An increasing number of photovoltaic (PV) systems are being installed worldwide and residential sector is responsible for a large part of this growth. Small scale PV systems do not have complex measuring devices and their breakdowns are not spotted immediately by the system owners. This might lead to prolonged time without generating power and creating both financial loss and environmental damage. This thesis presents a method of PV yield nowcasting laying foundations for remote monitoring. Early detection of faults is the first step towards eliminating the described issues. In this project four machine learning models for predicting solar yield were developed: ElasticNet, Polynomial Regression, Random Forest, and Extreme Gradient Boosting (XGBoost). The models were created both for daily and hourly data sets, as some inverters can log daily yields only. In both cases, the utilized data set consisted of data for the time between July 1st, 2018 and June 30th, 2019 and corresponding to 1,102 PV systems which is five times more than the largest data set studied in the found literature. The average PV system size in the data set is 4.44 kWp. Utilized inputs next to weather data and previous yields included shading factor describing fraction of direct light unable to reach PV system due to surrounding obstacles. Calculation of shading facor was based on 360⁰ pictures taken at the site. XGBoost algorithm turned out to be the most suitable for the task of PV yield nowcasting obtaining RMSE of 1.48 kWh and MAE of 0.877 kWh for hourly data aggregated to daily values and evaluated on future time steps. Currently used commercial software of Solar Monkey has RMSE equal 2.237 kWh and MAE equal 1.5 kWh. XGBoost model trained on daily data obtained RMSE 1.185 kWh and MAE 0.698 kWh outperforming hourly model most likely due to utilization of Hidden Markov Model for data cleaning. Next to overall performance, per system metrics were calculated for the hourly XGBoost. Mean individual RMSE for previously seen systems is 1.656 kWh while for unseen systems it equals 1.666 kWh. This means the model scales well to previously unseen systems and implies that its parallelized version is not necessary. Also, the model's learning saturates after seeing data corresponding to one year and 278 PV systems. Precalculation of GPOA worsened performance with respect to the model utilizing GHI. Hourly XGBoost has hourly RMSE of 0.281 kWh under clear sky and 0.377 kWh under partly cloudy sky which indicates it is more mistaken for cloudy conditions. This could be caused by low quality of cloud coverage data. The model also has large relative errors for small irradiance values which occur mostly in January and December, as well as just after sunrise and just before sunset. This issue is caused by using squared error as loss function during model training. Despite these shortcomings, the conclusive results recommend industrial implementation of the developed model.

A quick-scan algorithm has been developed in order to evaluate rooftop PV potential in the Netherlands. Both its panel fitting and yield prediction functions have been validated with existing systems monitored by Solar Monkey. The calculation times of different parts of the algorithm were measured and decreased while keeping the accuracy of the algorithm in the desired range.

First, a new approach to determine the roof segment orientation was introduced, using both the normal vector and the longest side of the roof segment polygon. A visual inspection was carried out, in which recent aerial images were compared to 3D roof segments that were provided by Readaar. For 145 roofs, the roof segments on which PV was placed were manually selected, such that they could validate the quick-scan. From the visual inspection it was deduced that in-roof obstacles were often not detected. Sometimes the customer had no desire to use the full potential of the roof, or preferred a rectangular panel layout instead of fitting the maximum amount of panels. Another finding was that the distance kept from the roof segment edge was much smaller than initially expected. For pitched roofs, there was virtually no distance between the roof edge and installed panels, while for flat roofs around 20 cm was kept. Using zero distance from the roof edge, the panel placement algorithm still underestimates the roof potential by 17.5% on average, with a relative standard deviation of 46.3%.

Three yield calculation methods were compared: the Solar Monkey method, the SVF & SCF method, and the method without obstacles. The predicted performance or final annual AC yield was compared with the actual measured performance, both measured in kWh/kW_p per year. For the three methods, relative standard deviation values of 7.2%, 7.5% and 9.1% were found respectively. The three methods could generate yield predictions for 91.0%, 92.7%, and 93.1% of the 145 roofs. For highly shaded roofs with Sun Coverage Factor values above 0.25, the method neglecting obstacles performed significantly worse. Additionally, the performance of large roofs with an average segment area above 70m² was generally under-predicted, while the relative standard deviation was highest. It is expected that using one obstacle view is not accurate enough for large roof segments.

The computational speed of the panel fitting algorithm for pitched roofs without internal obstacle segments was found to be 20.1± 5.0 m²s^-1, whereas for flat roofs it was found to be 56.9±12.0 m²s^-1. In order to optimise the quick-scan algorithm in both speed and accuracy, two filtering steps were carried out. Segments with a pitch angle over 10° and an orientation between 0° to 60° or 300° to 360º were filtered out, since they would have an annual performance below 650 kWh/kW_p. Moreover, segments with an area less than 8.4 m² were filtered out, since they were observed to fit less than 2 panels. The quick-scan calculation times for different yield prediction methods were found to be 15.50±1.01, 14.58±1.13 and 2.75±0.44 seconds per roof, respectively. These times were measured for data sets that had around 2 segments per roof after filtering on segment area and pitched segment orientation. It can be concluded that the method without obstacles is preferred when the calculation time is a limiting factor, whereas for accuracy the Solar Monkey method is preferred.