A method for online monitoring and intelligent diagnosis of power generation efficiency of a photovoltaic module

Abstract

The application discloses the technical field of photovoltaic module, and particularly relates to a photovoltaic module power generation efficiency online monitoring and intelligent diagnosis method, which comprises the following steps: step one, multi-dimensional synchronous acquisition; step two, benchmark efficiency dynamic modeling; step three, online efficiency deviation calculation; step four, multi-feature fusion abnormality determination; step five, intelligent classification of fault modes; and step six, positioning and hierarchical output. The application has the beneficial effects that: manual analysis is not required, rapid fault determination is realized, the diagnosis efficiency is greatly improved, the abnormal area and abnormal data category can be quickly positioned, accurate fault positioning is realized, operation and maintenance instructions containing fault positions, types and processing suggestions can be automatically generated, fault levels are divided according to influence degrees for preferential processing, the response time is significantly shortened, the power generation loss is reduced, and the operation and maintenance cost is reduced.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic module technology, specifically to a method for online monitoring and intelligent diagnosis of photovoltaic module power generation efficiency. Background Technology

[0002] Photovoltaics, or solar photovoltaic power generation, is a clean energy technology that uses the photovoltaic effect to directly convert solar radiation energy into electrical energy. It relies on photovoltaic modules to complete the photoelectric conversion. It requires no fuel, produces no noise, and causes no pollution. It can supply power through grid connection or off-grid methods and is widely used in power plant construction, industrial and commercial plants, residential rooftops, transportation facilities, and various new energy power supply scenarios. It is currently the most widely promoted and most mature form of renewable energy in the world.

[0003] Photovoltaic modules are the core power generation units that directly convert solar energy into electrical energy. They are usually composed of multiple photovoltaic cells connected in series and parallel and packaged together. They are combined with tempered glass, backsheet, frame and junction box to form an overall protective structure. They can continuously output direct current through the photoelectric effect under sunlight. They are the most basic and critical power generation components in photovoltaic power plants, distributed photovoltaics and residential photovoltaic systems. They are widely used in industrial and commercial, residential and various new energy power supply scenarios.

[0004] However, existing photovoltaic module monitoring and diagnostic methods mostly rely on manual experience analysis for fault diagnosis, which cannot achieve automatic classification and accurate location. Fault identification is lagging, processing efficiency is low, and there is no ability to collect and compare data in different areas. It is difficult to quickly locate abnormal areas and specific fault types, and maintenance personnel cannot obtain clear processing instructions in a timely manner, resulting in large losses in photovoltaic power generation and high operation and maintenance costs.

[0005] To address this, we propose an online monitoring and intelligent diagnostic method for photovoltaic module power generation efficiency. Summary of the Invention

[0006] In view of the problems existing in the above and / or existing methods for online monitoring and intelligent diagnosis of photovoltaic module power generation efficiency, the present invention is proposed.

[0007] Therefore, the purpose of this invention is to provide an online monitoring and intelligent diagnosis method for photovoltaic module power generation efficiency. By improving the existing photovoltaic module monitoring and diagnosis methods, this invention can solve the problems mentioned above. In terms of fault diagnosis, the existing photovoltaic module monitoring and diagnosis methods mostly rely on manual experience analysis, which cannot achieve automatic classification and accurate location. Fault identification is lagging, processing efficiency is low, and there is no ability to collect and compare data in different areas. It is difficult to quickly locate abnormal areas and specific fault types, and maintenance personnel cannot obtain clear processing instructions in a timely manner, resulting in large losses in photovoltaic power generation and high operation and maintenance costs.

[0008] To address the aforementioned technical problems, according to one aspect of the present invention, the present invention provides the following technical solution:

[0009] A method for online monitoring and intelligent diagnosis of photovoltaic module power generation efficiency, the method comprising the following steps:

[0010] Step 1: Multi-dimensional synchronous data acquisition. Real-time acquisition of photovoltaic module IV curve data, output voltage and current, operating temperature, irradiance, ambient temperature and humidity, junction box temperature and historical power generation data through edge acquisition terminals. Data acquisition is performed in partitions and time-series aligned according to module topology areas to form a unified time-series dataset.

[0011] Step 2: Dynamic modeling of benchmark efficiency. Based on real-time irradiance and module temperature, a temperature-irradiance dual-factor correction theoretical power generation model is established to calculate the theoretical maximum power generation efficiency of photovoltaic modules under the current environment, forming a dynamic benchmark value.

[0012] Step 3: Online efficiency deviation calculation. The difference and ratio between the real-time collected actual power generation and the theoretical reference power are calculated to obtain the relative efficiency decay rate and the absolute efficiency deviation value.

[0013] Step 4: Multi-feature fusion anomaly determination. Efficiency deviation, voltage dispersion, current fluctuation coefficient, and temperature anomaly gradient are fused together. Based on preset thresholds and statistical discrimination rules, it is determined whether the component has an anomaly.

[0014] Step 5: Intelligent classification of fault modes. Based on the fusion feature input, a lightweight classification model is used to identify seven typical faults: occlusion, dust, hot spots, PID attenuation, diode failure, component breakage, and branch abnormality.

[0015] Step 6: Location and Hierarchical Output. Based on the data collected from the partitions and the topology relationship of the components, the fault is accurately located, and the fault level is classified according to the degree of impact on power generation, generating alarms and operation and maintenance instructions.

[0016] As a preferred embodiment of the photovoltaic module power generation efficiency online monitoring and intelligent diagnosis method described in this invention, the multi-dimensional synchronous acquisition in step one divides different photovoltaic module areas and collects the IV curve data, output voltage and current, operating temperature, irradiance, ambient temperature and humidity, junction box temperature and historical power generation data of photovoltaic modules in different areas respectively. The collected data are compared with the same type of data to determine whether there is abnormal data in the photovoltaic modules in different areas, which can ensure that the photovoltaic modules in different areas can generate electricity normally.

[0017] The location and hierarchical output described in step six involves acquiring comparative data on the IV curve data, output voltage and current, operating temperature, irradiance, ambient temperature and humidity, junction box temperature, and historical power generation data of photovoltaic modules in different regions when an anomaly occurs. This helps identify abnormal data in these data, pinpointing the areas and categories of anomalies. Locating the abnormal areas helps staff locate the location, and identifying the abnormal data categories helps staff find corresponding solutions immediately, thus improving the efficiency of staff in handling anomalies.

[0018] As a preferred embodiment of the online monitoring and intelligent diagnosis method for photovoltaic module power generation efficiency described in this invention, the dynamic modeling of the benchmark efficiency in step two can automatically adjust the benchmark parameters according to the actual operating environment of the photovoltaic module, so that the calculation results of the theoretical power generation efficiency are more in line with the on-site working conditions, and improve the accuracy of subsequent efficiency deviation judgment.

[0019] As a preferred embodiment of the online monitoring and intelligent diagnosis method for photovoltaic module power generation efficiency described in this invention, the online efficiency deviation calculation in step three can continuously track the changing trend of power generation efficiency, making it easier to identify the efficiency degradation of photovoltaic modules in a timely manner.

[0020] As a preferred embodiment of the online monitoring and intelligent diagnosis method for photovoltaic module power generation efficiency described in this invention, the multi-feature fusion anomaly determination in step four comprehensively judges multiple operating features, which can reduce misjudgments caused by single data fluctuations and improve the reliability of anomaly identification.

[0021] As a preferred embodiment of the online monitoring and intelligent diagnosis method for photovoltaic module power generation efficiency described in this invention, the statistical discrimination rule in step four can automatically filter out interference data caused by environmental fluctuations, ensuring that only real faults are judged.

[0022] As a preferred embodiment of the online monitoring and intelligent diagnosis method for photovoltaic module power generation efficiency described in this invention, the fault mode intelligent classification in step five can distinguish and identify different types of faults, facilitating targeted corresponding treatment measures.

[0023] As a preferred embodiment of the online monitoring and intelligent diagnosis method for photovoltaic module power generation efficiency described in this invention, the lightweight classification model described in step five can run directly on the edge acquisition terminal to achieve real-time fault diagnosis without relying on external computing devices.

[0024] As a preferred embodiment of the online monitoring and intelligent diagnosis method for photovoltaic module power generation efficiency described in this invention, the fault level classification in step six is ​​based on the degree of impact of the fault on power generation efficiency, which facilitates prioritizing the handling of faults with greater impact.

[0025] As a preferred embodiment of the online monitoring and intelligent diagnosis method for photovoltaic module power generation efficiency described in this invention, the operation and maintenance instructions in step six include the fault location, fault type, and handling suggestions, which can directly guide operation and maintenance personnel to carry out their work.

[0026] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0027] 1. In this invention, a lightweight classification model is used to automatically classify and identify seven typical faults in real time at the edge terminal without manual analysis, enabling rapid fault determination and significantly improving diagnostic efficiency. By adopting a partitioned collection and horizontal comparison of similar data, abnormal areas and abnormal data categories can be quickly located, achieving accurate fault location and avoiding blind troubleshooting. It can automatically generate operation and maintenance instructions containing fault location, type, and handling suggestions, and prioritize fault handling according to the degree of impact, significantly shortening response time, reducing power generation loss, and lowering operation and maintenance costs.

[0028] 2. In this invention, a dynamic benchmark model is constructed using a temperature-irradiation dual-factor correction theoretical power generation model. This model can calculate the theoretical power generation efficiency in real time, closely matching the actual operating conditions of the site. It accurately reflects the degradation of the actual power generation performance of the components, thus solving the problem of inaccurate efficiency assessment. By fusing multiple features such as efficiency deviation, voltage dispersion, current fluctuation coefficient, and temperature anomaly gradient, and filtering environmental interference data through statistical discrimination rules, false judgments are significantly reduced, and the reliability of anomaly identification is improved. By continuously tracking the trend of power generation efficiency changes through online efficiency deviation calculation, latent degradation of the components can be identified in advance, enabling preventive monitoring and avoiding the expansion of faults.

[0029] 3. In this invention, by collecting data synchronously from multiple dimensions and comparing data in different areas, comprehensive operating data of photovoltaic modules can be obtained in real time, abnormal data can be quickly identified, stable power generation of the modules can be ensured, and power generation loss caused by the expansion of abnormalities can be avoided. By continuously tracking changes in power generation efficiency through online efficiency deviation calculation, hidden degradation and early faults of the modules can be detected in a timely manner, achieving preventive monitoring. Multiple operating characteristics are integrated for judgment, and interference data is filtered through statistical rules, which effectively reduces misjudgment caused by fluctuations in single data and improves the reliability of anomaly identification. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the diagnostic method steps of the present invention;

[0031] Figure 2This is a detailed flowchart of the diagnostic method of the present invention;

[0032] Figure 3 This is a schematic diagram of the multi-dimensional partitioning data acquisition method of the present invention;

[0033] Figure 4 This is a schematic diagram illustrating the baseline efficiency of the present invention;

[0034] Figure 5 This is a schematic diagram of the multi-feature fusion of the present invention;

[0035] Figure 6 This is a schematic diagram of the fault classification of the present invention. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0037] This invention provides an online monitoring and intelligent diagnosis method for photovoltaic module power generation efficiency. It features automatic classification and identification of seven typical faults in real time at the edge terminal through a lightweight classification model, eliminating the need for manual analysis and enabling rapid fault determination, thus significantly improving diagnostic efficiency. By adopting a partitioned data collection and horizontal comparison of similar data, it can quickly locate abnormal areas and abnormal data categories, achieving accurate fault location and avoiding blind troubleshooting. It can automatically generate operation and maintenance instructions containing fault location, type, and handling suggestions, and prioritize fault handling according to the degree of impact, significantly shortening response time, reducing power generation loss, and lowering operation and maintenance costs.

[0038] Please see Figure 1-6 A method for online monitoring and intelligent diagnosis of photovoltaic module power generation efficiency, the method comprising the following steps:

[0039] Step 1: Multi-dimensional synchronous data acquisition. Real-time acquisition of photovoltaic module IV curve data, output voltage and current, operating temperature, irradiance, ambient temperature and humidity, junction box temperature, and historical power generation data is obtained through edge acquisition terminals. Data is collected in partitions according to module topology and aligned with time series to form a unified time-series dataset. Step 2: Dynamic modeling of benchmark efficiency. Based on real-time irradiance and module temperature, a temperature-irradiance dual-factor correction theoretical power generation model is established to calculate the theoretical maximum power generation efficiency of the photovoltaic module under the current environment, forming a dynamic benchmark value. Step 3: Online efficiency deviation calculation. The difference and ratio between the real-time acquired actual power generation and the theoretical benchmark power are calculated to obtain the relative efficiency degradation rate and absolute efficiency deviation value. Step 4: Multi-feature fusion anomaly judgment. Efficiency deviation, voltage dispersion, current fluctuation coefficient, and temperature anomaly gradient are fused. Based on preset thresholds and statistical discrimination rules, it is determined whether the module has experienced an anomaly.

[0040] Step 5: Intelligent Fault Mode Classification. Based on the fusion feature input lightweight classification model, seven typical faults are identified: shading, dust, hot spots, PID attenuation, diode failure, component breakage, and branch anomaly. Step 6: Location and Hierarchical Output. Faults are accurately located based on the partitioned data collection and component topology, and fault levels are classified according to their impact on power generation, generating alarms and maintenance instructions. The lightweight classification model automatically classifies and identifies the seven typical faults in real time at the edge terminal, eliminating the need for manual analysis and enabling rapid fault determination, significantly improving diagnostic efficiency. By adopting partitioned data collection and horizontal comparison of similar data, abnormal areas and abnormal data categories can be quickly located, achieving accurate fault location and avoiding blind troubleshooting. Maintenance instructions containing fault location, type, and handling suggestions can be automatically generated. Fault levels are classified according to their impact for priority handling, significantly shortening response time, reducing power generation loss, and lowering maintenance costs.

[0041] In step one, multi-dimensional synchronous data acquisition divides different photovoltaic module areas and collects IV curve data, output voltage and current, operating temperature, irradiance, ambient temperature and humidity, junction box temperature, and historical power generation data for photovoltaic modules in different areas. The collected data are then compared with similar data to determine if there are any abnormal data in the photovoltaic modules of different areas, ensuring that the photovoltaic modules in different areas can generate electricity normally. In step six, location and hierarchical output, if an anomaly occurs, compares the IV curve data, output voltage and current, operating temperature, irradiance, ambient temperature and humidity, junction box temperature, and historical power generation data of photovoltaic modules in different areas to identify abnormal data. This allows for the location of the anomaly and the identification of the abnormal data category, helping staff to quickly find corresponding solutions and improving the efficiency of handling anomalies.

[0042] Step two, dynamic modeling of the baseline efficiency, can automatically adjust the baseline parameters according to the actual operating environment of the photovoltaic modules, making the calculated results of theoretical power generation efficiency more consistent with the on-site operating conditions and improving the accuracy of subsequent efficiency deviation judgment. Step three, online efficiency deviation calculation, can continuously track the changing trend of power generation efficiency, making it easier to identify the efficiency degradation of photovoltaic modules in a timely manner. Step four, multi-feature fusion anomaly judgment, comprehensively judges multiple operating features, which can reduce the misjudgment caused by single data fluctuations and improve the reliability of anomaly identification. Step four, statistical discrimination rules, can automatically filter out interference data caused by environmental fluctuations, ensuring that only real faults are judged.

[0043] Step five, intelligent fault mode classification, can distinguish and identify different types of faults, facilitating targeted handling measures. The lightweight classification model in step five can run directly on the edge acquisition terminal, enabling real-time fault diagnosis without relying on external computing devices. Step six, fault level classification, distinguishes faults based on their impact on power generation efficiency, facilitating the priority handling of faults with greater impact. Step six, operation and maintenance instructions, includes fault location, fault type, and handling suggestions, which can directly guide operation and maintenance personnel in their work.

[0044] The workflow of this invention is as follows: First, multi-dimensional synchronous data acquisition is performed. The edge acquisition terminal divides the photovoltaic array into partitions according to the preset photovoltaic module topology areas, and synchronously acquires data from photovoltaic modules in different areas. Real-time acquisition of IV curve data, output voltage and current, operating temperature, irradiance, ambient temperature and humidity, junction box temperature, and historical power generation data is obtained. During the acquisition process, all data are time-series aligned, integrating data from the same time and area to form a unified time-series dataset, ensuring data consistency and availability. At the same time, the same type of data collected from different areas is compared horizontally to determine whether there are any abnormal data that deviate significantly from the normal range, and to initially screen out suspicious anomalies, providing a reliable data foundation for subsequent analysis and ensuring that the operating status of modules in each area can be monitored in real time.

[0045] After data acquisition, a dynamic model of the baseline efficiency is performed. Based on the real-time collected irradiance intensity and module operating temperature, a temperature-irradiance dual-factor correction theoretical power generation model is established. This model automatically adjusts the baseline parameters according to the actual operating environment of the photovoltaic module. It does not rely on fixed standard conditions and can dynamically calculate the theoretical maximum power generation efficiency that the photovoltaic module can achieve under the current environment, forming a dynamic baseline value. The dynamic baseline is updated in real time with environmental changes, so that the theoretical power generation efficiency is highly consistent with the on-site operating conditions, avoiding deviations in subsequent judgments due to environmental differences, and providing a core basis for accurate assessment of power generation efficiency.

[0046] Next, online efficiency deviation calculation is performed. The actual power generation obtained in real time is compared with the dynamic benchmark value. By calculating the difference and ratio, the relative efficiency degradation rate and absolute efficiency deviation value of the photovoltaic module are obtained. The change of efficiency deviation is continuously tracked and the long-term trend is recorded. Once the efficiency degradation is found to be increasing continuously, it can be determined that the module may have performance degradation or abnormal potential, so as to realize real-time tracking and early warning of power generation status.

[0047] Subsequently, multi-feature fusion anomaly determination is performed. The system integrates four key features calculated as efficiency deviation, voltage dispersion, current fluctuation coefficient, and temperature anomaly gradient, instead of relying on a single indicator for judgment. At the same time, it combines preset thresholds and statistical discrimination rules to comprehensively analyze the data, automatically filtering out interference data caused by environmental fluctuations such as cloud cover and sudden changes in irradiance, excluding non-fault anomalies, and only judging actual component anomalies, effectively reducing the false judgment rate and improving the accuracy and reliability of anomaly identification.

[0048] Once an anomaly is detected, the system performs intelligent fault mode classification and inputs the fused features into a lightweight classification model. The model runs directly on the edge acquisition terminal without needing to be uploaded to the cloud or external computing devices, enabling real-time fault diagnosis. The system can automatically identify and distinguish seven typical faults: occlusion, dust, hot spots, PID attenuation, diode failure, component breakage, and branch abnormality, clarifying the specific fault type and providing accurate basis for subsequent processing.

[0049] Finally, the system performs location and hierarchical output. Based on the data collected from different zones and the topological relationship of photovoltaic modules, the system accurately locates the anomaly. By comparing the data differences in different areas, it identifies the specific area where the anomaly occurred and the corresponding anomaly data category, enabling maintenance personnel to quickly find the fault location. At the same time, the system classifies the fault level according to the degree of impact of the fault on power generation efficiency, prioritizing and marking faults with high impact and high risk for priority handling. Finally, the system generates complete alarm information and maintenance instructions containing fault location, fault type, and handling suggestions, which are directly pushed to the maintenance terminal to guide staff to quickly carry out maintenance work.

[0050] Although the present invention has been described above with reference to embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of the invention. In particular, as long as there is no structural conflict, the features in the disclosed embodiments can be combined with each other in any manner. The lack of an exhaustive description of these combinations in this specification is merely for the sake of brevity and resource conservation. Therefore, the present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A method for online monitoring and intelligent diagnosis of photovoltaic module power generation efficiency, characterized in that, The diagnostic method includes the following steps: Step 1: Multi-dimensional synchronous data acquisition. Real-time acquisition of photovoltaic module IV curve data, output voltage and current, operating temperature, irradiance, ambient temperature and humidity, junction box temperature and historical power generation data through edge acquisition terminals. Data acquisition is performed in partitions and time-series aligned according to module topology areas to form a unified time-series dataset. Step 2: Dynamic modeling of benchmark efficiency. Based on real-time irradiance and module temperature, a temperature-irradiance dual-factor correction theoretical power generation model is established to calculate the theoretical maximum power generation efficiency of photovoltaic modules under the current environment, forming a dynamic benchmark value. Step 3: Online efficiency deviation calculation. The difference and ratio between the real-time collected actual power generation and the theoretical reference power are calculated to obtain the relative efficiency decay rate and the absolute efficiency deviation value. Step 4: Multi-feature fusion anomaly determination. Efficiency deviation, voltage dispersion, current fluctuation coefficient, and temperature anomaly gradient are fused together. Based on preset thresholds and statistical discrimination rules, it is determined whether the component has an anomaly. Step 5: Intelligent classification of fault modes. Based on the fusion feature input, a lightweight classification model is used to identify seven typical faults: occlusion, dust, hot spots, PID attenuation, diode failure, component breakage, and branch abnormality. Step 6: Location and Hierarchical Output. Based on the data collected from the partitions and the topology relationship of the components, the fault is accurately located, and the fault level is classified according to the degree of impact on power generation, generating alarms and operation and maintenance instructions.

2. The method for online monitoring and intelligent diagnosis of photovoltaic module power generation efficiency according to claim 1, characterized in that, The multi-dimensional synchronous acquisition described in step one divides different photovoltaic module areas and collects the IV curve data, output voltage and current, operating temperature, irradiance, ambient temperature and humidity, junction box temperature and historical power generation data of photovoltaic modules in different areas. The collected data are compared with similar data to determine whether there is abnormal data in photovoltaic modules in different areas, which can ensure that photovoltaic modules in different areas can generate electricity normally. The location and hierarchical output described in step six involves acquiring comparative data on the IV curve data, output voltage and current, operating temperature, irradiance, ambient temperature and humidity, junction box temperature, and historical power generation data of photovoltaic modules in different regions when an anomaly occurs. This helps identify abnormal data in these data, pinpointing the areas and categories of anomalies. Locating the abnormal areas helps staff locate the location, and identifying the abnormal data categories helps staff find corresponding solutions immediately, thus improving the efficiency of staff in handling anomalies.

3. The method for online monitoring and intelligent diagnosis of photovoltaic module power generation efficiency according to claim 2, characterized in that, The dynamic modeling of the benchmark efficiency described in step two can automatically adjust the benchmark parameters according to the actual operating environment of the photovoltaic module, so that the calculation results of the theoretical power generation efficiency are more in line with the on-site operating conditions and improve the accuracy of subsequent efficiency deviation judgment.

4. The method for online monitoring and intelligent diagnosis of photovoltaic module power generation efficiency according to claim 3, characterized in that, The online efficiency deviation calculation described in step three can continuously track the changing trend of power generation efficiency, making it easier to identify the efficiency degradation of photovoltaic modules in a timely manner.

5. The method for online monitoring and intelligent diagnosis of photovoltaic module power generation efficiency according to claim 4, characterized in that, The multi-feature fusion anomaly determination described in step four integrates multiple operational features for comprehensive judgment, which can reduce misjudgments caused by fluctuations in a single data point and improve the reliability of anomaly identification.

6. The method for online monitoring and intelligent diagnosis of photovoltaic module power generation efficiency according to claim 5, characterized in that, The statistical discrimination rules described in step four can automatically filter out interference data caused by environmental fluctuations, ensuring that only real faults are judged.

7. The method for online monitoring and intelligent diagnosis of photovoltaic module power generation efficiency according to claim 6, characterized in that, The intelligent fault mode classification described in step five can distinguish and identify different types of faults, making it easier to take corresponding targeted measures.

8. The method for online monitoring and intelligent diagnosis of photovoltaic module power generation efficiency according to claim 7, characterized in that, The lightweight classification model described in step five can run directly on the edge acquisition terminal to achieve real-time fault diagnosis without relying on external computing devices.

9. The method for online monitoring and intelligent diagnosis of photovoltaic module power generation efficiency according to claim 8, characterized in that, The fault level classification described in step six is ​​based on the degree of impact of the fault on power generation efficiency, which facilitates prioritizing the handling of faults with greater impact.

10. The method for online monitoring and intelligent diagnosis of photovoltaic module power generation efficiency according to claim 9, characterized in that, The maintenance instructions described in step six include the fault location, fault type, and handling suggestions, which can directly guide maintenance personnel in carrying out their work.

Images