Periodic Reporting for period 1 - NoSoilPV (Novel Soiling Identification Logics for Photovoltaics)
Periodo di rendicontazione: 2018-12-03 al 2020-12-02
NoSoilPV aimed to tackle an issue affecting the performance of photovoltaic (PV) system worldwide: soiling. Soiling consists of the accumulation of dust, particles, and contaminants on the surface of PV modules and absorbs, reflects, deflects part of the sunlight. Therefore, the amount of light reaching the photovoltaic cell and converted into electricity is reduced if soiling is not removed. Soiling is estimated to have caused, in 2018 alone, the loss of 3% to 4% of the energy produced by photovoltaic worldwide and these losses are expected to raise in future as more PV capacity is installed in high insulation and high soiling risk areas.
However, differently from other reliability issues affecting PV, soiling is reversible and can therefore be mitigated. Cleanings are still nowadays the most common soiling mitigation strategy, but they have a cost. Therefore, the revenues made with the recovered energy after a cleaning have to be larger than its costs for it to be profitable. This means that the frequency and the timing of cleanings have to be correctly set to make soiling mitigation profitable.
This project provided the community with better tools for tackling soiling, through the investigation of different aspects of soiling monitoring and modelling. As part of this project, improvements have been made in the field of soiling extraction, monitoring and economics. The factors affecting the cleaning schedule profitability have been identified and a model has been developed for the advanced optimization of soiling mitigation. These studies have opened the way to the possibility of predicting in advance the optimal cleaning schedule through the analysis of environmental parameters only, meaning that O&M activities can be planned even before the PV site is operational.
Several reasons have been motivating this project. Indeed, an optimized soiling mitigation strategy increases the PV capacity factors with only limited additional costs. These would lead to higher profits, potentially attracting even more investments in PV and therefore favoring the installation of new PV capacity.
However, differently from other reliability issues affecting PV, soiling is reversible and can therefore be mitigated. Cleanings are still nowadays the most common soiling mitigation strategy, but they have a cost. Therefore, the revenues made with the recovered energy after a cleaning have to be larger than its costs for it to be profitable. This means that the frequency and the timing of cleanings have to be correctly set to make soiling mitigation profitable.
This project provided the community with better tools for tackling soiling, through the investigation of different aspects of soiling monitoring and modelling. As part of this project, improvements have been made in the field of soiling extraction, monitoring and economics. The factors affecting the cleaning schedule profitability have been identified and a model has been developed for the advanced optimization of soiling mitigation. These studies have opened the way to the possibility of predicting in advance the optimal cleaning schedule through the analysis of environmental parameters only, meaning that O&M activities can be planned even before the PV site is operational.
Several reasons have been motivating this project. Indeed, an optimized soiling mitigation strategy increases the PV capacity factors with only limited additional costs. These would lead to higher profits, potentially attracting even more investments in PV and therefore favoring the installation of new PV capacity.
Since the beginning, the project has been focusing on improving the analysis, the monitoring and the prediction of soiling losses for cleaning optimization purposes. This has been pursued by investigating different monitoring and forecast approaches, based either on soiling sensors, on PV performance data or on environmental parameters.
Several methods have been tested for improving the extraction of soiling loss profiles from PV performance data. The best results were obtained by using a “cleanings and change points” approach, which consists in fitting the data points in between cleanings using multiple linear functions rather than a single one as previously done in the literature. This way, dry periods experiencing different deposition rates can be modelled more accurately and cleaning optimization can be also improved. Piecewise regression returned the best results when compared with methods based on change point algorithms, lowering the modeling error by up to 40% compared to the case in which change points were neglected.
The project has also explored the possibility of estimating current and future soiling losses using only environmental data, as it would allow quantifying the soiling losses even if no soiling monitors or PV data are available. An analysis of the seasonality of rainfalls, the most common natural cleaning events, was conducted using weather generators. These algorithms stochastically produce a number of potential daily rainfall patterns using historical data. It was found that these can be used to evaluate the seasonality of rainfalls at a site and to identify typical high and low soiling periods, an essential step in planning an adequate cleaning schedule. Also, the possibility of identifying in advance a recommended cleaning schedule for PV sites was successfully investigated: it was possible to determine a priori a cleaning date less than a week away from the actual optimal cleaning date using only rainfall and particle matter data. This will make it possible to estimate the impact of the soiling losses and the cost of soiling mitigation even prior to the operation of a PV system. This way, the site selection, the PV plant design and the O&M planning can be optimized in advance depending also on the predicted severity and seasonality of soiling.
Last, the project investigated the economic impact of soiling, and analyzed the factors that influence the profitability of soiling mitigation. The possibility of cleaning only selected soiled strings of a plant rather than the full system was discussed through the analysis of the losses of several strings of a PV system in Chile.
The project has produced several important outputs. Some results have already been presented in peer-review journals and additional works have been submitted and are currently under review. The results were also presented at international conferences and PV- and soiling-related workshops. Some of the work conducted during the project was also the subject of invited talks to international working groups’ meetings. Effort was also spent to share the knowledge on solar energy and soiling and to disseminate the objectives and the progresses of the project to the general public through social media and public engagement events.
Several methods have been tested for improving the extraction of soiling loss profiles from PV performance data. The best results were obtained by using a “cleanings and change points” approach, which consists in fitting the data points in between cleanings using multiple linear functions rather than a single one as previously done in the literature. This way, dry periods experiencing different deposition rates can be modelled more accurately and cleaning optimization can be also improved. Piecewise regression returned the best results when compared with methods based on change point algorithms, lowering the modeling error by up to 40% compared to the case in which change points were neglected.
The project has also explored the possibility of estimating current and future soiling losses using only environmental data, as it would allow quantifying the soiling losses even if no soiling monitors or PV data are available. An analysis of the seasonality of rainfalls, the most common natural cleaning events, was conducted using weather generators. These algorithms stochastically produce a number of potential daily rainfall patterns using historical data. It was found that these can be used to evaluate the seasonality of rainfalls at a site and to identify typical high and low soiling periods, an essential step in planning an adequate cleaning schedule. Also, the possibility of identifying in advance a recommended cleaning schedule for PV sites was successfully investigated: it was possible to determine a priori a cleaning date less than a week away from the actual optimal cleaning date using only rainfall and particle matter data. This will make it possible to estimate the impact of the soiling losses and the cost of soiling mitigation even prior to the operation of a PV system. This way, the site selection, the PV plant design and the O&M planning can be optimized in advance depending also on the predicted severity and seasonality of soiling.
Last, the project investigated the economic impact of soiling, and analyzed the factors that influence the profitability of soiling mitigation. The possibility of cleaning only selected soiled strings of a plant rather than the full system was discussed through the analysis of the losses of several strings of a PV system in Chile.
The project has produced several important outputs. Some results have already been presented in peer-review journals and additional works have been submitted and are currently under review. The results were also presented at international conferences and PV- and soiling-related workshops. Some of the work conducted during the project was also the subject of invited talks to international working groups’ meetings. Effort was also spent to share the knowledge on solar energy and soiling and to disseminate the objectives and the progresses of the project to the general public through social media and public engagement events.
The project provided the community with improved modelling tools and new knowledge to better address photovoltaic soiling and to minimize its effects. The use of piecewise regression in soiling modelling makes it possible to improve cleaning optimization as it enhances the detection of periods of high and low soiling deposition rates. The factors affecting the profitability of soiling mitigation have been analyzed and the possibility of predicting typical soiling profiles of a site in advance has been investigated. These findings will favor the optimization of the cleaning schedule and will contribute limiting the economic impact of soiling. Additional results of the work will allow lowering the costs of the soiling sensors and will make it possible to model the spectral impact of soiling on various PV technologies. These will be continued as part of a new research line launched within the hosting group.
The results of the cleaning optimization models have been shared with an industrial partner and will be put in place in their utility scale systems. The developed methodologies have been presented in the literature and can be immediately used to assess the soiling loss of any PV system, and to identify the most convenient cleaning schedules. Additional research will be carried out in the future to further refine the forecast of soiling and mitigation.
The results of the cleaning optimization models have been shared with an industrial partner and will be put in place in their utility scale systems. The developed methodologies have been presented in the literature and can be immediately used to assess the soiling loss of any PV system, and to identify the most convenient cleaning schedules. Additional research will be carried out in the future to further refine the forecast of soiling and mitigation.