An intelligent cleaning method, device, equipment and medium of a photovoltaic power station

Abstract

The application provides an intelligent cleaning method, device, equipment and medium of a photovoltaic power station, relates to the technical field of photovoltaic power station cleaning control, and comprises the following steps: acquiring environmental state data, operation data and cleaning resource data of the photovoltaic power station in a current time period; inputting the environmental state data into a pre-trained dust accumulation evolution prediction model to obtain a predicted dust accumulation evolution trend in a future preset time period; inputting the predicted dust accumulation evolution trend and the operation data into a pre-trained power generation loss prediction model to obtain a predicted power generation loss in the future preset time period; determining an optimal cleaning scheme based on the cleaning resource data and the predicted power generation loss, and cleaning the photovoltaic power station based on the optimal cleaning scheme. The application realizes the intelligentization, refinement and scientization of the cleaning decision of the photovoltaic power station, effectively improves the power generation efficiency, reduces the cleaning cost, reduces unnecessary cleaning operation, and improves the overall operation and maintenance benefit of the power station.

Description

Technical Field

[0001] This application relates to the field of photovoltaic power plant cleaning control technology, and more specifically, to an intelligent cleaning method, device, equipment, and medium for photovoltaic power plants. Background Technology

[0002] With the rapid development of the new energy industry, photovoltaic power plants, as an important carrier of clean energy, have seen continuous growth in installed capacity. However, in areas prone to sandstorms, dust, sand, and other pollutants easily accumulate on the surface of photovoltaic modules, a phenomenon known as "dust accumulation." Dust accumulation obstructs the light-receiving surface of photovoltaic modules, reducing light absorption efficiency and causing a decrease in module output power, severely impacting the power generation and economic benefits of photovoltaic power plants. Related data indicates that severe dust accumulation can cause a 20%-40% loss in photovoltaic module power generation. Therefore, cleaning photovoltaic modules is a crucial aspect of photovoltaic power plant operation and management.

[0003] Currently, the cleaning decision-making process for photovoltaic power plants mainly adopts a passive mode of monitoring dust accumulation and triggering cleaning. The core logic is to monitor the degree of dust accumulation on the surface of photovoltaic modules through sensors, and trigger a cleaning command when the amount of dust accumulation reaches a preset threshold. However, the above method has a strong lag, and the cleaning decision depends on the dust accumulation monitoring results that have already occurred. By the time excessive dust accumulation is detected, the photovoltaic modules have already suffered a certain period of power generation loss, and the greater the amount of dust accumulation, the higher the difficulty and cost of subsequent cleaning. Summary of the Invention

[0004] The purpose of this application is to provide an intelligent cleaning method, device, equipment, and medium for photovoltaic power plants, which solves the above-mentioned problems existing in the prior art. It can realize the intelligent, precise, and scientific decision-making of photovoltaic power plant cleaning, effectively improve power generation efficiency, reduce cleaning costs, reduce unnecessary cleaning operations, and improve the overall operation and maintenance efficiency of the power plant.

[0005] Firstly, a smart cleaning method for photovoltaic power plants is provided, which may include: Obtain environmental status data, operational data, and cleaning resource data of the photovoltaic power station for the current time period; The environmental state data is input into a pre-trained ash accumulation evolution prediction model to obtain the predicted ash accumulation evolution trend for a future preset time period. The predicted ash accumulation evolution trend and the operating data are input into a pre-trained power generation loss prediction model to obtain the predicted power generation loss for a future preset time period. Based on the cleaning resource data and the predicted power generation loss, an optimal cleaning scheme is determined, and the photovoltaic power station is cleaned based on the optimal cleaning scheme.

[0006] In an optional implementation, the environmental status data includes: environmental data of the environment in which the photovoltaic power station is located and dust accumulation characteristic data of the photovoltaic power station; The predicted ash accumulation evolution trend includes: the predicted ash accumulation thickness and the predicted ash accumulation coverage change trend.

[0007] In an optional implementation, the ash accumulation evolution prediction model employs a bidirectional LSTM neural network architecture; The power generation loss prediction model adopts a gradient boosting tree architecture.

[0008] In an optional implementation, based on the cleaning resource data and the predicted power generation loss, an optimal cleaning scheme is determined, and based on the optimal cleaning scheme, the photovoltaic power station is cleaned, including: If the predicted ash accumulation thickness is greater than the configured ash accumulation warning threshold, the overall effectiveness of different configured cleaning schemes will be evaluated based on the cleaning resource data and the predicted power generation loss. Based on the overall effectiveness of different cleaning schemes, the optimal cleaning scheme is selected from among them, and the photovoltaic power station is cleaned based on the optimal cleaning scheme.

[0009] In one optional implementation, different cleaning schemes include cleaning timing, cleaning duration, and cleaning intensity; The predicted power generation loss for the future preset time period includes: the predicted power generation loss corresponding to different cleaning times.

[0010] In an optional implementation, the overall effectiveness of different configured cleaning schemes is evaluated based on the cleaning resource data and predicted power generation losses, including: For any cleaning scheme, the initial effectiveness of implementing the cleaning scheme is evaluated based on the predicted power generation loss corresponding to the cleaning timing of the cleaning scheme. The resource consumption of the cleaning scheme is evaluated based on the cleaning intensity, the cleaning resource data, and the area of ​​the photovoltaic modules in the configured photovoltaic power station. Based on the predicted power generation loss corresponding to the cleaning timing of the cleaning scheme and the cleaning duration, the efficiency loss corresponding to the cleaning scheme is evaluated. Based on the initial efficiency, the resource consumption, and the efficiency loss, the overall efficiency of the cleaning scheme is determined.

[0011] In one optional implementation, the optimal cleaning scheme is selected from different cleaning schemes based on their overall performance, including: From different cleaning solutions, those with overall performance greater than the configured performance threshold are selected as candidate cleaning solutions; The candidate cleaning scheme with the best overall performance is determined as the optimal cleaning scheme.

[0012] Secondly, a smart cleaning device for photovoltaic power plants is provided, which may include: The acquisition unit is used to acquire environmental status data, operational data, and cleaning resource data of the photovoltaic power station in the current time period. The first prediction unit is used to input the environmental state data into a pre-trained ash accumulation evolution prediction model to obtain the predicted ash accumulation evolution trend for a future preset time period. The second prediction unit is used to input the predicted ash accumulation evolution trend and the operating data into a pre-trained power generation loss prediction model to obtain the predicted power generation loss for a future preset time period. The cleaning unit is used to determine the optimal cleaning scheme based on the cleaning resource data and the predicted power generation loss, and to clean the photovoltaic power station based on the optimal cleaning scheme.

[0013] Thirdly, an electronic device is provided, which includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; When a processor executes a program stored in memory, it implements any of the steps described in the first aspect above.

[0014] Fourthly, a computer-readable storage medium is provided, wherein a computer program is stored therein, and when executed by a processor, the computer program implements the steps of any of the methods described in the first aspect above.

[0015] This application uses a dust accumulation evolution prediction model to predict future dust accumulation trends in advance, predicts power generation loss based on the prediction results, and makes cleaning decisions based on power generation loss prediction and cleaning resource data. This completely changes the traditional passive response mode of cleaning only when dust accumulation exceeds the standard. It can avoid large power generation losses caused by dust accumulation in advance, improve operational initiative, and respond to photovoltaic cleaning in a timely manner.

[0016] This application acquires real-time environmental status data, operational data, and cleaning resource data of photovoltaic power plants. It uses a dust accumulation evolution prediction model to accurately predict future dust accumulation trends and combines it with a power generation loss prediction model to quantitatively predict power generation losses. Without human intervention, it can comprehensively match cleaning resources with power generation revenue losses, automatically determine the optimal cleaning time and plan, realize intelligent, refined, and scientific decision-making for photovoltaic power plant cleaning, effectively improve power generation efficiency, reduce cleaning costs, reduce unnecessary cleaning operations, and improve the overall operation and maintenance efficiency of the power plant. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 An architecture diagram of an intelligent cleaning system for a photovoltaic power station provided in an embodiment of this application; Figure 2 A schematic diagram of a data acquisition component provided in an embodiment of this application; Figure 3 A schematic flowchart illustrating an intelligent cleaning method for a photovoltaic power station provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an intelligent cleaning device for a photovoltaic power station provided in an embodiment of this application; Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application. Unless otherwise defined, the technical or scientific terms used in this application should have the ordinary meaning understood by those skilled in the art. The words "first," "second," and similar terms used in this application do not indicate any order, quantity, or importance, but are only used to distinguish different components. The words "comprising" or "including," etc., mean that the element or object preceding the word covers the element or object listed after the word and its equivalents, but do not exclude other elements or objects. The words "connected," "coupled," or "connected," etc., are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. "Up," "down," "left," "right," etc., are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0020] The intelligent cleaning method for photovoltaic power plants provided in this application embodiment can be applied to... Figure 1 In the system architecture shown, such as Figure 1As shown, the system may include: a server, multiple data acquisition components, and a cleaning component; the multiple data acquisition components are used to collect environmental status data, operational data, and cleaning resource data of the photovoltaic power station in the current time period; the cleaning component is used to execute the cleaning task of the photovoltaic power station according to the cleaning instructions output by the server; the server is used to execute the intelligent cleaning method of the photovoltaic power station provided in this application embodiment and generate cleaning instructions; the server may be a physical server, a server cluster composed of multiple physical servers, or a distributed system, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms. Each data acquisition component and the server can be directly or indirectly connected through wired or wireless communication methods, which is not limited herein.

[0021] In one embodiment of this application, such as Figure 2 As shown, the multiple data acquisition components may include: an environmental data acquisition component, a first operational data acquisition component, a dust accumulation characteristic data acquisition component, a flow sensor, a second operational data acquisition component, and a cleaning resource acquisition component, which are used to collect environmental data, operational data, dust accumulation characteristic data, and cleaning resource data of the photovoltaic power station's environment, respectively; the environmental data acquisition component may include: an outdoor weather station data acquisition interface, a photometer, and a particulate matter monitor; the first operational data acquisition component may include: an infrared thermal imager, a temperature sensor, and a laser particle size analyzer; the dust accumulation characteristic data acquisition component may include: a vision camera, a laser rangefinder, and a spectral analyzer; the flow sensor may include: a flow meter, a water quality analyzer, and a pressure transmitter; the second operational data acquisition component may include: a current sensor, a voltage monitoring module, and a power tester; the cleaning resource acquisition component is used to receive user-defined cleaning resource data through a 5G communication module, a WIFI communication module, and a Loara gateway.

[0022] The preferred embodiments of this application are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit this application. Furthermore, the embodiments and features in the embodiments of this application can be combined with each other without conflict.

[0023] Figure 3 This is a schematic flowchart illustrating an intelligent cleaning method for a photovoltaic power station provided in an embodiment of this application. Figure 3 As shown, the method may include: Step S310: Obtain environmental status data, operation data, and cleaning resource data of the photovoltaic power station in the current time period.

[0024] The environmental status data may include: environmental data of the environment in which the photovoltaic power station is located and dust accumulation characteristics data of the photovoltaic power station; environmental data may include: wind speed, wind direction, precipitation, air humidity, PM10 concentration, PM2.5 concentration, solar radiation intensity, and ambient temperature; dust accumulation characteristics data may include: current dust accumulation thickness, current dust accumulation coverage, current dust accumulation composition, and current dust accumulation distribution uniformity; operational data may include the output voltage, output current, output power, surface temperature, and inverter efficiency of the photovoltaic modules of the photovoltaic power station; cleaning resource data may include the number of cleaning equipment, the status of the cleaning equipment, the allocation of cleaning personnel, the supply of cleaning water, and the cost of cleaning agents.

[0025] In practice, meteorological stations and particulate matter monitors are used to collect data on wind speed, wind direction, precipitation, air humidity, PM10 concentration, PM2.5 concentration, solar radiation intensity, and ambient temperature, with a collection frequency of once every 15 minutes. Photovoltaic inverters and module-level monitoring devices are used to collect data on module output voltage, output current, output power, module surface temperature, and inverter efficiency, with a collection frequency of once every 5 minutes. Machine vision cameras, laser rangefinders, and infrared thermal imagers are used to collect data on current dust accumulation thickness, dust coverage, dust composition, and dust distribution uniformity, with a collection frequency of once every hour. The dust composition can be obtained through spectral analysis of the dust accumulation images collected by the infrared thermal imager. The cleaning equipment management system and personnel scheduling system collect data on the number and status (normal / faulty) of cleaning equipment, the number of cleaning personnel, the supply of cleaning water, and the cost of cleaning agents, with a collection frequency of once per day.

[0026] In another embodiment of this application, after acquiring environmental status data, operational data, and cleaning resource data of the photovoltaic power station using various data acquisition components, the method may further include: The collected environmental status data, operational data, and cleaning resource data are preprocessed to obtain standardized environmental status data, operational data, and cleaning resource data. Specifically, the preprocessing can include outlier removal, missing value completion, data standardization, and data fusion. Outlier removal adopts the Grubbs criterion (significance level α=0.05), missing value completion adopts the linear interpolation method based on time series, data standardization adopts min-max normalization to transform to the [0,1] interval, and data fusion adopts the Kalman filter algorithm.

[0027] Step S320: Input the environmental state data into the pre-trained ash accumulation evolution prediction model to obtain the predicted ash accumulation evolution trend for a future preset time period.

[0028] The prediction of ash accumulation evolution trends includes: predicting ash accumulation thickness and predicting ash accumulation coverage changes. The ash accumulation evolution prediction model uses a 3-layer bidirectional LSTM neural network. The input layer has 9 dimensions (corresponding to wind speed, wind direction, precipitation, PM10 concentration, PM2.5 concentration, air humidity, solar radiation intensity, ambient temperature, and current ash accumulation thickness). Each hidden layer has 64 neurons. The activation functions for the forget gate, input gate, and output gate are all Sigmoid functions. The initial value of the forget gate weight is set to 0.9. The input and output gate weights are initialized using Xavier. The cell state update activation function is tanh. The output layer is a fully connected layer (with 2 neurons, corresponding to ash accumulation thickness). The predicted dust accumulation thickness and dust coverage rate are calculated using a linear activation function and an Adam optimizer (initial learning rate 0.001, decaying by 0.9 times every 50 rounds). The batch size is 32, and the number of training rounds is 200. L2 regularization (weight decay coefficient 0.0001) is used to prevent overfitting. This model is used to predict the trend of dust accumulation thickness and dust coverage rate changes within a preset time period. The dust accumulation evolution prediction model is trained using historical current dust accumulation thickness data, historical environmental data, and corresponding historical predicted dust accumulation thickness and historical predicted dust coverage rate change trends of the corresponding photovoltaic power station. The model parameters are optimized using 5-fold cross-validation to ensure that the model prediction error is less than 5%.

[0029] Step S330: Input the predicted ash accumulation evolution trend and operating data into the pre-trained power generation loss prediction model to obtain the predicted power generation loss for the future preset time period.

[0030] The future preset time period can include the predicted power generation loss for each day within the next 1 day, 3 days, 5 days, 7 days, and 15 days or longer. The predicted power generation loss can include the predicted power generation loss value and the predicted power generation loss rate. The predicted power generation loss rate is calculated based on the predicted power generation loss value and the configured predicted theoretical power generation. The power generation loss prediction model adopts a gradient boosting tree architecture. The power generation loss prediction sub-model adopts a gradient boosting tree algorithm based on CART trees, with 100 decision trees, a learning rate of 0.05, a maximum tree depth of 6, a minimum number of sample splits of 20, a minimum number of sample leaf nodes of 10, a feature sampling ratio of 0.8, and a loss function of mean squared error (MSE). Through training, a quantitative mapping relationship between the predicted ash accumulation thickness, the predicted ash accumulation coverage change trend, and the power generation loss rate is established to obtain the predicted power generation loss for the future preset time period. The power generation loss prediction model is trained using the historical predicted ash accumulation thickness, the historical predicted ash accumulation coverage change trend, and the corresponding historical power generation of the corresponding photovoltaic power station. The model parameters are optimized using the 5-fold cross-validation method to ensure that the model prediction error is less than 5%.

[0031] Step S340: Based on the cleaning resource data and the predicted power generation loss, determine the optimal cleaning scheme, and clean the photovoltaic power station based on the optimal cleaning scheme.

[0032] The different cleaning schemes include cleaning timing, cleaning duration, and cleaning intensity. Cleaning timing can be 1 day, 3 days, 5 days, 7 days, 15 days, or longer. Cleaning duration can be [not specified]. Cleaning intensity can be light, moderate, or heavy. Light cleaning removes only surface dust; moderate cleaning removes surface dust and some stubborn dust; heavy cleaning completely removes all dust. A cleaning scheme can be formed by selecting one of the cleaning timing, cleaning duration, and cleaning intensity. The predicted power generation loss for the future preset time period includes: the predicted power generation loss corresponding to different cleaning timings; and different amounts of reagent and water consumption per square meter corresponding to different cleaning intensities.

[0033] In practice, if the predicted dust accumulation thickness exceeds the configured dust accumulation warning threshold, the overall effectiveness of different cleaning schemes is evaluated based on cleaning resource data and predicted power generation loss. The dust accumulation warning threshold is determined based on the dust accumulation thickness corresponding to the historical best power generation of the photovoltaic power station. Specifically, the initial effectiveness of the cleaning scheme is evaluated based on the predicted power generation loss corresponding to the cleaning timing, including: the initial effectiveness is the product of the total power generation loss recovered after cleaning and the configured feed-in tariff; where the total power generation loss recovered after cleaning = Σ (dust accumulation loss power generation on day i); i represents the number of days from the current time period to the cleaning timing of the cleaning scheme; the dust accumulation loss power generation on day i is the product of the configured baseline power generation and the predicted power generation loss rate on day i, where the baseline power generation can be the average daily power generation of the photovoltaic power station without dust accumulation interference for the past 30 days; the resource consumption of the cleaning scheme is evaluated based on the cleaning intensity, cleaning resource data, and the area of ​​the photovoltaic modules of the configured photovoltaic power station, including: resource consumption = cleaning equipment operating consumption × number of equipment units + cleaning personnel consumption × number of personnel + cleaning water and The cost of cleaning equipment is calculated as follows: Chemical consumption × area of ​​the configured photovoltaic modules; The operating cost of the cleaning equipment includes equipment depreciation, energy consumption, and maintenance fees. The single-run cost per unit = daily average equipment depreciation ÷ 365 + single-run energy consumption × electricity price + single-run maintenance fee; Cleaning personnel costs = daily average wage ÷ daily average working hours × single-run cleaning duration; Cleaning water and chemical consumption = (tap water unit price × water consumption per square meter) + (cleaning chemical unit price × chemical dosage per square meter); Water consumption and chemical dosage per square meter are determined based on the cleaning intensity; The pre-cleaning schedule corresponds to the cleaning timing in the cleaning plan. Measure power generation loss and cleaning time, and evaluate the corresponding efficiency loss of the cleaning scheme, including: Efficiency loss = Baseline power generation × Predicted power generation loss rate during cleaning period × Cleaning time percentage × Grid-connected electricity price; Wherein, power generation loss rate during cleaning period = Cleaning time ÷ Average daily sunshine duration × 100%, and Cleaning time percentage = Single cleaning operation time ÷ Average daily sunshine duration; Based on the initial efficiency, resource consumption, and efficiency loss, determine the comprehensive efficiency of the cleaning scheme; The difference between the initial efficiency and the sum of resource consumption and efficiency loss is taken as the comprehensive efficiency of the cleaning scheme; Based on the overall performance of different cleaning schemes, the optimal cleaning scheme is selected from among them. Specifically, for any cleaning scheme, cleaning schemes with an overall performance greater than the configured performance threshold are selected as candidate cleaning schemes. The candidate cleaning scheme with the best overall performance is determined as the optimal cleaning scheme. The performance threshold can be 0. Based on the optimal cleaning scheme, the photovoltaic power station is cleaned. Specifically, the cleaning components are controlled to clean the photovoltaic power station according to the cleaning intensity in the optimal cleaning scheme, starting at the optimal cleaning time and continuing for the optimal cleaning duration. The control parameters of each cleaning device in the cleaning components are determined according to the cleaning intensity. Specifically, when the cleaning intensity is light cleaning, the water pressure of the cleaning equipment is 0.3-0.5MPa, the rotation speed of the cleaning brush head is 300-500r / min, and the moving speed is 0.5-1m / s; when the cleaning intensity is medium cleaning, the water pressure of the cleaning equipment is 0.5-0.8MPa, the rotation speed of the cleaning brush head is 500-800r / min, and the moving speed is 0.3-0.5m / s; when the cleaning intensity is heavy cleaning, the water pressure is 0.8-1.2MPa, the rotation speed of the cleaning brush head is 800-1200r / min, and the moving speed is 0.1-0.3m / s.

[0034] In another embodiment of this application, if the current dust accumulation coverage rate is greater than the configured dust accumulation coverage rate threshold, the comprehensive effectiveness of different configured cleaning schemes is evaluated based on cleaning resource data and predicted power generation loss; wherein, the dust accumulation coverage rate threshold can be 15%; the dust accumulation coverage rate threshold is determined based on the dust accumulation coverage rate corresponding to the historical best power generation of the photovoltaic power station.

[0035] In another embodiment of this application, if the overall performance of each cleaning scheme is not greater than 0, the cleaning is suspended, and the predicted ash accumulation evolution trend and overall performance are updated every 12 hours, and the screening is carried out again.

[0036] In another embodiment of this application, the cleaning scheme may further include: a cleaning strategy; the method may further include: determining a cleaning strategy for different areas of the photovoltaic power station based on the current dust accumulation coverage rate, the current dust accumulation composition, and the current dust accumulation distribution uniformity; specifically, a slow-walking multiple-pass cleaning strategy is adopted for areas where the current dust accumulation coverage rate is not less than a configured first coverage rate threshold; a fast-walking once cleaning strategy is adopted for areas where the current dust accumulation coverage rate is not greater than a configured second coverage rate threshold; wherein, the first coverage rate threshold may be 20%; and the second coverage rate threshold may be 10%.

[0037] In another embodiment of this application, the method may further include: after the cleaning task is completed, the dust accumulation feature data acquisition device immediately re-inspects the surface of the photovoltaic modules of the photovoltaic power station.

[0038] In another embodiment of this application, the intelligent cleaning method for the photovoltaic power station may include: Taking a 100MW photovoltaic power station in a windy and sandy area in Northwest my country as an example, the power station covers an area of ​​about 2,000 acres, has 400,000 monocrystalline silicon photovoltaic modules, an average annual sunshine duration of 3,200 hours, an average annual wind speed of 3.5m / s, and an average annual atmospheric particulate matter concentration (PM10) of 120μg / m³. The problem of dust accumulation is prominent, and the traditional passive cleaning mode results in an average annual power generation loss rate of 18%.

[0039] The specific implementation process is as follows: 1. Multi-source data acquisition deployment A multi-source data acquisition system was constructed within the photovoltaic power station: (1) Environmental data collection: Three automatic weather stations are evenly deployed in the power station to collect wind speed, wind direction, precipitation, air humidity and ambient temperature; two particulate matter monitors are deployed to collect PM10 and PM2.5 concentrations; one solar intensity meter is deployed to collect solar intensity; the collection frequency is once every 15 minutes. (2) Operational data acquisition: The output voltage, output current and output power of the photovoltaic inverter are acquired through the RS485 interface; 1 photovoltaic module is selected out of every 10 modules, and a module surface temperature sensor is installed to acquire the module surface temperature; Inverter efficiency data is acquired; The acquisition frequency is once every 5 minutes. (3) Data collection of dust accumulation characteristics: 10 representative areas (each area contains 100 modules) were selected in the power plant. One machine vision camera and one laser rangefinder were deployed in each area to collect dust accumulation thickness and dust accumulation coverage. One spectral analysis module was deployed in the center of the power plant to collect dust accumulation samples for composition analysis at regular intervals. The surface images of the modules were captured by an infrared thermal imager to analyze the uniformity of dust accumulation distribution. The collection frequency was once per hour. (4) Cleaning resource data collection: The operating status (normal / fault) of 5 high-pressure cleaning vehicles in the power station is collected through the cleaning equipment management system; the number of cleaning personnel (10 people in total) is collected through the personnel dispatch system; the water supply for cleaning is collected through the water management system; the unit cost of cleaning agents (neutral detergent) is calculated; the collection frequency is 1 time / day.

[0040] 2. Data Preprocessing Implementation Preprocess the collected multi-source data: (1) Outlier removal: The Grubbs criterion (α=0.05) was used to remove abnormal peak data of PM10 concentration caused by sandstorm weather (such as PM10 concentration reaching 1500μg / m³ at a certain moment, far exceeding the normal range) and abnormal data of component output power of 0 caused by sensor failure. A total of 326 abnormal data were removed. (2) Missing value completion: For a meteorological station with 2 hours of missing data due to communication interruption, linear interpolation was used to complete the missing data based on the valid data before and after the missing period. The completed data showed a consistent trend with the data in the adjacent period. (3) Data standardization: The min-max normalization method is used to convert data of different dimensions such as wind speed (0-10m / s), PM10 concentration (0-500μg / m³), and component output power (0-250W) to the [0,1] interval; (4) Data fusion: The Kalman filter algorithm is used to fuse the dust thickness data collected by the machine vision camera and the laser rangefinder. The error of the fused dust thickness data is reduced from ±0.2mm to ±0.05mm, and the reliability is significantly improved.

[0041] 3. Implementation of quantitative prediction of ash accumulation and power generation Based on the preprocessed standardized data, a quantitative prediction model for ash accumulation and power generation is constructed: (1) Construction of the ash accumulation evolution prediction sub-model: Using wind speed, wind direction, PM10 concentration, and air humidity as input variables, an LSTM neural network model was constructed. The core parameters were refined as follows: the input layer has a dimension of 9, the hidden layer is a 3-layer bidirectional LSTM structure with 64 neurons per layer, the initial value of the forget gate weight is set to 0.9, the initial values ​​of the input gate and output gate weights are initialized using Xavier, the cell state update activation function is tanh, and the output layer is a fully connected layer (with 2 neurons, corresponding to the predicted values ​​of ash accumulation thickness and ash accumulation coverage). The optimizer is the Adam optimizer with an initial learning rate of 0.001, a decay of 0.9 times every 50 rounds, a batch size of 32, and 200 training rounds. L2 regularization (weight decay coefficient of 0.0001) was used to prevent overfitting, and the loss function was the mean squared error (MSE). The model was trained using historical environmental data and ash accumulation data of the power station over the past 3 years. The trained model can predict the trend of ash accumulation thickness and ash accumulation coverage in the next 7 days. According to the dust coverage rate forecast curve, the dust coverage rate will gradually increase from the current 8% to 22% in the next 7 days. (2) Construction of the sub-model for predicting power generation loss: Using the ash thickness, ash coverage rate, and photovoltaic module operation data (solar intensity, module surface temperature, etc.) output by the ash accumulation evolution prediction sub-model as input variables, a gradient boosting tree model is constructed. The core parameters are refined as follows: the basic model is a CART regression tree, the number of decision trees is 100, the learning rate is 0.05, the maximum tree depth is 6, the minimum number of sample splits is 20, the minimum number of sample leaf nodes is 10, the feature sampling ratio is 0.8, the loss function is the mean squared error (MSE), and the weights of the tree are updated using the gradient descent method. Through training, a quantitative mapping relationship between ash accumulation features and power generation loss rate is established, and the predicted value of power generation loss under different ash accumulation levels is output. (3) Model verification: The quantitative prediction model was verified by the 5-fold cross-validation method. The verification results showed that the prediction error of ash coverage rate was 3.2% and the prediction error of power generation loss rate was 2.8%, both less than 5%, which met the requirements of engineering application.

[0042] 4. Implementation of Economic Benefit Evaluation The economic benefit assessment parameters for this photovoltaic power station are defined as follows: the operating cost of the cleaning equipment is 500 yuan / time; the cost of cleaning personnel is 200 yuan / person / time; the cost of cleaning water and chemicals is 0.8 yuan / ㎡; the grid-connected electricity price is 0.38 yuan / kWh; the cleaning operation time is 4 hours / time; and the power generation loss rate during the cleaning period is 2%.

[0043] Based on four cleaning opportunities (1 day, 3 days, 5 days, and 7 days) and three cleaning intensities (light, moderate, and heavy), 12 cleaning plans are combined, and the net benefit of each plan is calculated: Taking the next 3 days and moderate cleaning plan as an example, the following precise calculations are performed based on the refined parameters: ① Determine the basic parameters: Baseline power generation = 2.8 million kWh / day (reasonable daily power generation for a 100MW photovoltaic power station), total photovoltaic module area = 500,000㎡ (100MW module efficiency 20%, 100×10 6 W÷(1000W / ㎡×20%)), 5 cleaning equipment units and 10 cleaning personnel were deployed; ② Calculate cleaning revenue: According to the quantitative prediction model, the power generation loss rate due to ash accumulation in the next 3 days is 10%, 12%, and 14% respectively. Therefore, the total power generation loss in 3 days = 2.8 million kWh × (10% + 12% + 14%) = 2.8 million × 0.36 = 1.008 million kWh, and the cleaning revenue = 1.008 million kWh × 0.38 yuan / kWh = 383,040 yuan; ③ Calculate total cleaning cost: Equipment operating cost = 140.75 yuan / unit × 5 units ≈ 703.75 yuan; Personnel cost = 150 yuan / person × 10 people = 1500 yuan; Water and chemicals Agent cost = 0.85 yuan / ㎡ × 500,000㎡ = 425,000 yuan; Total cleaning cost = 703.75 + 1,500 + 425,000 = 427,203.75 yuan; ④ Calculate the cost of power generation loss during cleaning: Cleaning time is 4 hours, average daily sunshine is 8 hours, loss rate is 50%, power generation loss = 2.8 million kWh × 50% × (4 ÷ 8) = 700,000 kWh, loss cost = 700,000 kWh × 0.38 yuan / kWh = 266,000 yuan; ⑤ Calculate the net profit: Net profit = 383,040 - 427,203.75 - 266,000 = -310,163.75 yuan (the net profit of this plan is negative and it is not economical).

[0044] Similarly, other options are calculated. Taking the next 5 days and moderate cleaning options as examples, the precise calculation process is as follows: ① Cleaning revenue: The power generation loss rates due to ash accumulation in the next 5 days are 10%, 12%, 14%, 16%, and 18%, respectively. The total power generation loss = 2.8 million kWh × (10% + 12% + 14% + 16% + 18%) = 2.8 million × 0.7 = 1.96 million kWh. The cleaning revenue = 1.96 million kWh × 0.38 yuan / kWh = 744,800 yuan; ② Total cleaning cost: Consistent with the "next 3 days + moderate cleaning" option, it is 427,203.75 yuan; ③ Power generation loss cost during cleaning: Consistent with the "next 3 days + moderate cleaning" option, it is 266,000 yuan; ④ Net revenue = 744,800 - 427,203.75 - 266,000 = 51,596.25 yuan (this option has the highest and positive net revenue, and is therefore determined to be the optimal cleaning option). This refined calculation logic allows for accurate assessment of the economic viability of different cleaning solutions, providing reliable data support for decision-making.

[0045] 5. Proactive control decision-making and cleanup execution The dust accumulation warning threshold was set at a dust coverage rate ≥ 15%, and the net benefit threshold was 0. The dust accumulation evolution prediction sub-model predicted that the dust coverage rate would reach 18% in the next 5 days, exceeding the dust accumulation warning threshold and triggering the screening of economic benefit assessment results. The screening found that the "5-day + moderate cleaning" option had the highest net benefit (445,126.67 yuan), which was also greater than the net benefit threshold. Therefore, this option was determined to be the optimal cleaning option.

[0046] The server outputs a cleaning command: the cleaning time is from 9:00 AM to 1:00 PM five days from now (when the sunlight intensity is moderate to avoid damage to the components from high-temperature cleaning); the cleaning intensity is medium; the key cleaning areas are three representative areas with a dust accumulation coverage rate of ≥20%.

[0047] After receiving the command, the cleaning execution system adjusted the water pressure of the five high-pressure cleaning trucks to 0.6 MPa, the cleaning brush head speed to 600 r / min, and the moving speed to 0.4 m / s. A "slow pass twice" cleaning strategy was adopted for key cleaning areas, while a "fast pass once" cleaning strategy was used for other areas. After the cleaning operation was completed, the dust accumulation characteristic data acquisition equipment showed that the average dust accumulation coverage of the modules decreased to 3%, and the power generation loss rate decreased to 1.2%, achieving the expected results.

[0048] 6. Comparison of Implementation Results After adopting the intelligent cleaning method for photovoltaic power plants provided in this application, the average annual power generation loss rate of the photovoltaic power plant has been reduced from 18% in the traditional passive cleaning mode to 6%, with an average annual increase in power generation of 1.008 million kWh, corresponding to an increase in revenue of 383,000 yuan. At the same time, it avoids the cost waste caused by over-cleaning, reducing the average annual cleaning cost from 2 million yuan in the traditional mode to 1.2 million yuan, and increasing the average annual net profit by 1.183 million yuan, demonstrating significant economic benefits. In addition, through continuous optimization of model parameters by the closed-loop control system, the prediction error has been further reduced to within 2%, and the accuracy of decision-making has been continuously improved.

[0049] Corresponding to the above method, this application also provides an intelligent cleaning device for photovoltaic power plants, such as... Figure 4 As shown, the device includes: The acquisition unit 410 is used to acquire environmental status data, operation data and cleaning resource data of the photovoltaic power station in the current time period; The first prediction unit 420 is used to input environmental state data into a pre-trained ash accumulation evolution prediction model to obtain the predicted ash accumulation evolution trend for a future preset time period. The second prediction unit 430 is used to input the predicted ash accumulation evolution trend and operating data into the pre-trained power generation loss prediction model to obtain the predicted power generation loss for a future preset time period. The cleaning unit 440 is used to determine the optimal cleaning scheme based on cleaning resource data and predicted power generation loss, and to clean the photovoltaic power station based on the optimal cleaning scheme.

[0050] The functions of each functional unit of the intelligent cleaning device for photovoltaic power plants provided in the above embodiments of this application can be realized through the above methods and steps. Therefore, the specific working process and beneficial effects of each unit in the intelligent cleaning device for photovoltaic power plants provided in the embodiments of this application will not be repeated here.

[0051] This application also provides an electronic device, such as... Figure 5 As shown, it includes a processor 510, a communication interface 520, a memory 530, and a communication bus 540, wherein the processor 510, the communication interface 520, and the memory 530 communicate with each other through the communication bus 540.

[0052] Memory 530 is used to store computer programs; When the processor 510 executes the program stored in the memory 530, it performs the following steps: Obtain environmental status data, operational data, and cleaning resource data of the photovoltaic power station for the current time period; Environmental status data is input into a pre-trained ash accumulation evolution prediction model to obtain the predicted ash accumulation evolution trend for a future preset time period. The predicted ash accumulation evolution trend and operational data are input into a pre-trained power generation loss prediction model to obtain the predicted power generation loss for a future preset time period. Based on cleaning resource data and predicted power generation losses, the optimal cleaning scheme is determined, and the photovoltaic power station is cleaned based on the optimal cleaning scheme.

[0053] The communication bus mentioned above can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not mean that there is only one bus or one type of bus.

[0054] The communication interface is used for communication between the aforementioned electronic devices and other devices.

[0055] The memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0056] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0057] The implementation methods and beneficial effects of the various components of the electronic device in the above embodiments for solving the problem can be found in [reference needed]. Figure 3 The steps in the illustrated embodiments are used to implement the electronic device. Therefore, the specific working process and beneficial effects of the electronic device provided in this application will not be repeated here.

[0058] In another embodiment provided in this application, a computer-readable storage medium is also provided, which stores instructions that, when executed on a computer, cause the computer to perform the intelligent cleaning method for a photovoltaic power station as described in any of the above embodiments.

[0059] In another embodiment provided in this application, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute the intelligent cleaning method for photovoltaic power plants described in any of the above embodiments.

[0060] Those skilled in the art will understand that the embodiments in this application can be provided as methods, systems, or computer program products. Therefore, the embodiments in this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the embodiments in this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0061] This application describes embodiments of methods, apparatus (systems), and computer program products according to embodiments of this application with reference to flowchart illustrations and / or block diagrams. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0062] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0063] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0064] Although preferred embodiments have been described in this application, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of this application.

[0065] Obviously, those skilled in the art can make various modifications and variations to the embodiments of this application without departing from the spirit and scope of the embodiments of this application. Therefore, if these modifications and variations to the embodiments of this application fall within the scope of this application and its equivalents, then these modifications and variations are also intended to be included in the embodiments of this application.

Claims

1. A smart cleaning method for photovoltaic power plants, characterized in that, The method includes: Obtain environmental status data, operational data, and cleaning resource data of the photovoltaic power station for the current time period; The environmental state data is input into a pre-trained ash accumulation evolution prediction model to obtain the predicted ash accumulation evolution trend for a future preset time period. The predicted ash accumulation evolution trend and the operating data are input into a pre-trained power generation loss prediction model to obtain the predicted power generation loss for a future preset time period. Based on the cleaning resource data and the predicted power generation loss, an optimal cleaning scheme is determined, and the photovoltaic power station is cleaned based on the optimal cleaning scheme.

2. The method as described in claim 1, characterized in that, The environmental status data includes: environmental data of the environment in which the photovoltaic power station is located and dust accumulation characteristics data of the photovoltaic power station; The predicted ash accumulation evolution trend includes: the predicted ash accumulation thickness and the predicted ash accumulation coverage change trend.

3. The method as described in claim 2, characterized in that, The ash accumulation evolution prediction model adopts a bidirectional LSTM neural network architecture; The power generation loss prediction model adopts a gradient boosting tree architecture.

4. The method as described in claim 2, characterized in that, Based on the cleaning resource data and the predicted power generation loss, an optimal cleaning scheme is determined, and the photovoltaic power station is cleaned according to the optimal cleaning scheme, including: If the predicted ash accumulation thickness is greater than the configured ash accumulation warning threshold, the overall effectiveness of different configured cleaning schemes will be evaluated based on the cleaning resource data and the predicted power generation loss. Based on the overall effectiveness of different cleaning schemes, the optimal cleaning scheme is selected from among them, and the photovoltaic power station is cleaned based on the optimal cleaning scheme.

5. The method as described in claim 4, characterized in that, Different cleaning solutions include the timing of cleaning, the duration of cleaning, and the intensity of cleaning; The predicted power generation loss for the future preset time period includes: the predicted power generation loss corresponding to different cleaning times.

6. The method as described in claim 5, characterized in that, Based on the cleaning resource data and predicted power generation losses, the overall effectiveness of different cleaning schemes is evaluated, including: For any cleaning scheme, the initial effectiveness of implementing the cleaning scheme is evaluated based on the predicted power generation loss corresponding to the cleaning timing of the cleaning scheme. The resource consumption of the cleaning scheme is evaluated based on the cleaning intensity, the cleaning resource data, and the area of ​​the photovoltaic modules in the configured photovoltaic power station. Based on the predicted power generation loss corresponding to the cleaning timing of the cleaning scheme and the cleaning duration, the efficiency loss corresponding to the cleaning scheme is evaluated. Based on the initial efficiency, the resource consumption, and the efficiency loss, the overall efficiency of the cleaning scheme is determined.

7. The method as described in claim 5, characterized in that, Based on the overall effectiveness of different cleaning solutions, the optimal cleaning solution is selected from among them, including: From different cleaning solutions, those with overall performance greater than the configured performance threshold are selected as candidate cleaning solutions; The candidate cleaning scheme with the best overall performance is determined as the optimal cleaning scheme.

8. An intelligent cleaning device for a photovoltaic power station, characterized in that, The device includes: The acquisition unit is used to acquire environmental status data, operational data, and cleaning resource data of the photovoltaic power station in the current time period. The first prediction unit is used to input the environmental state data into a pre-trained ash accumulation evolution prediction model to obtain the predicted ash accumulation evolution trend for a future preset time period. The second prediction unit is used to input the predicted ash accumulation evolution trend and the operating data into a pre-trained power generation loss prediction model to obtain the predicted power generation loss for a future preset time period. The cleaning unit is used to determine the optimal cleaning scheme based on the cleaning resource data and the predicted power generation loss, and to clean the photovoltaic power station based on the optimal cleaning scheme.

9. An electronic device, characterized in that, The electronic device includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor, when executing a program stored in memory, implements the method of any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method described in any one of claims 1-7.

Images