6 steps to find the causes of low power generation using string-level data

When you feel that the power output from a solar power system is "low", the first thing to check is not just the plant's total output. Looking only at the overall figure makes it difficult to determine whether the cause is weather or a fault in part of the equipment. An effective approach is to narrow down the cause of reduced output using string-level data. By comparing current, voltage, and generation trends for each string, it becomes easier to determine whether the decline is system-wide or confined to specific circuits. In this article, we explain six steps to investigate causes of low output from string-level data, so that field personnel can more easily prioritize on-site checks.

Table of Contents

• Before reviewing per-string data, outline the range of power output reduction.

• As the first step, isolate the effects of solar irradiance and weather

• As Step 2, compare the current differences between strings.

• As Step 3, suspect an open circuit or a poor connection from variations in voltage.

• As Step 4, check the decline patterns by time of day.

• As Step 5, overlay site photos and inspection records to narrow down the cause.

• Step 6: Verify the effectiveness of the improvements using post-recovery data.

• Approach to Utilizing Per-String Data for Daily Management

• To quickly detect low power generation

Clarify the range of power generation decline before looking at per-string data

When you feel that power generation is low, immediately chasing detailed numbers at the string level can make diagnosing the cause more complicated. First, it is important to determine whether the decline is occurring across the entire plant, concentrated around specific power conditioners, or biased toward some of the strings within them. Even when overall generation is low, not all strings will necessarily decline in the same way. On days with low irradiance the output will fall across the board, but if certain strings remain low even on sunny days, you should suspect equipment-related causes.

A string is a unit of several solar modules connected in series. In a power plant, multiple strings are gathered and, via combiner boxes and collection equipment, connected to power conversion equipment. Therefore, string-level data serve as a granular diagnostic unit for viewing the entire plant. Small abnormalities that are hard to detect from the overall power output alone can appear as differences when comparing strings individually. In particular, comparing strings with the same orientation, the same tilt, and the same number of modules makes it easier to determine whether an anomaly exists.

However, when looking at string-level data, it is important not to make simple comparisons while ignoring differences in design conditions. If azimuth and tilt, number of modules, shading, cable length, or the configuration of connected equipment differ, generated power and current values can vary even when operating normally. For example, a string facing east and a string facing west will have different peak production times. Higher output from the east-facing string in the morning and lower output from the west-facing string does not necessarily indicate an anomaly. Conversely, if one string among a group under the same conditions is clearly lower, the likelihood of a fault is high.

When investigating the cause of low power generation, it is important to establish a baseline comparison group first. Treat strings that share the same combiner box, the same orientation, identical installation conditions, and the same equipment configuration as a single group. By checking for differences within that group, it becomes easier to distinguish natural variations caused by weather and design conditions from anomalies. While string-level data can provide detailed information, viewing it without organizing the underlying assumptions can lead to misjudgments; therefore, it is advisable to first review the plant configuration diagram, the single-line wiring diagram, the string layout diagram, and past inspection records.

Also, it's important not to judge that generation is low based on the impression from a single day. Generation can drop due to short-term external factors such as cloudy skies, rain, snowfall, yellow dust, fog, or dust from nearby construction. The purpose of looking at per-string data is not simply to find locations with low numbers, but to check whether the decrease is persistent, whether it is reproducible under specific conditions, and whether there is an anomalous difference compared with other strings. When dealing with decreased generation in practice, by sequentially comparing overall values, solar irradiance, weather conditions, per-string data, and on-site conditions, you can narrow down the cause without forcing conclusions.

Step 1: Separate the effects of solar irradiance and weather

The first step in investigating causes with string-level data is to separate the effects of solar irradiance and weather. Even on days with low power generation, if the solar irradiance itself is low it does not necessarily indicate an equipment abnormality. Because photovoltaic generation is strongly influenced by irradiance, output can fluctuate greatly on cloudy or rainy days, on days with rapidly moving clouds, and during the low-irradiance periods in the morning and evening. Before checking string-level data, you should review that day’s weather and the trend in solar irradiance to determine whether the drop in generation is within a natural range.

The basic check is to see whether power output and solar irradiance move in tandem. If irradiance drops and the current of all strings decreases similarly, it is likely due to weather. Conversely, if irradiance is sufficient but only some strings show low current, suspect equipment-related causes. If irradiance has recovered but a particular string does not, that string should be prioritized for inspection.

When separating the effects of weather, it is useful not only to look at clear-sky days but also to find periods when solar irradiance remained stable. Even if the whole day is not completely sunny, if there are periods with little cloud influence you can compare per-string data for those periods. On days when clouds are moving, each string’s output can change in a short time, so it is risky to judge an anomaly from instantaneous values alone. Observing trends over a span of several minutes to tens of minutes and confirming whether a low condition persists will reduce false detections.

Also, when investigating low power output, comparing with data from the previous day, the previous week, or the same season in past years can be helpful. If, compared with days with similar weather conditions, only a specific string’s output has dropped, possibilities include age-related degradation, soiling, shading, poor connections, or equipment malfunction. On the other hand, if the entire plant’s output is similar to past days with the same conditions, the current drop may not be a significant anomaly. If you conclude a string fault without looking at the relationship between irradiance and power output, unnecessary site visits and inspection work can increase.

Snow, frost, yellow sand, pollen, bird droppings, fallen leaves, and similar can also block solar irradiance. These are external, weather-related factors, but in per-string data they can appear as partial decreases. For example, if snow remains on part of a roof or on some rows of racking, differences can occur between strings within the same plant. In such cases, it should be treated as a surface shading or soiling issue rather than an electrical fault. Considering not only solar irradiance but also the local weather and seasonal factors around the site makes it easier to classify the cause.

At this stage, it is important to categorize occurrences of low power output into 'natural decreases due to insufficient solar irradiance', 'temporary decreases due to external factors', and 'decreases suspected to indicate equipment faults.' Per-string data is useful for detecting equipment faults, but you cannot make a correct assessment if you ignore irradiance conditions. First, confirm the effects of solar irradiance and weather, select time periods or dates suitable for comparison, and then examine the differences between strings in detail using the following steps.

Step 2: Compare the current differences between strings

After isolating the effects of solar irradiance and weather to some extent, next compare the differences in current between strings. When looking for a decline in power output using string-level data, differences in current are particularly important. Solar modules generate current when exposed to sunlight. Therefore, strings installed under the same conditions tend to have similar current values during periods when they receive roughly the same irradiance. If only one string shows a lower current, it is a cue to suspect soiling, shading, module faults, connection abnormalities, or the operation of bypass diodes.

When comparing, first group strings under the same conditions. As a basic rule, compare those with similar installation orientation, tilt, number of modules, connection destination, and surrounding shading conditions. Mixing strings with different conditions can lead you to misidentify legitimate differences as abnormalities. Within the same group, check for any strings with persistently low current, and prioritize investigating those with the largest deviations from the average.

When looking at current differences, it's important to examine the time-series behavior, not just instantaneous values. Cloud cover or temporary shading can cause the current to drop briefly. A strong indication of an anomaly is when, during stable sunny periods, only a specific string remains continuously low. Also, if the drop occurs only at the same time each day, shadows from nearby structures, trees, rows of racking, utility poles, or fences may be the cause. If the current is low all day, more consideration should be given to equipment-side factors such as soiling, module faults, poor connections, or circuit problems.

Even if a string’s current is lower than others, there’s no need to immediately assume a serious fault. For example, even partial soiling on module surfaces can affect the entire string’s current. If bird droppings, fallen leaves, dust, or mud splashes are adhering to some modules, they can impact power generation more than they appear. Shadows from vegetation growth may not be a problem at installation, but as the seasons progress they will show up as differences in per-string data.

On the other hand, if the current of the same string suddenly drops significantly and does not recover, the priority for on-site inspection should be raised. Loose connections, poor connector contact, cable damage, operation of protective devices, or a partial malfunction of a module may be involved. If the current is close to zero, conditions such as the string being disconnected from the power circuit, a break in the wiring, or a protective device having tripped should also be considered. However, because actual response requires safety verification, do not decide on work based solely on remote data; it is important to confirm in accordance with the inspection procedures for electrical equipment.

An effective use of comparing current differences is to rank candidate anomalies. If you inspect all strings equally when power generation is low, the scope of work becomes too large. By using per-string data, you can prioritize checks on locations with noticeable declines, places that have changed significantly compared to the past, and locations with large deviations from the average under the same conditions. This improves on-site inspection efficiency and makes it easier to shorten the time to identify the cause.

Step 3: Suspect a broken wire or poor connection from voltage fluctuations

In per-string data, not only current but also voltage is an important criterion. Current is easily affected by solar irradiance and shading, whereas voltage is influenced by string configuration, connection condition, the number of modules, and temperature. When investigating the causes of low power output, if you only look at differences in current you can miss signs of poor connections or open circuits. By also checking voltage variations, it becomes easier to narrow down causes that are more likely electrical in nature.

In normal conditions, the voltages of strings with the same configuration generally fall within a similar range. Of course, they vary with module temperature and operating conditions, but if only one deviates significantly under the same conditions, attention is required. If the voltage is extremely low, faults such as poor connections within the string, a broken conductor, tripped protective devices, or a partial module failure may be suspected. If voltage cannot be obtained, or the value fluctuates unstably, checking the measurement system and connection points may also be necessary.

If both current and voltage are low, the entire string may not be functioning adequately as a power-generating circuit. If current is low while voltage is relatively maintained, issues affecting the irradiated surface—such as shading, dirt, or surface blockage of the module—may be involved. If only the voltage is abnormally high or low, it is necessary to check the operation control and connection status. In practice, it is important to interpret current and voltage together rather than looking at them separately.

Also, when comparing voltages, ambient and module temperatures should be taken into account. In general, higher module temperatures tend to lower the voltage, while lower temperatures tend to raise it. Therefore, it is not appropriate to conclude that a voltage is abnormal simply because it is lower than yesterday's, while ignoring seasonal or time-of-day variations. If you are going to make comparisons, the basic practice is to compare strings with the same installation conditions at the same time on the same day. Even when comparing with past data, choosing days with similar solar irradiance and ambient temperature makes assessment easier.

If a break or poor connection is suspected, it can be difficult to confirm this from remote data alone. String-level data indicate candidates for anomalies, and a safe on-site inspection is required for a final determination. Electrical equipment inspections include checking for open-circuit conditions, the condition of connection points, cable damage, loose terminals, the condition of protective devices, and indications inside junction boxes. However, because inspecting energized equipment involves hazards, it is a prerequisite that the person in charge has the appropriate authority and that safety procedures are followed.

The purpose of checking voltage variations when power output is low is to clarify the direction of the inspection. Depending on whether shading or soiling is strongly suspected, or an electrical connection abnormality is more likely, the locations to inspect on site will differ. By looking at voltage as well as current, it becomes easier to carry out a targeted inspection by narrowing down checkpoints in advance instead of searching blindly after arriving on site.

Step 4: Check time-of-day decline patterns

When investigating the causes of low power generation in string-level data, it is very important to look at time-of-day patterns of decline. Even for a drop in the same string, the suspected causes change depending on how it appears—low only in the morning, only at midday, only in the evening, low all day, or low only during certain seasons. If you only look at the daily total energy generation, you can easily miss these characteristics.

If the power output of a specific string is low only in the morning, shadows from the east, nearby buildings, trees, terrain, or shadows cast by rows of mounting racks may be involved. If it is low only in the evening, check for shadows from the west or the long shadows that occur at low solar altitude. If it decreases around midday, review shadows from above, soiling, partial shading of the module surface, and the placement of equipment. The cause of low power output is not necessarily singular, and because shadow positions change with the time of day, examining the shape of the data can help identify the cause.

A string that is low all day suggests not only shading but also soiling buildup, module faults, poor connections, cable damage, or tripping of protective devices. In particular, if the same string is lower than others from morning to evening even on sunny days, on-site inspection should be a high priority. By contrast, if the drop is brief and corresponds with changes in solar irradiance, it may be due to clouds or temporary irradiance fluctuations. Viewing by time of day makes it easier to distinguish external factors from equipment-related causes.

Seasonal changes should not be overlooked. In winter, the sun's altitude is lower, so shadows that were not a problem in summer can fall on the strings. From spring to summer, vegetation grows and shadows from surrounding plants can affect power generation. In autumn, fallen leaves, and in winter, snow or frost can also have an impact. If low power output occurs only in specific seasons, you need to consider seasonal factors as well as equipment faults.

When checking time-of-day decline patterns, it's easier to understand if you pay attention to the shape of the graph. A healthy string shows a smooth, gentle hill-shaped power generation trend on sunny days when solar irradiance is stable. By contrast, a shaded string will show dips at specific times of day, or delayed ramp-up or ramp-down. If there's a poor connection or equipment fault, it may appear as a sudden drop, irregular fluctuations, or a sustained near-zero value.

This procedure clarifies "when it occurs" for anomalies. During on-site inspections, being able to check at the time of day when a drop occurs makes it easier to detect shadows or soiling. For example, if you go to inspect a drop that occurs in the morning at noon, it can be difficult to determine the cause of the shadow. If you know from the data the time when the drop occurs, it is easier to adjust the timing of on-site photos and inspections. To efficiently find the cause of low power output, reading the flow of time as well as the numbers is indispensable.

Step 5: Narrow down the cause by overlaying on-site photos with inspection records

Once potential causes of a decline become apparent from string-level data, the next step is to narrow down the cause by cross-referencing on-site photos and inspection records. Remotely obtained data is extremely useful, but numbers alone cannot fully capture shadows, dirt, damage, vegetation, snow accumulation, installation condition, or changes in the surrounding environment. To determine the reasons for low power generation in practice, it is necessary to combine the data with on-site information.

First, what I want to confirm is the on-site location of the underperforming string. Match the string layout diagram with the data to verify which row, which section, which roof surface, and which junction box it belongs to. If this remains unclear, there is a risk of inspecting the wrong location on-site. Confirm that the string number, junction box number, power conditioner circuit, and module layout correspond, and clarify the position of the suspected anomaly.

Next, compare them with on-site photos. In the photos, check for module surface dirt, bird droppings, fallen leaves, shadows from vegetation, shadows from the racking and surrounding structures, panel cracks, tilting, misalignment, sagging wiring, the condition around junction boxes, and so on. By confirming whether there are visible abnormalities at the locations of strings with low power output, you can narrow down the possible causes. If past photos are available, you can compare them to see whether vegetation has grown, whether new structures have been built nearby, or whether module surface soiling has increased.

Inspection records are also important. Check whether the same system has experienced abnormalities in the past, whether there is a history of cleaning or repairs, whether components have been replaced, and whether conditions have changed after strong winds, typhoons, heavy rain, or snowfall. If a string with low power output has repeatedly declined in the past, the root cause may not have been resolved. If it falls again after once recovering, consider loose connections or a recurrence of environmental factors.

When overlaying on-site photos with data, pay attention to the photo’s date and time. If the time period of low power generation differs significantly from the photo’s shooting time, you may not be able to correctly assess shadow conditions. If you photograph a spot that is shaded in the morning at midday, the photo may not reveal the problem. If possible, using photos taken close to the time period when the drop appears in the data makes judgment easier. For checking shadows in particular, it is important to match the season and time of day.

Also, during on-site checks it is important to record even small anomalies. Even if the cause cannot be identified on the spot, if you document photos, location information, the time of inspection, weather, surrounding conditions, and work performed, you can later cross-reference them with the data. The cause of reduced power generation may not be discernible from a single inspection. By compiling multiple sets of data and on-site records, the repeatability of the decline and seasonal patterns will become apparent.

String-level data is the entry point for improving the accuracy of on-site inspections. Identify potential anomalies from the numerical data, verify the situation with photos and records, and proceed to electrical inspections if necessary. By following this order, you can reduce unnecessary checks and more easily pinpoint the causes of reduced power generation.

Step 6: Confirm the improvement effects using post-recovery data

After performing responses to candidate causes—such as cleaning, tree cutting, repairs, checking connections, replacing components, and verifying settings—always check the post-restoration string-level data. Even if you take some action to address low power generation, if no improvement is confirmed in the data, the cause may lie elsewhere. In practice, it is important not to focus on the fact that an action was taken, but on how the differences in power output and current changed after the action.

When confirming recovery, make sure to compare like-for-like before and after the intervention. Simply comparing a sunny day with a cloudy day makes it difficult to discern the improvement effect. If possible, check on days with similar solar irradiance conditions, at the same time of day, and between strings in the same comparison group. Look to see whether the current of the string that had been reduced has returned to levels similar to the surrounding strings, whether the time-of-day dips have been eliminated, and whether the tendency to be low throughout the day has improved.

For example, if dirt or fallen leaves were the cause, the current difference is expected to narrow after cleaning. If shading from vegetation was the cause, cutting back or weeding may improve the drop that occurs during specific times of day. If a poor connection was the cause, repairs may eliminate unstable fluctuations or sudden drops. However, if improvement is only partial, multiple causes may have overlapped. Because the reason for low power generation is not necessarily a single factor, the need for additional checks is determined by reviewing the data after restoration.

After restoration, it is important to check both the short term and the medium term. Even if things improve immediately after the work, they can decline again after a few days or weeks. For example, a problem at a connection may temporarily improve but recur due to vibration or temperature changes. Shadows from vegetation can begin to affect things again as the seasons progress. Dirt after cleaning can also reattach in a short time depending on the surrounding environment. Therefore, don’t stop with only an immediate post-action check; monitoring the trend over a certain period provides reassurance.

The results of recovery verification are kept as inspection records. We record which string experienced what kind of decline, what was checked, what actions were taken, and how the data changed after those actions. As these records accumulate, when a low power output condition occurs again, diagnosing the cause becomes faster. If a similar decline occurred at the same location in the past, the possibility of recurrence can be considered immediately. Conversely, if the pattern differs from past cases, it provides an opportunity to suspect a new cause.

Addressing a drop in power generation doesn't end with finding the cause. It also involves confirming the effectiveness of improvements, monitoring for recurrence, and keeping records that can be used for future decisions. Using string-level data to verify restoration makes it easier to justify the response and helps improve the quality of power plant management.

How to Apply Per-String Data to Daily Management

Per-string data is not something to be checked only when generation is low. By keeping track of trends on a daily basis, you can detect small abnormalities earlier. If you don't know the normal data in advance, it becomes difficult to judge how abnormal a drop is when it occurs. Knowing the typical current on sunny days, seasonal generation trends, time-of-day characteristics by orientation, and the normal variation between strings will speed up decision-making when an anomaly occurs.

In daily operations, it's also important not to treat reference values as too rigid. Solar power generation data change with the seasons, solar irradiance, temperature, clouds, soiling, and the surrounding environment. Therefore, rather than simply labeling a value below a fixed threshold as abnormal, it is more practical to look at the difference from the average of strings under the same conditions, the difference compared with past days under the same conditions, and the persistence of any decline. String-level data are advantageous not only for absolute values but also for relative comparisons.

Also, to detect low power generation quickly, it is useful to organize rules for anomaly detection. For example, deciding which conditions to check—such as when only a particular string is persistently low within the same group; when the difference from the average during daytime on sunny days is large; when current suddenly drops and does not recover; when voltage is abnormal compared with surrounding units; or when a system that previously had an anomaly is declining again—can reduce variation in judgments among operators.

However, you should avoid judging everything on site based solely on remote data. String-level data is a powerful resource for narrowing down causes, but to confirm the actual condition of equipment you need on-site photos, inspection records, and safe electrical inspections. Especially when electrical abnormalities are suspected, work procedures and safety management must be thoroughly enforced. Establishing a workflow of narrowing candidates with data, confirming them on site, and evaluating with data after restoration leads to stable operations.

To make string-level data useful for daily management, it is also important to have a system that links the data to the field. Just looking at a string number is not enough—if you cannot tell where it is on site, inspections take time. If layout diagrams, photos, inspection history, and generation data are correlated, potential anomalies can be immediately identified on a map or as specific field locations. What matters for operational staff is not just reading the numbers, but being able to turn those numbers into on-site actions.

To quickly detect low power generation

When investigating the causes of low power generation using string-level data, first determine whether the decline is system-wide or localized, and separate the effects of irradiance and weather. Then compare current differences among strings under the same conditions and check for possible open circuits or poor connections from voltage variations. Additionally, examining decline patterns by time of day makes it easier to diagnose shading, soiling, seasonal factors, or equipment abnormalities. Finally, narrow down the cause by cross-referencing site photos and inspection records, and confirm the effectiveness of measures with post-action data.

A state of low power generation, if left unchecked, will accumulate losses. In particular, a decline in some strings can be overlooked if you only look at the overall plant figures. Even small drops, if they persist for a long time, cannot be ignored in terms of their impact on power generation. That is why it is important to use string-level data to detect potential anomalies early and lead to on-site verification.

On the other hand, string-level data, being more granular, tends to create a larger amount of information to review. In practice, you need to do more than just look at the data: you must clearly define what to check, in what order to inspect the site, and what will be considered evidence of improvement after taking action. By handling power generation, irradiance, string current, voltage, time-of-day trends, site photos, and inspection history together, you can more efficiently identify the causes of reduced power generation.

In power plant management, rather than scrambling to investigate after an anomaly occurs, it is important to routinely standardize how data is viewed and how records are kept. By understanding per-string data under normal conditions, keeping comparison baselines for performance drops, and linking these to on-site location information and photos, you can more quickly detect low generation states. Establishing such routine management procedures reduces the chance of overlooking causes and makes it easier to prioritize inspections and restoration actions.