An operation and maintenance scheduling method for a distributed photovoltaic cluster

Abstract

The present application relates to photovoltaic power generation technical field, specifically relates to a kind of operation and maintenance scheduling method for distributed photovoltaic cluster.Get photovoltaic unit operation time sequence and the regional environment data it is in, calculate the daily light transmittance time series data that can reflect its surface light transmittance dynamic change, while obtaining the predicted rainfall probability of each day in future period;Analysis light transmittance time series data, determine the daily light transmittance observation value, analyze the same day observation value difference and determine light transmittance deviation index, reflect the difference between the unit and dust accumulation, then analyze fluctuation characteristics and calculate dust accumulation confidence, accurately reflect the degree of dust accumulation and development trend;For the problem that dust accumulation signal is weak and easy to be submerged by noise, based on a variety of indexes to construct cost function, determine the optimal light transmittance prediction sequence, reflect the real clean degree of unit;Finally, combined with the optimal sequence change trend, numerical characteristics and rainfall probability, determine the total expected dust accumulation loss in future period, according to which the photovoltaic unit is operated and maintained, and cleaning is reasonably arranged.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power generation technology, and specifically to an operation and maintenance scheduling method for distributed photovoltaic clusters. Background Technology

[0002] With the increasing global demand for clean energy, distributed photovoltaic (PV) power generation, as an important method of renewable energy utilization, has been widely applied and rapidly developed. During the operation of distributed PV clusters, the light transmittance of the PV units is one of the key factors affecting their power generation efficiency. The accumulation of dust and dirt on the surface of the PV panels significantly reduces light transmittance, thereby reducing the power generation of the PV units, thus requiring maintenance and cleaning.

[0003] Currently, the operation and maintenance of distributed photovoltaic (PV) clusters mainly relies on periodic inspections and cleaning. However, these regular inspections and cleaning cannot be precisely tailored to the actual dust accumulation on the PV units, potentially leading to over- or under-cleaning of some units. Furthermore, rainfall naturally removes dust from the PV units. Therefore, cleaning according to a fixed schedule undoubtedly increases unnecessary operation and maintenance costs in areas with abundant rainfall; while in areas with scarce rainfall, failure to clean in a timely manner will exacerbate the dust accumulation problem, resulting in a significant decrease in power generation. Summary of the Invention

[0004] To address the technical problem that performing cleaning at fixed intervals would undoubtedly increase unnecessary operation and maintenance costs in areas with high rainfall, while in areas with low rainfall, failure to clean in a timely manner would exacerbate the dust accumulation problem, leading to a significant decrease in power generation, this invention aims to provide an operation and maintenance scheduling method for distributed photovoltaic clusters. The specific technical solution adopted is as follows: The system acquires the runtime sequence data of each photovoltaic unit and the environmental data of the area where it is located, in order to calculate the daily transmittance time series data of each photovoltaic unit; and acquires the predicted rainfall probability for each day in the future period for the area where the photovoltaic unit is located. In each transmittance time series data, the numerical characteristics of transmittance are analyzed to determine the daily transmittance observation value of each photovoltaic unit; the differences between the transmittance observation values ​​of photovoltaic units on the same day are analyzed to determine the transmittance deviation index of each photovoltaic unit on each day; based on the fluctuation characteristics of the transmittance deviation index corresponding to each photovoltaic unit, the daily dust accumulation confidence level of each photovoltaic unit is obtained. The numerical characteristics of the light transmittance observations of each photovoltaic unit and their deviation from the pre-assigned light transmittance estimates are analyzed. A cost function is constructed by combining the dust accumulation confidence and light transmittance deviation index to determine the optimal light transmittance prediction sequence for each photovoltaic unit. In the optimal transmittance prediction sequence for each photovoltaic unit, based on the changing trend and numerical characteristics of the data values, and combined with the daily rainfall probability and preset economic parameters in the future period of the region, the total expected dust accumulation loss in the future period is determined and used for operation and maintenance scheduling of each photovoltaic unit.

[0005] Furthermore, the method for obtaining the transmittance observation value includes: In the daily transmittance time-series data of each photovoltaic unit, all transmittance values ​​are sorted in ascending order to obtain the sorted sequence; In the sorting sequence, the transmittance at the preset percentile is used as the daily transmittance observation value for each photovoltaic unit.

[0006] Furthermore, the method for obtaining the transmittance deviation index includes: On the same day, the transmittance observations of all photovoltaic units were sorted in descending order to obtain the sorting sequence; In the arrangement sequence, the average of the first preset number of transmittance observations is used as a reference benchmark value; The difference between the reference benchmark value and the observed transmittance value of each photovoltaic unit is used as the transmittance deviation index of each photovoltaic unit on a daily basis.

[0007] Furthermore, the method for obtaining the confidence level of the ash accumulation includes: For any photovoltaic (PV) unit, determine the preset time neighborhood corresponding to the PV unit for each day. Within the preset time neighborhood for each day, perform negative correlation mapping on the variance of the transmittance deviation index of all PV units and normalize the value, which is used as the confidence level of dust accumulation of the PV unit for each day.

[0008] Furthermore, the method for obtaining the cost function includes: The cost function includes boundary residuals, regional residuals, and trend residuals, and the environmental data includes daily rainfall data for the area where the photovoltaic unit is located. Multiple transmittance estimates are set, and a transmittance estimate is assigned to each photovoltaic unit every day, thereby exhaustively obtaining all transmittance estimate sequences for each photovoltaic unit; For any photovoltaic unit, under each transmittance estimation sequence, if the transmittance observation value of the photovoltaic unit on a certain day is greater than or equal to the preset cloudy and rainy threshold, then the transmittance observation value of that day is set as a valid observation value; otherwise, the transmittance observation value of that day is set as an invalid observation value. Select any photovoltaic unit as the unit under test, and perform the following under each transmittance estimation sequence; Based on the deviation between the observed and estimated transmittance values ​​of the unit under test each day, and in conjunction with the confidence level of dust accumulation, the boundary residuals of the unit under test each day are determined. The numerical characteristics of the estimated and observed transmittance values ​​of the unit under test are analyzed daily, and combined with the confidence level of dust accumulation and the transmittance deviation to obtain the regional residuals of the unit under test daily. If the rainfall data of the unit under test on a certain day is greater than or equal to the preset cleaning threshold, the trend residual of the unit under test on that day will be set to 0. If the rainfall data of the unit under test on a certain day is less than the preset cleaning threshold: if the estimated transmittance of the unit under test on that day is greater than the estimated transmittance of the previous day, the trend residual of the unit under test on that day is set to infinity; if the estimated transmittance of the unit under test on that day is less than or equal to the estimated transmittance of the previous day, the absolute value of the difference between the estimated transmittance of the unit under test on that day and the estimated transmittance of the previous day is calculated as the dust accumulation factor; if the dust accumulation factor is greater than the preset dust accumulation threshold, the trend residual of the unit under test on that day is set to infinity; if the dust accumulation factor is less than or equal to the preset dust accumulation threshold, the dust accumulation factor of the unit under test on that day is set to the trend residual. Under each transmittance estimation sequence, the sum of the boundary residuals, regional residuals, and trend residuals of the unit under test for each day is used as the cost factor of the unit under test for each day. The sum of the cost factors of the unit under test over all days is used as the value of the cost function of the unit under test for each transmittance estimation sequence.

[0009] Furthermore, the method for obtaining the boundary residual includes: If the transmittance observation value of the unit under test on a certain day is an invalid observation value, then the boundary residual of the unit under test on that day is set to 0; otherwise, the formula model for the boundary residual includes: in, This represents the boundary residual of the unit under test on day d; This indicates the confidence level of dust accumulation in the unit under test on day d. This represents the observed transmittance value of the unit under test on day d. This represents the estimated transmittance of the unit under test on day d. Indicates the heavy penalty coefficient; This indicates the light penalty coefficient.

[0010] Furthermore, the method for obtaining the regional residual includes: If the transmittance observation value of the unit under test on a certain day is an invalid observation value, then the regional residual of the unit under test on that day is set to 0; otherwise, the formula model for the regional residual includes: in, This represents the regional residual of the unit under test on the d-th day; This indicates the confidence level of dust accumulation in the unit under test on day d. This represents the estimated transmittance of the unit under test on day d. This indicates the deviation of the transmittance of the unit under test from the index on day d. This represents the spatial constraint coefficient.

[0011] Furthermore, the method for obtaining the optimal transmittance prediction sequence includes: Among all transmittance estimation sequences for each photovoltaic unit, the transmittance estimation sequence corresponding to the minimum value of the cost function is taken as the optimal transmittance prediction sequence for each photovoltaic unit.

[0012] Furthermore, the method for obtaining the total expected ash accumulation loss includes: The preset economic parameters include the preset electricity price and the preset daily power generation. In the optimal transmittance prediction sequence for each photovoltaic unit, the last transmittance estimate is negatively correlated and mapped to the value, which is then used as the current ash loss rate for each photovoltaic unit. In the optimal transmittance prediction sequence for each photovoltaic unit, the least squares method is used to fit all transmittance estimates to a straight line, and the slope of the fitted line is used as the average daily deposition rate. In the future, the probability of no rainfall will be obtained by negatively mapping the predicted rainfall probability for each photovoltaic unit for each day. In the future period, for any photovoltaic unit, the product of all the probabilities of no rainfall for that photovoltaic unit on each day and before each day is taken as the no rainfall index for that photovoltaic unit on each day in the future period. In the future period, the product of the daily sequence number of the photovoltaic unit and the average daily deposition rate of the photovoltaic unit will be used as the daily ash accumulation growth factor of the photovoltaic unit. The sum of the ash accumulation growth factor and the current ash loss rate of the photovoltaic unit will be used as the daily ash accumulation index of the photovoltaic unit in the future period. The product of the daily dust accumulation index, no rainfall index, preset electricity price, and preset daily power generation of the photovoltaic unit in the future period is used as the expected loss level of the photovoltaic unit in the future period. In the future period, the sum of the expected loss values ​​of the photovoltaic unit on each day will be taken as the total expected ash accumulation loss of the photovoltaic unit in the future period.

[0013] Furthermore, the operation and maintenance scheduling for each photovoltaic unit includes: The difference between the total expected dust accumulation loss of each photovoltaic unit in the future period and the preset cost of a single cleaning is taken as the cleaning revenue; If the cleaning benefit is less than or equal to 0, then no maintenance cleaning will be performed; if the cleaning benefit is greater than 0, then maintenance cleaning will be performed.

[0014] The present invention has the following beneficial effects: Daily transmittance time-series data is calculated by acquiring operational sequence data for each photovoltaic (PV) unit and environmental data of its location. This transmittance time-series data reflects the dynamic changes in the light transmittance performance of the PV unit surface over time, representing data calculated at the physical level. Simultaneously, the predicted daily rainfall probability for the area where the PV unit is located is also acquired, fully considering the significant impact of weather factors on PV unit operation and maintenance. Since cloud cover affects PV transmittance as a random negative pulse (i.e., a significant decrease in transmittance), and the transmittance of the module in unshaded conditions reflects its true physical upper limit, the numerical characteristics of transmittance are analyzed in each transmittance time-series data set to determine the daily transmittance observation value for each PV unit. Furthermore, the differences in transmittance observation values ​​between PV units on the same day are analyzed to determine the daily transmittance deviation index for each PV unit, reflecting the differences in dust accumulation between units. Based on this, the fluctuation characteristics are analyzed to calculate the dust accumulation confidence level, which more accurately reflects the degree and trend of dust accumulation. Furthermore, addressing the issue of weak dust accumulation signals that are easily masked by noise, a cost function can be constructed based on the aforementioned indicators to determine the optimal transmittance prediction sequence. This sequence reflects the true cleanliness of each photovoltaic unit under physical constraints and other conditions each day. Finally, within the optimal transmittance prediction sequence for each photovoltaic unit, based on the changing trends and numerical characteristics of the data values, and combined with data such as rainfall probability, the total expected dust accumulation loss for future periods is determined for operation and maintenance scheduling of each photovoltaic unit. This precise operation and maintenance scheduling method can effectively improve the overall power generation efficiency of the photovoltaic cluster, reduce operation and maintenance costs, and maximize economic benefits. Attached Figure Description

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating an operation and maintenance scheduling method for a distributed photovoltaic cluster, provided as an embodiment of the present invention. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an operation and maintenance scheduling method for distributed photovoltaic clusters proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0019] The following description, in conjunction with the accompanying drawings, details a specific scheme for the operation and maintenance scheduling method of a distributed photovoltaic cluster provided by the present invention.

[0020] Please see Figure 1 The diagram illustrates a method flowchart for operation and maintenance scheduling of a distributed photovoltaic cluster according to an embodiment of the present invention. The method includes the following steps: Step S1: Obtain the operating sequence data of each photovoltaic unit and the environmental data of the area where it is located, in order to calculate the daily transmittance time series data of each photovoltaic unit; obtain the predicted rainfall probability for each day in the future period of the area where the photovoltaic unit is located.

[0021] Because the electrical sampling frequency of photovoltaic inverters is typically on the order of minutes (e.g., 5 or 15 minutes), while the release frequency of public meteorological data is typically on the order of hours, there is a misalignment between the two in the time dimension. Furthermore, when the solar altitude angle is too low, the spectral redshift effect caused by atmospheric refraction will cause the photoelectric conversion efficiency to deviate from the linear model; when the inverter reaches its rated power, the peak clipping protection mechanism will lead to output power saturation distortion. Therefore, to ensure the data quality and model applicability of subsequent calculations, time-domain alignment and data filtering based on physical characteristics must be performed.

[0022] For each photovoltaic (PV) unit in the PV cluster, the operating sequence data of each PV unit and the environmental data of its location are acquired. In this embodiment of the invention, the operating sequence data includes the measured active power on the AC side of the PV unit, and the environmental data mainly includes two types: ambient temperature time-series data and rainfall data, where the rainfall data is in daily units. Then, using the sampling time of the active power time-series data as a reference, linear interpolation is performed on the low-frequency ambient temperature time-series data to calculate the synchronous ambient temperature at the sampling time of the active power time-series data. Next, using an astronomical algorithm (such as the SPA algorithm), the latitude and longitude of the PV unit and the time (the time should be synonymously converted to Greenwich Mean Time) are input to calculate the solar altitude angle at each time. Then, an effective dataset is constructed: if the solar altitude angle at a certain time is less than a preset altitude angle threshold (in the low-light nonlinear region, the preset altitude angle threshold is taken as...), the solar altitude angle is calculated as follows: If the measured active power is greater than or equal to 98% of the rated installed capacity (limited power saturation zone), the data point is marked as invalid and removed. The rated installed capacity can be read in advance from the equipment ledger.

[0023] Since the physical essence of photovoltaic power generation is to convert received solar radiation energy into electrical energy, and this conversion efficiency decreases as the cell temperature rises, directly using measured active power cannot distinguish whether the power decrease is caused by reduced sunlight (such as cloud cover) or by dust accumulation on the module surface. Therefore, by constructing a theoretical output model under ideal clean conditions, a standardized reference denominator can be provided for the measured active power, thereby decoupling environmental factors. For each valid sampling moment, the system performs the following physical modeling calculations: The first step is to calculate the slope irradiance. The total horizontal irradiance is calculated using a clear-sky irradiance model (e.g., the Ineichen-Perez model or the Haurwitz model). The atmospheric turbidity coefficient required by the Ineichen-Perez model is set to a default value of 3.0, or can be obtained from meteorological data of the area where the photovoltaic (PV) units are located. Then, combining the installation azimuth and tilt angle of each PV unit, an anisotropic model (e.g., the Hay model or the Perez model) is used to convert the horizontal irradiance into the effective received irradiance on the slope of the modules. The installation azimuth and tilt angle required for this step are obtained from the static configuration ledger of the PV units.

[0024] The second step is to estimate the battery operating temperature. Using the Ross thermal balance model or other equivalent thermal models, the junction temperature of the module's solar cells is calculated based on the ambient temperature and the slope irradiance. ;in, Let be the junction temperature at the t-th valid sampling time. Let be the synchronous ambient temperature at the t-th valid sampling time; Let be the slope irradiance at the t-th effective sampling time; This refers to the empirical temperature rise coefficient related to its installation structure; for example, for roof-mounted installations, due to poor ventilation and heat dissipation, a value of [value missing] is recommended. For bracket-supported, overhead installations, the recommended value is [value to be inserted here]. .

[0025] The third step is to calculate the theoretical clear-sky baseline. Based on the standard test conditions (STC) of photovoltaic modules, a temperature correction factor is introduced to calculate the theoretical maximum output at that moment. : ;in, For standard test irradiance ( ), The standard test temperature is (°C). The power temperature coefficient of photovoltaic modules (usually a negative value, such as...) ), The nominal conversion efficiency of the inverter (e.g.) ); This represents the theoretical maximum output of the photovoltaic unit at time t. Indicates the rated installed capacity of the photovoltaic unit; This represents the synchronous ambient temperature at the t-th valid sampling time. Let be the slope irradiance at the t-th effective sampling time.

[0026] because This characterizes the upper limit of the physical output of photovoltaic modules under the current illumination geometry and temperature, while the measured active power... It is the final product after atmospheric attenuation, cloud cover, and dust accumulation on the component surface. Therefore, by calculating the ratio of the two, we can eliminate the two main environmental variables of irradiance and temperature, and obtain a normalized index that only reflects the light path transmittance.

[0027] For each valid sampling time Calculate the normalized transmittance : in, This represents the transmittance of the photovoltaic unit at the t-th effective sampling time. This represents the measured active power of the photovoltaic unit at the t-th effective sampling time. This represents the theoretical maximum output of the photovoltaic unit at time t; a small positive number is added to the denominator. (Recommended) This is to prevent the theoretical calculation value from being inaccurate. In extremely small cases (such as the instants of sunrise and sunset), numerical calculation precision issues can lead to a denominator of zero or an infinitely large calculation error. An upper limit threshold is introduced. (Recommended) Due to the cloud edge effect, the instantaneous measured irradiance may be higher than the theoretical clear-sky irradiance, or there may be slight deviations in the theoretical model parameters, which could cause the calculated ratio to be slightly greater than the theoretical ratio. Therefore, through Truncation of functions can prevent individual abnormally high values ​​from causing excessive interference to subsequent statistical characteristics (such as daily peak values).

[0028] At the end of the day, the system will record all valid sampling times for that day. The calculated transmittance is arranged in chronological order to generate daily transmittance time-series data for each photovoltaic unit. This time-series data intuitively reflects the dynamic changes in the transmittance performance of the photovoltaic module surface over time. High-frequency, large downward fluctuations are usually caused by cloud shading, while long-term trend-like attenuation corresponds to dust accumulation.

[0029] Furthermore, since rainfall can naturally remove dust from photovoltaic units, the impact of rainfall needs to be considered in subsequent analyses. Therefore, the predicted rainfall probability for each photovoltaic unit's area during the future period was also obtained, which can be obtained through daily forecasts.

[0030] It should be noted that in this embodiment of the present invention, the number of days for the light transmittance time series data collection is set to 15 days, and the number of days for the future period is set to 10 days. The specific number of days can be adjusted according to the implementation scenario and is not limited here. The clear sky irradiance model and other related technologies used in this embodiment of the present invention are all technologies well known to those skilled in the art, and the specific process will not be described in detail here.

[0031] Step S2: In each transmittance time series data, analyze the numerical characteristics of transmittance to determine the daily transmittance observation value of each photovoltaic unit; analyze the differences between the transmittance observation values ​​of photovoltaic units on the same day to determine the transmittance deviation index of each photovoltaic unit on each day; based on the fluctuation characteristics of the transmittance deviation index corresponding to each photovoltaic unit, obtain the daily dust accumulation confidence level of each photovoltaic unit.

[0032] Because the true level of dust accumulation in photovoltaic modules only becomes apparent when clouds clear, the transmittance calculated in step S1, while eliminating the influence of light temperature, is still affected by cloud cover. Furthermore, clouds only decrease transmittance (negative noise), not increase it. Therefore, the impact of cloud cover on photovoltaic transmittance manifests as random negative pulses (i.e., a significant drop in transmittance), while the transmittance of the photovoltaic unit in unobstructed conditions reflects its true physical upper limit. Therefore, by first analyzing the daily transmittance time-series data for each photovoltaic unit, we determine the daily transmittance observation value for each unit to eliminate cloud noise and pinpoint the optimal performance for each day.

[0033] Preferably, in one embodiment of the present invention, the method for obtaining the transmittance observation value includes: In the daily transmittance time-series data of each photovoltaic unit, all transmittance values ​​are sorted in ascending order to obtain a sorted sequence. Since cloud cover may cause most of the low values ​​in the daily transmittance time-series data to be cloud-contaminated data, which cannot represent the true transmittance status of the components, the transmittance at the preset percentile is used as the daily transmittance observation value of each photovoltaic unit in the sorted sequence. In this embodiment of the invention, the preset percentile is set to the 95th percentile, which can filter out the low values ​​caused by cloud cover, retain the numerical characteristics close to the true transmittance capability, and avoid positive errors introduced by measurement glitch or instantaneous over-irradiation.

[0034] Since photovoltaic units in the same area are under the same atmospheric radiation background, the systematic deviation caused by the inaccurate estimation of atmospheric aerosol parameters in the theoretical clear sky model manifests as a common-mode signal in the same direction among the photovoltaic units in the cluster. Therefore, by analyzing the differences in the transmittance observation values ​​of photovoltaic units in the area on the same day, the transmittance deviation index of each photovoltaic unit on each day can be determined. This can offset the common-mode meteorological error and the model systematic error, and only retain the signal caused by the difference in dust accumulation.

[0035] Preferably, in one embodiment of the present invention, the method for obtaining the transmittance deviation index includes: On the same day, the transmittance observations of all photovoltaic (PV) units are sorted in descending order to obtain a sorting sequence. The average of the first preset number of transmittance observations in this sequence is used as a reference benchmark value, reflecting the transmittance of the best-performing PV unit. In this embodiment of the invention, the preset number is set to 20% of the total number; the specific value can be adjusted according to the implementation scenario and is not limited here.

[0036] Then, the difference between the reference value and the observed transmittance value of each photovoltaic unit is calculated. A positive and larger difference indicates a greater negative deviation between the photovoltaic unit and the reference value, suggesting severe dust accumulation. Conversely, a negative and smaller difference indicates a cleaner photovoltaic unit than the reference value. Therefore, the obtained difference is directly used as the daily transmittance deviation index for each photovoltaic unit. Based on the aforementioned analysis, a smaller transmittance deviation index indicates a cleaner photovoltaic unit, while a larger value indicates more severe dust accumulation.

[0037] In theory, the transmittance deviation index of each photovoltaic unit on each day reflects the difference in dust accumulation between photovoltaic units and eliminates the meteorological error of common mode. Furthermore, given that the relative performance degradation caused by dust accumulation is stable over several consecutive days, while the relative deviation caused by local cloud clusters has random fluctuations, we can analyze the fluctuation characteristics of the transmittance deviation index of each photovoltaic unit over multiple days to determine the confidence level of dust accumulation for each photovoltaic unit on each day.

[0038] Preferably, in one embodiment of the present invention, the method for obtaining the confidence level of ash accumulation includes: For any given photovoltaic (PV) unit, a preset time neighborhood is determined for each day. In this embodiment, the preset time neighborhood is set to each day and the two days closest to its time sequence. Since dust accumulation is a slow variable that typically doesn't change abruptly, while cloud cover and localized shadows are fast variables that change rapidly, the variance of the transmittance deviation index for all PV units is calculated within the preset time neighborhood for each day. The smaller the variance, the more stable the transmittance deviation index, and the more likely it is caused by dust accumulation, resulting in higher confidence. Conversely, the larger the variance, the more chaotic the transmittance deviation index, and the more likely it is noise interference, leading to lower confidence. Therefore, the obtained variance is negatively correlated and normalized to correct the logical relationship, thereby obtaining the confidence level of dust accumulation for the PV unit each day. The negative correlation mapping and normalization process can be performed using the formula... ,in, Let x represent an exponential function with the natural constant e as the base, and let x represent the independent variable.

[0039] Step S3: Analyze the numerical characteristics of the light transmittance observations of each photovoltaic unit and the deviation between them and the pre-assigned light transmittance estimates. Combine the dust accumulation confidence level and the light transmittance deviation index to construct a cost function, thereby determining the optimal light transmittance prediction sequence for each photovoltaic unit.

[0040] Since the power attenuation caused by photovoltaic dust accumulation is a weak, slow variable, while the power fluctuations caused by cloud cover and changes in atmospheric turbidity are drastic, fast variables, traditional single-unit time-series analysis is prone to misinterpreting environmental noise as dust accumulation signals. However, the actual dust accumulation state is essentially a smooth variable. Therefore, this step introduces the irreversibility (monotonically cumulative) of the dust accumulation process and the abrupt reset characteristics of rainwater cleaning to pre-allocate transmittance estimates for each day and analyze the deviation between these estimates and observed transmittance values. Combining the dust accumulation confidence level and transmittance deviation index, a cost function is constructed to determine the optimal transmittance prediction sequence for each photovoltaic unit. The aim is to use dynamic programming algorithms to decouple the actual dust accumulation state of each photovoltaic unit under strong noise background and achieve accurate quantification of even the smallest dust accumulation signals.

[0041] First, a cost function is constructed. Preferably, in one embodiment of the present invention, the method for obtaining the cost function includes: Multiple transmittance estimates are set, and one transmittance estimate is assigned to each photovoltaic unit per day, thereby exhaustively obtaining all transmittance estimate sequences for each photovoltaic unit. This step can transform the problem of solving a continuous function into a discrete path search problem. In this embodiment of the invention, since the transmittance observation value is between 0 and 1, the transmittance estimate value is also between 0 and 1. A discretization step size of 0.001 is set to divide 0 to 1 into many discrete points. At this time, each discrete point is a transmittance estimate.

[0042] Then, a data validity label is generated: for any photovoltaic unit, under each transmittance estimation sequence, if the transmittance observation value of the photovoltaic unit on a certain day is greater than or equal to the preset cloudy / rainy threshold, then the transmittance observation value of that day is set as a valid observation value; otherwise, the transmittance observation value of that day is set as an invalid observation value.

[0043] The cost function includes boundary residuals, regional residuals, and trend residuals. Among them, the boundary residuals mainly restrict the final predicted value to roughly match the actual power generation performance of the photovoltaic unit itself and cannot be separated from the observation data. Therefore, it can be determined based on the deviation between the observed value and the estimated value of the transmittance of the photovoltaic unit and in combination with the confidence level of dust accumulation.

[0044] For ease of subsequent explanation and illustration, a photovoltaic unit is selected as the unit under test, and the method for obtaining the boundary residuals under each transmittance estimation sequence includes: If the transmittance observation value of the unit under test on a certain day is an invalid observation value, then the boundary residual of the unit under test on that day is set to 0; otherwise, the formula model for the boundary residual includes: in, This represents the boundary residual of the unit under test on day d; This indicates the confidence level of dust accumulation in the unit under test on day d. This represents the observed transmittance value of the unit under test on day d. This represents the estimated transmittance of the unit under test on day d. Indicates the heavy penalty coefficient; This indicates the light penalty coefficient.

[0045] In the above formula model, a higher confidence level of dust accumulation on a given day indicates more reliable data among photovoltaic units within the region. Therefore, the weight of the boundary residual should be reduced to decrease the reliability of unilateral data. Conversely, a lower confidence level of dust accumulation indicates less reliable data, thus increasing the weight of the boundary residual. Therefore, a negative correlation mapping is applied to the dust accumulation confidence level in the boundary residual formula model. Because cloud cover can only lower, not raise, the situation is considered highly unreasonable when the observed transmittance value is higher than the estimated transmittance value; therefore, a heavy penalty coefficient is applied. (Set to 100) the boundary residual cost increases sharply; conversely, when the transmitted light observation value is less than or equal to the transmitted light estimate value, it is considered to be a normal situation, so a light penalty coefficient (set to 1) is applied to keep the boundary residual cost within the normal range.

[0046] Next, since the transmittance observations of a single unit may still contain the influence of thin clouds that are difficult to detect, it is necessary to consider the general characteristics among photovoltaic units in a certain area for further constraints. Therefore, the numerical characteristics of the transmittance estimates and transmittance observations of the unit under test are analyzed every day, and combined with the dust accumulation confidence level and transmittance deviation, the regional residuals of the unit under test are obtained every day: Similarly, if the transmittance observation value of the unit under test on a certain day is an invalid observation value, then the regional residual of the unit under test on that day is set to 0; otherwise, the formula model for the regional residual includes: in, This represents the regional residual of the unit under test on the d-th day; This indicates the confidence level of dust accumulation in the unit under test on day d. This represents the estimated transmittance of the unit under test on day d. This indicates the deviation of the transmittance of the unit under test from the index on day d. This represents the spatial constraint coefficient.

[0047] In the above formula model, the dust accumulation confidence level reflects the stability of the transmittance deviation index. A larger value indicates higher data stability from the photovoltaic units within the region, meaning the data is more reliable and the calculation of the regional residuals is more reliable. Conversely, a smaller value indicates a lower reliability in the calculation of the regional residuals. The '1' in the formula represents the ideal cleanliness state, and the transmittance deviation index reflects the deviation of the tested unit from the cleanliness level within the region. A larger value indicates more severe dust accumulation on the photovoltaic unit, while a smaller value indicates a cleaner photovoltaic unit. This can reflect the current transmittance level that the unit under test should have. Then, the absolute value of the difference between the estimated transmittance value and the current transmittance level that the unit under test should have is calculated. The larger this value is, the larger the estimated residual is and the less ideal it is. Multiply it by the spatial constraint coefficient (set to 10), which is mainly used to balance the magnitude of the regional residual and the boundary residual.

[0048] Since rainfall has a cleaning effect on the dust accumulation of photovoltaic units, the rainfall situation is analyzed in this embodiment of the invention to further constrain the cost function.

[0049] Rainfall inputs external energy, cleans the photovoltaic modules, and breaks the original dust accumulation process. Therefore, if the rainfall data of the unit under test on a certain day is greater than or equal to the preset cleaning threshold, the trend residual of the unit under test on that day will be set to 0.

[0050] Conversely, if the rainfall data for a given day is less than the preset cleaning threshold: Under normal circumstances, without external rainfall or cleaning, dust will only accumulate, meaning light transmittance will only decrease. Therefore, if the estimated light transmittance of the tested unit on that day is greater than the estimated light transmittance of the previous day, it is considered illogical and impossible. Therefore, the trend residual for that day is set to infinity (which can be 10). 6 (This is done by substituting the previous day's data). When the estimated transmittance of the unit under test on that day is less than or equal to the estimated transmittance of the previous day, the absolute value of the difference between the estimated transmittance of the unit under test on that day and the estimated transmittance of the previous day can be calculated as the dust accumulation factor. However, dust accumulation is a natural deposition process and will not increase suddenly. Therefore, if the dust accumulation factor is greater than the preset dust accumulation threshold, it is also considered unreasonable. In this case, the trend residual of the unit under test on that day is set to infinity. Only when the dust accumulation factor is less than or equal to the preset dust accumulation threshold is it considered to be in accordance with normal conditions, and the dust accumulation factor of the unit under test on that day is set as the trend residual. In this embodiment of the invention, the preset cleaning threshold is essentially the rainfall, set to 5 mm, and the preset dust accumulation threshold is set to 0.02. Since the estimated transmittance of the first day is not compared with that of the previous day, it is only compared with 1 (ideal clean state).

[0051] Finally, for each transmittance estimation sequence, the sum of the boundary residuals, regional residuals, and trend residuals of the unit under test for each day is taken as the cost factor of the unit under test for each day. Based on the aforementioned analysis, it can be seen that the smaller the boundary residual, the closer the estimated value is to the measured data, which is consistent with the facts; the smaller the regional residual, the more consistent the estimated value is with the commonality; and the smaller the trend residual, the more consistent with the experiment. Therefore, the smaller the sum cost factor, the more reasonable the selection of transmittance estimates in the transmittance estimation sequence is. Thus, the sum of the cost factors of the unit under test for all days is taken as the value of the cost function of the unit under test for each transmittance estimation sequence.

[0052] It should be noted that the spatial constraint coefficient, heavy penalty coefficient, light penalty coefficient, etc. in the embodiments of the present invention can all be obtained by calibration training based on the historical data of the photovoltaic unit.

[0053] Thus, we can obtain the value of each transmittance estimation sequence for each photovoltaic unit under the cost function. Based on this value, we can determine the optimal and most reasonable transmittance estimation sequence for each photovoltaic unit and use it as the optimal transmittance prediction sequence.

[0054] Preferably, in one embodiment of the present invention, the method for obtaining the optimal transmittance prediction sequence includes: Based on the aforementioned logic, it can be seen that the smaller the value of the cost function, the better and more reasonable the estimate is. Therefore, among all the transmittance estimation sequences of each photovoltaic unit, the transmittance estimation sequence corresponding to the minimum value of the cost function is taken as the optimal transmittance prediction sequence for each photovoltaic unit.

[0055] Step S4: In the optimal transmittance prediction sequence of each photovoltaic unit, based on the changing trend and numerical characteristics of the data values, and combined with the daily rainfall probability and preset economic parameters in the future period of the region, determine the total expected dust accumulation loss in the future period, which is used for operation and maintenance scheduling of each photovoltaic unit.

[0056] The optimal transmittance prediction sequence for each photovoltaic unit can reflect the actual transmittance level of each photovoltaic unit on a daily basis. Therefore, in this step, the numerical characteristics and trends of the data can be analyzed based on the prediction sequence. Combined with the probability of each day in the future period and the preset economic parameters, the total expected ash accumulation loss in the future period can be determined for operation and maintenance scheduling of each photovoltaic unit.

[0057] Preferably, in one embodiment of the present invention, the total expected ash accumulation loss includes: In this embodiment of the invention, the preset economic parameters include a preset electricity price and a preset daily power generation. The preset electricity price can be obtained based on market pricing, and the preset daily power generation can be taken from the historical daily power generation on the same day of the same month last year.

[0058] Since light transmittance is typically negatively correlated with the loss caused by dust accumulation, in the optimal light transmittance prediction sequence for each photovoltaic (PV) unit, the last estimated light transmittance value, after negative correlation mapping, is used as the current dust loss rate for each PV unit. Furthermore, in the optimal light transmittance prediction sequence for each PV unit, the least squares method is used to fit a straight line to all estimated light transmittance values, and the slope of the resulting fitted line is used as the daily average deposition rate. The daily average deposition rate can be used to characterize the potential dust accumulation growth rate for each PV unit in the future. This negative correlation mapping can be achieved using... , where x represents the independent variable; the least squares method is a well-known technique, and the specific process will not be elaborated here.

[0059] In the future period, the predicted rainfall probability for each photovoltaic unit is negatively correlated and mapped to the value of no rainfall for each day. Since rainfall has a cleaning effect on dust, there is a positive correlation between the no rainfall probability and dust accumulation. Therefore, in the future period, for any photovoltaic unit, if no effective rainfall has occurred before a certain day, the degree of dust accumulation will accumulate. So, the product of all no rainfall probabilities for that photovoltaic unit on each day and before that day is used as the no rainfall index for that photovoltaic unit on each day in the future period.

[0060] Furthermore, since dust accumulates over time, the daily sequence number of the photovoltaic unit represents the time process. Multiplying this value by the average daily deposition rate yields the daily dust accumulation increase. Therefore, the product of the daily sequence number of the photovoltaic unit and its average daily deposition rate is used as the daily dust accumulation growth factor for the photovoltaic unit. The sum of the dust accumulation growth factor and the current dust loss rate of the photovoltaic unit is used as the daily dust accumulation index for the photovoltaic unit in the future.

[0061] The higher the dust accumulation index, the greater the degree of dust accumulation; the higher the no-rainfall index, the higher the likelihood of dust accumulation. The preset electricity price and preset power generation reflect the economic value that should exist each day. Therefore, the product of the dust accumulation index, no-rainfall index, preset electricity price, and preset daily power generation of the photovoltaic unit in the future period is taken as the expected loss level of the photovoltaic unit in the future period.

[0062] Finally, the sum of the expected daily loss values ​​of the photovoltaic unit in the future period will be used as the total expected dust accumulation loss of the photovoltaic unit in the future period.

[0063] This allows us to obtain the total expected dust accumulation loss for each photovoltaic unit in the future period. The larger the value, the greater the economic loss caused by dust accumulation. Then, we can use this indicator to schedule the operation and maintenance of the photovoltaic units.

[0064] Preferably, in one embodiment of the present invention, the operation and maintenance scheduling of each photovoltaic unit includes: The difference between the total expected dust accumulation loss of each photovoltaic unit in the future period and the preset cost of a single cleaning is calculated as the cleaning benefit. If the cleaning benefit is greater than 0, it means that the economic loss caused by the current dust accumulation is higher than the cost of a single cleaning, so cleaning can bring positive benefits. Conversely, if the cleaning benefit is less than or equal to 0, it means that the economic loss caused by the current dust accumulation is lower than or equal to the cost of a single cleaning, so cleaning will result in a loss. Therefore, if the cleaning benefit is less than or equal to 0, maintenance cleaning is not performed; if the cleaning benefit is greater than 0, maintenance cleaning is performed.

[0065] In summary, by acquiring the operational sequence data of each photovoltaic (PV) unit and the environmental data of its location, daily transmittance time-series data is calculated. This transmittance time-series data reflects the dynamic changes in the light transmittance performance of the PV unit surface over time, representing data calculated at the physical level. Simultaneously, the predicted daily rainfall probability for the area where the PV unit is located is also obtained, fully considering the significant impact of weather factors on PV unit operation and maintenance. Since cloud cover affects PV transmittance as a random negative pulse (i.e., a significant decrease in transmittance), and the transmittance of the module in unshaded conditions reflects its true physical upper limit, the numerical characteristics of transmittance are analyzed in each transmittance time-series data. This determines the daily transmittance observation value for each PV unit and analyzes the differences between the transmittance observation values ​​of PV units on the same day, identifying the daily transmittance deviation index for each PV unit. This index reflects the differences in dust accumulation between units, and based on this, the fluctuation characteristics are analyzed to calculate the dust accumulation confidence level, which more accurately reflects the degree and trend of dust accumulation. Furthermore, addressing the issue of weak dust accumulation signals that are easily masked by noise, a cost function can be constructed based on the aforementioned indicators to determine the optimal transmittance prediction sequence. This sequence reflects the true cleanliness of each photovoltaic unit under physical constraints and other conditions each day. Finally, within the optimal transmittance prediction sequence for each photovoltaic unit, based on the changing trends and numerical characteristics of the data values, and combined with data such as rainfall probability, the total expected dust accumulation loss for future periods is determined for operation and maintenance scheduling of each photovoltaic unit. This allows for the rational arrangement of cleaning time and frequency according to the dust accumulation loss of different units. This precise operation and maintenance scheduling method can effectively improve the overall power generation efficiency of the photovoltaic cluster, reduce operation and maintenance costs, and maximize economic benefits.

[0066] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0067] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for operation and maintenance scheduling of distributed photovoltaic clusters, characterized in that, The method includes: The system acquires the runtime sequence data of each photovoltaic unit and the environmental data of the area where it is located, in order to calculate the daily transmittance time series data of each photovoltaic unit; and acquires the predicted rainfall probability for each day in the future period for the area where the photovoltaic unit is located. In each transmittance time series data, the numerical characteristics of transmittance are analyzed to determine the daily transmittance observation value of each photovoltaic unit; the differences between the transmittance observation values ​​of photovoltaic units on the same day are analyzed to determine the transmittance deviation index of each photovoltaic unit on each day; based on the fluctuation characteristics of the transmittance deviation index corresponding to each photovoltaic unit, the daily dust accumulation confidence level of each photovoltaic unit is obtained. The numerical characteristics of the light transmittance observations of each photovoltaic unit and their deviation from the pre-assigned light transmittance estimates are analyzed. A cost function is constructed by combining the dust accumulation confidence and light transmittance deviation index to determine the optimal light transmittance prediction sequence for each photovoltaic unit. In the optimal transmittance prediction sequence for each photovoltaic unit, based on the changing trend and numerical characteristics of the data values, and combined with the daily rainfall probability and preset economic parameters in the future period of the region, the total expected dust accumulation loss in the future period is determined and used for operation and maintenance scheduling of each photovoltaic unit.

2. The operation and maintenance scheduling method for distributed photovoltaic clusters according to claim 1, characterized in that, The method for obtaining the transmittance observation value includes: In the daily transmittance time-series data of each photovoltaic unit, all transmittance values ​​are sorted in ascending order to obtain the sorted sequence; In the sorting sequence, the transmittance at the preset percentile is used as the daily transmittance observation value for each photovoltaic unit.

3. The operation and maintenance scheduling method for distributed photovoltaic clusters according to claim 1, characterized in that, The method for obtaining the transmittance deviation index includes: On the same day, the transmittance observations of all photovoltaic units were sorted in descending order to obtain the sorting sequence; In the arrangement sequence, the average of the first preset number of transmittance observations is used as a reference benchmark value; The difference between the reference benchmark value and the observed transmittance value of each photovoltaic unit is used as the transmittance deviation index of each photovoltaic unit on a daily basis.

4. The operation and maintenance scheduling method for distributed photovoltaic clusters according to claim 1, characterized in that, The method for obtaining the confidence level of the ash accumulation includes: For any photovoltaic (PV) unit, determine the preset time neighborhood corresponding to the PV unit for each day. Within the preset time neighborhood for each day, perform negative correlation mapping on the variance of the transmittance deviation index of all PV units and normalize the value, which is used as the confidence level of dust accumulation of the PV unit for each day.

5. The operation and maintenance scheduling method for distributed photovoltaic clusters according to claim 1, characterized in that, The method for obtaining the cost function includes: The cost function includes boundary residuals, regional residuals, and trend residuals, and the environmental data includes daily rainfall data for the area where the photovoltaic unit is located. Multiple transmittance estimates are set, and a transmittance estimate is assigned to each photovoltaic unit every day, thereby exhaustively obtaining all transmittance estimate sequences for each photovoltaic unit; For any photovoltaic unit, under each transmittance estimation sequence, if the transmittance observation value of the photovoltaic unit on a certain day is greater than or equal to the preset cloudy and rainy threshold, then the transmittance observation value of that day is set as a valid observation value; otherwise, the transmittance observation value of that day is set as an invalid observation value. Select any photovoltaic unit as the unit under test, and perform the following under each transmittance estimation sequence; Based on the deviation between the observed and estimated transmittance values ​​of the unit under test each day, and in conjunction with the confidence level of dust accumulation, the boundary residuals of the unit under test each day are determined. The numerical characteristics of the estimated and observed transmittance values ​​of the unit under test are analyzed daily, and combined with the confidence level of dust accumulation and the transmittance deviation to obtain the regional residuals of the unit under test daily. If the rainfall data of the unit under test on a certain day is greater than or equal to the preset cleaning threshold, the trend residual of the unit under test on that day will be set to 0. If the rainfall data of the unit under test on a certain day is less than the preset cleaning threshold: if the estimated transmittance of the unit under test on that day is greater than the estimated transmittance of the previous day, the trend residual of the unit under test on that day is set to infinity; if the estimated transmittance of the unit under test on that day is less than or equal to the estimated transmittance of the previous day, the absolute value of the difference between the estimated transmittance of the unit under test on that day and the estimated transmittance of the previous day is calculated as the dust accumulation factor; if the dust accumulation factor is greater than the preset dust accumulation threshold, the trend residual of the unit under test on that day is set to infinity; if the dust accumulation factor is less than or equal to the preset dust accumulation threshold, the dust accumulation factor of the unit under test on that day is set as the trend residual. Under each transmittance estimation sequence, the sum of the boundary residuals, regional residuals, and trend residuals of the unit under test for each day is used as the cost factor of the unit under test for each day. The sum of the cost factors of the unit under test over all days is used as the value of the cost function of the unit under test for each transmittance estimation sequence.

6. The operation and maintenance scheduling method for distributed photovoltaic clusters according to claim 5, characterized in that, The method for obtaining the boundary residual includes: If the transmittance observation value of the unit under test on a certain day is an invalid observation value, then the boundary residual of the unit under test on that day is set to 0; otherwise, the formula model for the boundary residual includes: in, This represents the boundary residual of the unit under test on day d; This indicates the confidence level of dust accumulation in the unit under test on day d. This represents the observed transmittance value of the unit under test on day d. This represents the estimated transmittance of the unit under test on day d. Indicates the heavy penalty coefficient; This indicates the light penalty coefficient.

7. The operation and maintenance scheduling method for distributed photovoltaic clusters according to claim 5, characterized in that, The method for obtaining the regional residual includes: If the transmittance observation value of the unit under test on a certain day is an invalid observation value, then the regional residual of the unit under test on that day is set to 0; otherwise, the formula model for the regional residual includes: in, This represents the regional residual of the unit under test on the d-th day; This indicates the confidence level of dust accumulation in the unit under test on day d. This represents the estimated transmittance of the unit under test on day d. This indicates the deviation of the transmittance of the unit under test from the index on day d. This represents the spatial constraint coefficient.

8. The operation and maintenance scheduling method for distributed photovoltaic clusters according to claim 5, characterized in that, The method for obtaining the optimal transmittance prediction sequence includes: Among all transmittance estimation sequences for each photovoltaic unit, the transmittance estimation sequence corresponding to the minimum value of the cost function is taken as the optimal transmittance prediction sequence for each photovoltaic unit.

9. The operation and maintenance scheduling method for distributed photovoltaic clusters according to claim 1, characterized in that, The method for obtaining the total expected ash accumulation loss includes: The preset economic parameters include the preset electricity price and the preset daily power generation. In the optimal transmittance prediction sequence for each photovoltaic unit, the last transmittance estimate is negatively correlated and mapped to the value, which is then used as the current ash loss rate for each photovoltaic unit. In the optimal transmittance prediction sequence for each photovoltaic unit, the least squares method is used to fit all transmittance estimates to a straight line, and the slope of the fitted line is used as the average daily deposition rate. In the future, the probability of no rainfall will be obtained by negatively mapping the predicted rainfall probability for each photovoltaic unit for each day. In the future period, for any photovoltaic unit, the product of all the probabilities of no rainfall for that photovoltaic unit on each day and before each day is taken as the no rainfall index for that photovoltaic unit on each day in the future period. In the future period, the product of the daily sequence number of the photovoltaic unit and the average daily deposition rate of the photovoltaic unit will be used as the daily ash accumulation growth factor of the photovoltaic unit. The sum of the ash accumulation growth factor and the current ash loss rate of the photovoltaic unit will be used as the daily ash accumulation index of the photovoltaic unit in the future period. The product of the daily dust accumulation index, no rainfall index, preset electricity price, and preset daily power generation of the photovoltaic unit in the future period is used as the expected loss level of the photovoltaic unit in the future period. In the future period, the sum of the expected loss values ​​of the photovoltaic unit on each day will be taken as the total expected ash accumulation loss of the photovoltaic unit in the future period.

10. The operation and maintenance scheduling method for distributed photovoltaic clusters according to claim 1, characterized in that, The operation and maintenance scheduling of each photovoltaic unit includes: The difference between the total expected dust accumulation loss of each photovoltaic unit in the future period and the preset cost of a single cleaning is taken as the cleaning revenue; If the cleaning benefit is less than or equal to 0, then no maintenance cleaning will be performed; if the cleaning benefit is greater than 0, then maintenance cleaning will be performed.

Images