6 methods to narrow down the causes of low power generation at the cluster level

When you feel that a solar PV system's "generation is low," looking only at the total output of the plant often won't reveal the cause. This is because many factors affect generation, such as weather, season, irradiance, temperature, shading, soiling, wiring, and equipment operating conditions. A useful approach is to divide the system into clusters and compare them, narrowing down the area of decline step by step. Here, a "cluster" refers to a generation area treated as a unit for on-site management, a power conditioner unit, an input-circuit unit, string groups, racking rows, groups of panels with similar orientation or tilt, and so on. Rather than viewing the whole at once, examining small groups with similar conditions makes it easier to determine whether an anomaly exists and to organize candidate causes.

Table of Contents

• Standardize cluster ranges to establish a foundation for comparison

• Normalize power generation by solar irradiance conditions and installed capacity

• Disentangle shadows, dirt, and orientation differences from time-of-day variations

• Identify electrical imbalances from voltage, current, and string differences

• Look for cluster-specific external factors during on-site inspections

• Keep records to verify recurrence and determine priorities for improvement.

• Summary

Align Cluster Ranges to Establish a Basis for Comparison

When narrowing down the cause of reduced power generation at the cluster level, the first thing to do is to clearly align the scope being compared. Even if you look at the entire plant and conclude that output is "lower than yesterday" or "lower than last year," if you don't know which section is declining or which equipment systems it is concentrated in, it becomes difficult to decide the order of inspections. To find the cause quickly, you should first divide the facilities into manageable units and organize them so they can be compared under the same conditions.

How clusters are divided depends on the site configuration. Options include dividing by power conditioner, by input circuit, by junction box, by racking row or section, or by groups of panels with the same azimuth or tilt. The important point is not to split simply by area, but to divide into units where generation conditions and electrical connectivity can be traced. For example, even within the same power plant, treating east-facing panel groups and south-facing panel groups as a single cluster will mix morning and midday generation differences. Forcing together areas with different tilt angles, azimuths, capacities, or connection destinations can cause normal differences to be mistaken for anomalies.

At the cluster level, it is easier to make judgments if you document each cluster’s equipment capacity, number of panels, connection destinations, installation orientation, tilt, and surrounding environment in a register. Rather than scrambling to search for drawings or site photos when power output is low, it is important to routinely maintain the ability to trace which cluster is connected to which equipment. Even if you only look at generation data, if you cannot correlate it with the physical layout, it will take time to narrow down the possible causes.

Also, cluster names and numbers need to be standardized on site. If the names on the monitoring screen, inspection records, drawings, and on-site labels are inconsistent, you may find a cluster with low power output but be unsure where to check in the field. This mismatch of names is especially likely to lead to missed inspections or duplicated checks at sites managed by multiple people. Even before starting cluster-level comparisons, simply aligning the names and the scope can greatly improve the accuracy of root-cause investigations.

At the stage of establishing a basis for comparison, it's important not to immediately assume a failure. Even if power output appears low, it may actually be caused by differences in capacity between clusters, shading, orientation differences, residual snow or dirt, missing communication data, or similar factors. In the initial stage, rather than concluding there is an anomaly, focus on creating units that can be compared under the same conditions. If clusters are correctly grouped, subsequent power output comparisons, voltage and current checks, and on-site inspections will all be easier to correlate.

Normalize power generation by solar irradiance and system capacity

When comparing power generation on a cluster basis, it is important not to judge solely by raw generation figures. If installed capacity differs between clusters, differences in generation are natural. Even with the same installed capacity, generation varies with solar irradiance, temperature, orientation, tilt, and the presence or absence of shading. Therefore, to narrow down the causes of low generation, you should first look at generation per unit of installed capacity and generation trends that take solar irradiance conditions into account.

For example, even if a cluster’s daily power output is lower than others, that can be a natural result if the cluster’s capacity is smaller. Conversely, if clusters have similar capacities and similar installation conditions but a large discrepancy persists, that is grounds to suspect abnormalities, degradation, soiling, shading, or connection faults. A common mistake in the field is judging a block as “low” based only on total power output. The correct approach is to check how much is being generated per unit of capacity and whether it is relatively dropping under the same solar irradiance conditions.

When checking solar irradiation conditions, it's easier to see trends if you look not only at clear-sky days but also at cloudy days, after rain, and during seasonal transitions. On clear days shadows and soiling tend to have a larger effect, while on cloudy days orientation differences may appear smaller. If a specific cluster drops significantly only during clear-sky periods and the differences become smaller on cloudy days, shadows, soiling, or the condition of the panel surface related to direct irradiance should be suspected. On the other hand, if values are consistently low regardless of weather, candidates include electrical connections, equipment shutdowns, settings, degradation, or communication/measurement issues.

The period you compare is also important. If you judge based on a single day’s generation, you can be affected by passing clouds, temporary shutdowns, or communication losses. You need to look at trends over several days to several weeks to determine whether the same cluster is repeatedly low or if the decline is temporary. In particular, if you feel that “power generation has suddenly dropped,” it is useful to find the day the decline began. Once you know the start date of the decline, you can cross-check it against records of work performed, power outages, equipment operations, mowing, cleaning, typhoons, heavy rain, snowfall, lightning strikes, and so on.

Also, when using measurements from pyranometers, thermometers, or similar instruments, confirm which cluster’s conditions those measurements represent. In a large power plant, cloud cover, shadowing, and terrain effects can vary even within the same site. If there is only one pyranometer, that single value may not fully describe all clusters. Solar irradiance conditions should be used only as an aid for comparison; ultimately, it is safer to make a judgment by combining the relative differences between clusters, changes by time of day, and on-site conditions.

By normalizing for equipment capacity and solar irradiance conditions, it becomes easier to separate normal differences from abnormal ones. In cluster-level root-cause investigations, rather than immediately moving on to parts replacement or detailed inspections, first confirming whether that cluster is truly underperforming and whether capacity differences or weather variations can explain it is the quickest way to reduce unnecessary work.

Isolate shadows, dirt, and orientation differences from time-of-day variations

When investigating a drop in power generation at the cluster level, looking only at daily totals can lead you to overlook the cause. Even if the daily totals appear similarly low, the candidate causes change when viewed by time of day—for example, it may be low only in the morning, drop only at midday, show differences only in the evening, or be low throughout the daytime. To narrow down the cause of low generation, it is important to compare the generation curves for each cluster and check in which time periods the differences occur.

If only a specific cluster shows a delayed ramp-up in the morning hours, shadows from eastern buildings, trees, terrain, racking rows, or equipment structures may be involved. If generation drops significantly only in the evening, check for shadows on the west side and azimuth conditions. If the drop occurs only around midday, candidates include soiling of the panel surface, bird damage, partial shading, shadows from racks or cables, shadows from surrounding structures, or effects of equipment-side control or temperature rise. Examining time-of-day patterns makes it easier to narrow down the directions and areas to inspect on site.

Orientation and tilt differences are also important when making time-of-day comparisons. East-facing clusters tend to generate more power in the morning, while west-facing clusters tend to generate more power in the afternoon. If south-facing and east-/west-facing clusters are compared as the same curve, normal generation patterns can be misinterpreted as anomalies. Therefore, as a basic rule, time-of-day comparisons should be made between clusters with as similar orientation and tilt as possible. When comparing clusters with different conditions, separate whether the differences that appear at certain times of day are meaningful or are simply natural generation tendencies due to the design.

The effects of soiling appear not only by time of day but also in changes after weather events. If differences narrow after rain, surface dirt or deposits are likely responsible. Conversely, if differences remain unchanged after rain, causes other than soiling should be considered. However, rain does not necessarily remove all dirt; bird droppings, mud splatter, fallen leaves, pollen, dust, and drainage marks can remain. Lower-row panels close to the ground, areas sheltered from wind, and locations surrounded by trees or bare soil can show differences in soiling at the cluster level.

When checking for shading, it is important not to finish after viewing the site just once. Shading changes with the seasons and the time of day. In winter the sun's altitude is lower, and shadows that were unlikely to appear in summer can lengthen. Morning and evening shadows also change in extent with the seasons, so it is ideal to inspect at a time close to the same conditions as the date and time when the power output decline occurred. Even if on-site inspection is difficult, you can assess the possibility of shading by checking whether drops in the generation curve occur at the same time every day. If the output falls at almost the same time each day, it is more likely caused by a fixed object's shadow than by clouds.

Comparisons by time of day are also useful for determining the priority of on-site responses. Clusters that are low throughout the day may indicate equipment issues or widespread soiling and require prompt inspection. Conversely, when output is reduced only for a short period in the morning, the prioritization of countermeasures should be considered while looking at the annual impact. By identifying not only that power generation is low but also when, by how much, and in which clusters the differences occur, root-cause investigations can proceed more practically.

Identifying electrical imbalances from voltage, current, and string differences

If a comparison of power generation by cluster shows a decline in a specific area, the next things to check are voltage, current, and string-to-string differences. The causes of low generated power are not limited to external factors such as shading or soiling. Electrical causes such as open circuits, poor contacts, blown fuses, faulty connectors, incorrect connections, reversed polarity, equipment stoppage, input circuit abnormalities, protective operations, and measurement errors can cause a decline at the cluster level. Because power generation alone cannot determine the cause, it is important to check a combination of electrical values.

When looking at voltage and current, first check for any extreme differences between clusters under identical conditions. If the voltage differs significantly, consider differences in the number of series-connected modules, the connection condition, the operation of the equipment, or an open/abnormal circuit. If the current is low, candidates include shading, dirt, panel defects, partial string shutdown, increased contact resistance, blown fuses, and so on. If both voltage and current are low, possibilities include the equipment not operating normally, insufficient input, or that the measurements themselves are not being taken correctly.

In sites where string-level comparisons are possible, examining differences between strings within the same cluster makes it easier to narrow down the cause. Strings that share the same orientation, the same tilt, and the same number of modules are expected to exhibit similar generation characteristics. If only some strings show low current, focus inspections on the panels, wiring, connections, fuses, shading, and soiling related to those strings. Whether the entire cluster is underperforming or only a few strings within the cluster are low greatly changes the inspection scope.

Electrical inspections also require attention to measurement safety. Solar power generation systems produce DC voltage during the day, creating a risk of electric shock and arcing. When taking actual measurements, you must confirm the equipment specifications, work procedures, personal protective equipment, isolation/shutdown procedures, work qualifications, and site rules, and you must not perform tasks that are beyond what is safe. It is dangerous to carelessly touch energized connection points just because you want to investigate the cause of low power output. Check trends using monitoring data and existing measurements, and, if necessary, have specialist personnel carry out inspections using safe methods.

When reading anomalies in voltage and current, check not only instantaneous values but also the time series. If a value is low only momentarily, it could be due to passing clouds or temporary shading, but if the same cluster remains low over a long period, equipment-side factors become more likely. There are also patterns such as being normal in the morning and dropping at noon, plateauing around a certain output, or the difference widening only on hot days. Causes that become apparent when combined with temporal changes include equipment temperature rise, protective control, poor ventilation, and the effects of connection resistance.

Do not overlook communication or measurement faults. Even if power output is shown as low, the cause may be missing data acquisition or an instrument malfunction rather than actual generation. If only one cluster suddenly drops to near zero, suspect communication stoppage or measurement loss in addition to equipment shutdown. Cross-check on-site meter readings, the monitoring screen, device displays, and historical data to determine whether this is a display issue or an actual decline in generation. When generation appears low, it is especially important to question the assumption that the visible data is correct.

Identify cluster-specific external factors during on-site verification

When underperforming clusters become apparent in the data, investigate external factors through on-site inspection. In solar power generation, causes that cannot be identified from data alone may be hidden at the site. Check for factors that can vary by cluster, such as shading, soiling, vegetation overgrowth, fallen leaves, bird damage, mud splatter, poor drainage, racking tilt, damage to panel surfaces, sagging cables, abnormalities around connectors, and the condition around junction boxes. When searching on-site for the cause of low power output, it is important to compare not only the degraded clusters but also clusters that are close to normal.

In on-site inspections, first pay attention to the time period when power generation is reduced. If a cluster is low in the morning, check for morning shadows; if a cluster is low in the evening, check for evening shadows. If a cluster is low throughout the daytime, broadly inspect for surface soiling, electrical abnormalities, orientation and tilt, airflow, and equipment condition. Visiting the site at times unrelated to the period of reduced power generation can lead to missing shadows or temporary environmental factors. When taking on-site photos, record the date and time, the direction, and the target cluster so that later comparisons are easier.

Vegetation shading is a factor to be aware of as a cause of declines at the cluster level. Even if there were no problems at installation, vegetation can grow over time and cast shadows on specific rows or edge clusters. Shadows from vegetation and surrounding objects are especially likely in clusters with panels located at low elevations or near site boundaries. If the difference in power generation narrows after mowing, vegetation shading may have been a major influence. However, mowing or tree removal is subject to safety management and site management rules, so take planned action after confirming power generation data and on‑site conditions.

Dirt and deposits can also be concentrated in particular clusters. In areas close to roads, dust and mud splashing tend to increase, while areas near trees are more likely to have fallen leaves, pollen, and bird damage. Panels with a shallow tilt may not shed dirt easily, causing it to accumulate at the lower edge. In locations with poor drainage, mud can splash up or water can remain after rain. If dirt is noticeable in clusters with low power output, you can confirm the impact by comparing power generation trends before and after cleaning.

Visible anomalies include cracked panels, discoloration, scorch marks, warped frames, settlement or tilting of the mounting structure, damaged cables, loosened cable ties, and abnormalities around connection points. These do not necessarily immediately result in a large drop in power generation, but if they are concentrated in the area experiencing a cluster-level decline, they should be treated as candidate causes. In particular, when generation falls after typhoons, heavy rain, snowfall, lightning strikes, earthquakes, or ground deformation at development sites, carefully inspect the site for physical changes.

What’s important in on-site inspection is not to immediately conclude that an observed phenomenon is the cause. For example, rather than assuming dirt is the cause of a drop in power generation, check whether there is a difference in soiling between the cluster with reduced generation and the normal cluster, whether it matches the time period of the decline, and whether it improved after cleaning or mowing. Multiple factors can overlap. If light soiling, partial shading, partial string malfunction, and communication loss occur simultaneously, individually small differences can appear as a large overall drop. Treat on-site inspection as a process of verifying the hypotheses suggested by the data to reduce misidentifying the cause.

Keep records to confirm recurrence and determine improvement priorities

Narrowing down the causes of low power generation at the cluster level cannot be completed with a single inspection. You need to identify candidate causes, implement countermeasures, and only after confirming how generation changed afterward can you judge the effectiveness of the response. For that reason, it is important to keep records of the investigation details and the rationale for your judgments. If a drop in generation recurs, having past records makes it easier to determine whether it is the same cause or a new one.

Items to record include the days when a drop in power generation was observed, the affected cluster, the comparison cluster, weather, solar irradiance trends, generation differences by time of day, voltage and current conditions, on-site photos, inspection results, actions taken, and subsequent changes in power generation. Rather than insisting on a specific format, it is important that a third party can follow the situation later. If the cluster name, date and time, inspector, and reasons for the judgment are recorded, it will be easier to repeat the same procedures at the next inspection.

When setting priorities for improvements, consider not only the magnitude of the power generation decline but also its persistence, extent, safety implications, and ease of response. Declines that persist for long periods or spread across multiple clusters take precedence over temporary, minor drops. Also, even if the impact on power generation is still small, issues that could affect safety—such as damaged cables or abnormalities at connection points—should be checked promptly. Relying solely on power output as the criterion risks overlooking signs that are important for safety.

In post-countermeasure checks, we judge not only by the single day immediately after the work but by multiple days with similar weather. Even if it appears temporarily improved after cleaning, that may be due to weather differences. When mowing, checking connections, rebooting devices, replacing parts, or reviewing settings, we compare days before and after the measures that are as close as possible in conditions. By continuing comparisons at the cluster level, it becomes easier to confirm whether the countermeasures truly worked or whether another cause remains.

Accumulating records also helps in understanding seasonal variations. If a cluster's generation consistently declines at the same time each year, seasonal factors such as seasonal shading, vegetation growth, fallen leaves, snow accumulation, pollen, or rising temperatures can be considered. Conversely, if generation suddenly drops around a certain day and does not recover afterward, equipment or connection issues, physical damage, or configuration changes are more likely suspected. Trends that are not visible from a single inspection become easier to grasp by keeping records at the cluster level.

It is also important to keep records in a state where they can be shared among stakeholders. If the plant manager, maintenance personnel, on-site workers, and the person receiving reports each hold different information, cause investigations will become indirect and time-consuming. Sharing cluster-level decline status, items already checked, items not yet checked, and the next areas to inspect reduces duplicated work and overlooked checks. The problem of low power generation can be addressed efficiently only when both data management and on-site management are in place.

Summary

When investigating the causes of low power generation, it is important not to judge solely by the plant-wide totals but to narrow the scope to the cluster level. First, align cluster boundaries and names, and compare them taking into account installed capacity and solar irradiation conditions. Then, from generation differences by time of day, distinguish shading, soiling, and orientation differences, and confirm electrical imbalances from voltage, current, and string differences. Additionally, search for external factors during on-site inspections, and by recording the investigation findings and changes after countermeasures, you can reduce overlooked causes and confusion in the event of recurrence.

In cluster-level investigations, it is important not to attribute a cause based on a single piece of data. A decrease in power generation can result from a combination of factors such as weather, season, shading, dirt, equipment capacity, electrical abnormalities, measurement errors, and site conditions. By combining total generation, time-of-day data, voltage and current, onsite photos, and inspection records, you can realistically narrow down the causes of low generation. In particular, keeping multiple clusters comparable under the same criteria makes it easier to distinguish normal differences from abnormal ones.

In responding to a decline in power generation, avoiding mistakes in the order of checks is more important than quickly pinning down the cause. Viewed at the cluster level, even a large power plant can have its inspection scope narrowed down in stages, improving the efficiency of on-site verification and maintenance decisions. If you want to continuously monitor declines in power generation and link inspection records with field information, it is effective to promote condition management and the use of on-site records by cluster. To clearly organize the plant’s status and connect understanding of the causes of decline to improvement decisions, it is important to establish a system that can manage data, drawings, site photographs, and inspection histories using the same unit.