    Team members:
    Paulos Bousoulas
    Petros Savvakis
  • Profession:
  • Software used for this entry:
    Node RED , Word , Easy Eda , Raspbian
  • Patent status: