application of artificial intelligence in fault detection and classification of solar power plants and prediction of power generation of combined cycled power plants
abstract
solar energy is one of the most dependable renewable energy technologies, as it is feasible
almost everywhere globally and is environmentally friendly. photovoltaic-based renewable energy
systems are highly susceptible to power grid transients. their operation may suffer drastically
during faults in the solar arrays, power electronics, and the inverter. thus, it is vital to develop an
intelligent mechanism to detect any type of fault or abnormalities within the shortest possible time
that will increase reliability and decrease the maintenance cost of the solar system. to accomplish
that, in this research, different artificial intelligence (ai) techniques are utilized to develop
classification, fault detection, and optimization algorithms for solar photovoltaic (pv) panels.
initially, a convolutional neural network (cnn) model was designed to detect and classify pv
modules based on the images taken from the solar panels. it is found that the proposed cnn model
can identify the fault with an accuracy of 91.1% for binary (i.e., healthy and faulty) and 88.6% for
multi-classification (i.e. cracked, shadowy, dusty and normal). however, sometimes the fault in
the solar panel may not be detectable from the images of the solar panels. that is why an adaptive
neuro-fuzzy inference system (anfis) model is developed to detect and classify the defects of pv
systems based on the signals collected from the solar panels. the performance of the developed
defect detection and classification algorithms was tested using real-life solar farm datasets. the
performance of the proposed anfis-based fault detection scheme has been compared with
different machine learning algorithms. it is found from the comparative results that the proposed
anfis-based fault detection technique is robust and straightforward. thus, the developed anfisbased intelligent technique will enhance the reliability of the pv system by minimizing the
maintenance cost and saving energy.
finally, another anfis model is developed to predict the power generation in a combined
cycle power plant. the codes were written in matlab, and their validity is confirmed with the
available anfis toolboxes in matlab. the proposed anfis is found capable of successfully
predicting power generation with extremely high accuracy and being much faster than the built-in
anfis of matlab toolbox. thus, the developed anfis model could be utilized as a promising
tool for energy generation applications.