Five thresholds for daily variability and anomalies to watch when power generation is low

When you feel that a solar photovoltaic system's power generation is low, the first point of uncertainty is often the boundary between "is this natural variation due to the weather?" and "is there an abnormality in the equipment?" Daily generation varies with many factors: clear skies, cloudy skies, rain, temperature, shading, output curtailment, maintenance shutdowns, communication status, and so on. Therefore, judging an anomaly based on the figure for a single day can lead to unnecessary inspections or incorrect root-cause investigations. Conversely, if you dismiss it as natural variation, you may miss signs such as panel soiling, racking tilt, poor connections, equipment outages, or grid-side constraints.

In this article, aimed at practitioners who search for "low power generation", we organize the boundaries between normal variability and abnormalities in daily data from five perspectives. As an initial check before specialized analysis, we explain from a practical viewpoint which figures to line up, which changes to suspect, and at what stage to escalate to on-site inspection or maintenance actions.

Table of Contents

• Boundary 1: Check whether the day-to-day decline can be explained by weather differences.

• Boundary 2: Check whether low days are isolated or consecutive.

• Boundary 3: Check whether there is a bias in drop-off patterns by time of day

• Boundary 4: See whether the gap is widening when compared with equipment within the facility or with nearby equipment

• Boundary 5: Check whether the records, site conditions, and equipment status align with the numbers

• Summary: For day-to-day variability, determine the comparison method and don't overlook anomalies

Boundary 1: Check whether the daily decline can be explained by weather differences

When you find a day with low generation, the first thing to check is whether that drop can be explained by weather differences. Because solar power generation is strongly affected by solar irradiance, daily generation can vary greatly between clear, cloudy, and rainy days. In particular, on days when clouds move quickly, days that were sunny only in the morning but turned cloudy in the afternoon, or days with prolonged thin cloud cover, perceived weather and actual generation may not match. On site, queries like "It looked sunny both yesterday and today, so why is the generation different?" are common, but in fact daily generation fluctuates due to differences in irradiance strength, cloud cover, temperature, and solar elevation.

At this stage, the important thing is not to determine an anomaly solely from the absolute value of generation. For example, the fact that generation is lower than the previous day alone does not necessarily indicate an equipment fault. When making comparisons, you need to look at similar-weather days, historical data from the same month, the same period in the previous year, neighboring facilities, or metrics such as generation divided by solar irradiance. For sites that record solar irradiance, check whether daily generation and solar irradiance generally move in tandem. If irradiance has decreased and generation has also decreased, it may be a natural variation. Conversely, if solar irradiance is similar but generation is clearly lower, that becomes a starting point for suspecting equipment or operational factors.

The influence of ambient temperature should not be overlooked. Solar panels are easier to generate power the stronger the sunlight, but their output tends to drop when panel temperature rises. Therefore, a clear midsummer day is not necessarily the day of maximum annual generation. On clear spring or autumn days with sufficient sunlight and temperatures that are not too high, generation can be higher. Thus, even if generation on a hot sunny day does not increase as much as expected, it is safer not to judge it abnormal on that basis alone. However, if the decline is conspicuous compared with past data under similar temperatures and weather conditions, other factors may be at play.

Also, on the day after rain or during periods with a lot of yellow sand, pollen, or dust, you may see changes that cannot be explained by weather alone. In some cases light dirt on the panel surface is washed away by rain and generation recovers, while in others splashed mud, bird damage, fallen leaves, or accumulated deposits remain and only some strings continue to produce low output. Even if daily data shows only a slight drop in total generation, part of the system may be affected. When comparing weather differences, don’t think only in terms of sunny versus rainy—consider the period before and after the rain, wind strength, dust from nearby work, and the condition of nearby trees and structures, as keeping these in mind makes it easier to follow up with the next inspection.

The boundary between normal variability and an anomaly is where you don’t stop at “it was low because the weather was bad.” Power output is lower on bad-weather days, but if the drop is larger than that observed on other days with the same weather conditions or than at other installations, an anomaly may still be present. Conversely, if changes in weather and power output move together naturally and recover from the next day onward, there is no need to hastily conclude a failure. When judging day-to-day variability, it is important to first separate the portion that can be explained by weather and then examine the residual that cannot be explained.

Boundary 2: Determine whether low days are isolated or consecutive

What you should check next is whether a day with low generation is an isolated incident or persists over multiple days. Daily generation naturally fluctuates, so a single low day does not necessarily indicate equipment failure. Generation can drop temporarily due to cloudy or rainy weather, output curtailment, maintenance or inspections, power outages, or communication loss. The important thing is to confirm whether the drop recovers on the following day(s) or whether the same trend continues for several days.

Even a single isolated drop should not be completely overlooked. For example, if there are records of inspections or power outages on that day, a day when communications were temporarily interrupted, or a day with clearly bad weather, it becomes easier to explain the reason for the drop. However, if the records show normal operation and the weather was not significantly bad, yet one day shows a large drop, it may indicate a momentary shutdown or a fault in some equipment. Because daily data alone cannot determine the cause, it is practical to proceed to check that day's hourly data and equipment-specific data.

If power output is low for consecutive periods, the likelihood of an anomaly increases. In particular, attention is required when output does not recover even though the weather has improved, or when levels remain lower than before for several days to several weeks. Suspect factors that continuously affect performance, such as accumulation of dirt, shading from weeds or trees, partial panel failures, problems at connection points, operational limits of the power conditioner, changes in communication settings, or faults on the measurement side. In daily data, focus on patterns of decline: a sudden drop that does not recover, a gradual decrease, or low values only on specific days of the week or under certain conditions.

If it suddenly falls and does not recover, it is possible that the operating state or site conditions changed as of a certain day. Check for changes around the equipment such as construction work, inspections, setting changes, equipment replacement, cable work, installation of nearby structures, growth or removal of vegetation, and movement of fences or temporary installations. If the daily power generation graph shows a step-like decline, cross-check whether any site events occurred on the dates immediately before or after that drop to help narrow down possible causes.

If output gradually decreases, suspect slow-progressing causes such as aging, dirt, increased shading, changes in ventilation conditions, or degradation of contacts. However, during periods when seasonal changes alter solar radiation conditions, natural decline and abnormalities can appear to overlap. As winter approaches, changes in daylight hours and solar altitude mean that daily generation can naturally appear to decline. In such cases, compare with the same period in the previous year and with nearby installations to determine whether the decrease is large even after accounting for seasonal factors.

If output is low only on certain days of the week, or on specific operating days, operational factors rather than equipment faults may be involved. For self-consumption systems, apparent generation or recorded values can change depending on load conditions during holidays or closed days, battery charge/discharge settings, control conditions, and the operating patterns of on-site equipment. If you are looking only at the amount sold to the grid, the figures may reflect consumption or control effects rather than the generation itself. It is important to distinguish whether a decline in daily data represents the system’s generation performance or reflects differences in measurement items or operating conditions.

The boundary between normal variation and an anomaly is indicated by whether the decline recovers. If a one-off drop can be explained and returns to normal levels the next day or thereafter, it can likely be treated as natural variation or a temporary factor. Conversely, if an unexplained decline persists, power generation does not recover even when the weather improves, levels shift from a specific day onward, or the magnitude of the decline gradually widens, you should move to anomaly verification. For daily data, reading it as a trend over time rather than just looking at single-day figures improves the accuracy of your assessment.

Boundary 3: Check whether there is a bias in how the drop-off varies by time of day

When considering the causes of low daily generation, the day's total alone can hide important details. Even for the same decline in daily generation, the suspected causes differ depending on whether output is low only in the morning, low around midday, drops only in the afternoon, or is low all day. If daily data suggests a possible anomaly, it is important to next check the time-of-day generation curve. The generation curve provides practical clues for distinguishing normal variability from true abnormalities.

If only the morning power output is low, suspect eastern-side shading, morning dew, frost, soiling, delayed startup, or the influence of nearby structures. When the sun angle is low in the morning, shadows from buildings, trees, fences, slopes, and adjacent equipment tend to extend farther. Because shadow length changes with the seasons, there may be periods when only the morning output appears reduced. Especially from autumn into winter, the sun angle drops and shadows from obstacles that previously had no effect can reach the panels. Even if the daily total shows only a slight decrease, if the loss is concentrated in the morning hours, it is worth prioritizing a shading inspection.

If generation is low only in the afternoon, consider west-side shading, ambient temperature rise, equipment temperature rise, output limits, and grid conditions during specific time periods. In the afternoon, panel and equipment temperatures tend to be higher, and depending on the installation environment, output may not rise as much. If this is the normal temperature effect, similar trends often appear in nearby equipment and on past days with the same conditions. However, if only specific equipment drops significantly in the afternoon, if the drop is more pronounced on clear days, or if it coincides with equipment alarms or temperature-related records, it is necessary to check ventilation, the installation environment, and the equipment condition.

If it only dips in a valley-like way around late morning, the passing of clouds may have temporarily reduced irradiance, but if the dip occurs at the same time every day you should suspect other factors. For example, a particular structure casting a short-lived shadow when the sun is high, a control system activating under certain conditions, equipment being limited above a certain output, or temporary interruptions in measurement or communication. These are often treated as “slightly low days” in daily totals, but when viewed by hour the pattern of the cause can become clear.

If generation is low throughout the day, consider poor weather, an overall lack of solar irradiance, operational restrictions across the facility, measurement anomalies, main equipment shutdowns, or grid-side constraints. If it is low all day due to cloudy or rainy conditions, it may fall within the range of natural variation, but if it is low all day under conditions close to clear skies, caution is required. In particular, when the shape of the power generation curve is similar to normal but overall lower, check for soiling, age-related degradation, settings, measurement coefficients, and discrepancies with irradiance. If the shape of the power generation curve is distorted, suspect equipment stoppages or issues with the connection/system.

When looking at a power generation curve, check not only whether it is a smooth bell shape but also whether its shape changes when overlaid with a day that had the same weather. The normal curve on a sunny day rises in the morning, peaks around midday, and falls in the evening. However, the shape can vary depending on temperature, clouds, azimuth, tilt, shading, and control conditions. What matters is which time periods have changed compared with past normal periods. If only morning and evening are affected, suspect shading or solar altitude; if only midday is affected, suspect control or temperature; if it is affected all day, suspect overall/systemic factors. The time-of-day bias helps narrow down candidate causes.

The boundary between normal variation and an anomaly is whether the timing of the drop is reproducible. Drops caused by passing clouds tend to vary in timing from day to day, whereas drops due to shading, settings, equipment temperature, or operational controls can recur at similar times. When daily generation is low, don’t stop at the total value; by checking how the drop occurs by time of day, it becomes easier to distinguish natural variability from anomalies.

Boundary 4: Check whether differences are widening when comparing within the facility or with nearby facilities

When judging that power generation is low, it is very important to have a reference for comparison. Looking at generation output alone makes it hard to tell whether the cause is weather or seasonal effects, or an equipment-specific fault. What helps is comparing by circuit, by device, by area within the facility, or with neighboring facilities. By comparing targets operating in the same region, under similar weather, and at the same time of day, you can more easily identify differences that are difficult to explain by natural variation.

When a facility has multiple power conditioners or multiple circuits, check each one’s daily generation and the proportion of output. Even if total generation appears low, if all devices are similarly low then weather or other system-wide factors may be the cause. On the other hand, if only specific devices or specific circuits are underperforming, you can narrow the investigation to that area. Candidates for causes that affect only a portion of the system include shading over a particular area, panel soiling, poor connections, broken wiring, blown fuses, switches/circuit breakers, equipment shutdowns, differences in settings, and measurement faults.

When making comparisons, it's also important not to take capacity differences at face value. Larger-capacity equipment generates more electricity, and smaller-capacity equipment generates less. Therefore, rather than looking at simple total generation, examine generation per unit of capacity and the ratios between sections with the same design conditions. If sections have different orientations or tilts, their generation characteristics can differ even on the same day. East-facing sections produce more in the morning, west-facing sections produce more in the afternoon, and their daily output patterns may differ from sections that are closer to south-facing. If you judge without matching the conditions for comparison, you may mistakenly interpret normal design differences as anomalies.

Comparing with nearby installations is also useful. If generation across the same area on the same day is generally low, it is easier to conclude that weather or solar irradiance conditions are having a strong effect. Conversely, if nearby installations are at normal levels but only your system is low, you should suspect factors on your side. However, when comparing with nearby installations, there may be differences in system capacity, orientation, tilt, installation environment, shading, commissioning date, maintenance status, measurement methods, and output control settings. You cannot simply say "it's abnormal because it's lower than the neighbor," but such comparisons are useful as a way to see whether your system deviates from the common trend.

When comparing within a facility, what you should pay particular attention to is whether a difference is temporary or persistent. If clouds pass over locally or a shadow falls only briefly, only some sections may show a temporary drop. However, if the same section is lower every day, if the gap doesn't narrow even on sunny days, or if the difference widens over time, the likelihood of an anomaly increases. Instead of just listing daily generation, looking at the trend in ratios makes it easier to detect changes in the difference.

Also, localized abnormalities within a facility may not be obvious in the overall power generation. In a large installation, if only some circuits decline, the facility’s daily total generation may appear to drop by only a few percent. However, those parts may be experiencing a clear stoppage or reduction. Checking device- and circuit-level variations regularly—not only when you feel the overall generation is low—leads to earlier detection. In particular, if you routinely know each device’s normal share of output, you will more easily notice differences during an abnormal condition.

The boundary between normal variation and an anomaly lies in whether a subject deviates from others under the same conditions. If the entire system is similarly low, first suspect the weather or overall control. If only a specific range is low, consider that the cause may lie within that range. When the gap has widened compared with neighboring facilities or other sections within the facility, it is a sign that you should proceed to anomaly investigation rather than attribute it to day-to-day variation. Comparison is the basic method to narrow down the causes of reduced power generation with numbers rather than by intuition.

Boundary 5: Check whether the records, site conditions, and equipment status correspond to the numbers

When daily power output is low, there is a limit to judging based solely on the numbers. Changes in power output involve a complex mix of factors such as weather, on-site conditions, equipment condition, operational logs, inspection history, and communication status. To determine the boundary of an anomaly, it is essential to verify whether the daily data connects with on-site events. If changes in the numbers align with on-site records, the explanatory power for the cause increases. Conversely, if the numbers have changed significantly but there is no recorded reason, it is necessary to look for overlooked factors.

First, you should check the history of inspections and work. On days when cleaning, mowing, equipment inspections, work during power outages, cable checks, equipment replacements, setting changes, or work on communication equipment took place, daily power generation can be affected. If there were periods of temporary shutdown due to the work, that day's generation will be lower. If generation recovered after the work, the cause of the drop is easier to explain. On the other hand, if generation is lower after the work, you need to check settings, connections, missed restorations, and possible measurement mix-ups.

Next, we look at changes in site conditions. Examples include weeds growing, temporary structures being placed nearby, construction starting nearby, increased soil dust, noticeable bird damage, more fallen leaves, remaining snow or frost, and mud splashing caused by poor drainage—changes at the site are reflected in daily power generation. These causes are hard to discern from daily data alone, but become easier to judge when combined with site photos and inspection records. In particular, shadows and soiling change their impact depending on the time of day and season, so even at the same location their effect on power generation can vary from day to day.

Keeping records of equipment status is also important. If there are alarms, shutdowns, restarts, temperature rises, communication losses, operational restrictions, grid abnormalities, insulation-related warnings, or the like, check whether those times or dates coincide with decreases in power generation. The absence of an alarm does not necessarily mean everything is normal, but if alarms or operational logs are available they can provide clues to narrow down potential causes. Conversely, if generation is low while equipment-side records appear normal and the discrepancy cannot be explained by weather differences, also check measurement methods, communication losses, data aggregation, and possible confusion between sold electricity and generated electricity.

Be careful about differences in what is being measured. Even if generation appears low, the meaning changes depending on whether the number you are looking at is the energy generated, the amount sold, the surplus after self-consumption, the equipment-side output, or an aggregated value on the monitoring screen. In self-consumption systems, it may not be that generation itself is low; consumption patterns and battery behavior can make the amount sold look small. If you think you are looking at generation but are actually looking at the amount sold, you may mistakenly interpret a normal condition as a decline when there is no equipment fault. Before assessing daily variations, it is important to confirm which meter or which aggregated value you are viewing.

Also, communication failures or delays in data aggregation can appear as a decrease in power generation. If monitoring data are partially missing, daily totals may be displayed lower even though generation actually occurred. In some cases, data are backfilled after communication is restored, while in others the missing data remain. When generation suddenly drops to near zero, or records are missing only during specific time periods, you need to distinguish between equipment shutdown and communication loss. Cross-checking with on-site meter readings and device-side logs can reduce incorrect judgments.

The boundary between normal variation and an anomaly is whether there is a plausible explanation for the change in the numbers. If records such as weather, inspections, shutdowns, output control, site conditions, and communication status correspond to a decline in daily power generation, it should initially be treated as an explainable fluctuation. However, if the records and the numbers do not match, or if an unexplained decline continues, you should proceed to on-site verification and a detailed investigation. Daily data should not be judged on its own; only when combined with records does it become information usable in practice.

Summary: Day-to-day variability — decide how to compare so you don't miss anomalies

The purpose of looking at daily data when generation is low is not simply to find low days. It is to distinguish between natural variability and true anomalies, and to move quickly to any necessary checks. Solar power generation output varies day to day due to weather, season, temperature, shading, soiling, operational conditions, and measurement conditions. Therefore, rather than judging an anomaly from a single day’s figure, it is important to view the data from multiple perspectives: weather differences, continuity, time-of-day patterns, comparisons within the plant, and on-site records.

First, confirm whether the difference can be explained by weather. On days with low solar irradiance, rainy or cloudy days, or days with high temperatures, power generation may be lower. However, if output is lower even compared with days with similar weather, or if the patterns of solar irradiance and power generation do not match, equipment-related causes are suspected. Next, check whether the low-output days are isolated or consecutive. If isolated and the reason is clear, it may be a temporary factor; but if output remains low even after the weather returns, you should proceed to check for anomalies.

Furthermore, by examining time-of-day patterns you can narrow down possible causes. If it is low only in the morning, suspect shadows on the east side or startup conditions; if it is low only in the afternoon, suspect shadows on the west side or temperature; if there is a dip only at midday, suspect control issues or local shading; if it is low all day, suspect system-wide factors. Abnormalities that are not visible in daily totals become apparent when viewed by time of day. Comparison within the facility and with neighboring facilities is also indispensable. By distinguishing whether the whole system is similarly low or only specific equipment or zones are low, you can narrow the scope of possible causes.

Finally, confirm that the numbers correspond with the records. By looking together at information that can affect daily power generation—inspections, shutdowns, communication losses, on-site work, changes in vegetation, soiling, alarms, differences in measurement items, etc.—you can improve the accuracy of your assessments. A low level of power generation does not necessarily indicate a failure. However, if an unexplained decline continues, if it falls outside the comparison baseline under the same conditions, or if the decline is biased toward particular times of day or areas of equipment, prompt verification is necessary.

In practice, rather than investigating causes from scratch each time, it is effective to establish a procedure for reviewing daily data. First check the weather and solar radiation conditions, then examine continuity, review time-of-day data, compare within the facility and with neighboring sites, and finally reconcile with on-site records; creating this flow stabilizes assessments. If each person views data differently, anomalies are easily missed or overreacted to. It is important to standardize within the company the items to check, the periods for comparison, the conditions that constitute an anomaly, and the criteria for proceeding to on-site verification.

Daily variations, when interpreted correctly, can help detect abnormalities early. However, if interpretation remains vague, you can be led astray by natural weather-driven fluctuations or miss small faults on the equipment side. Especially when you feel that power generation is low, it is important to make judgments based on comparisons and records rather than on intuition. Organizing daily generation data together with inspection records and on-site information, and establishing a system that can detect signs of decline early, leads to stable equipment operation.