A light power prediction system, method and electronic equipment suitable for photovoltaic power stations

Abstract

The present invention discloses a light power prediction system, method, and electronic device applicable to a photovoltaic power station, and belongs to the field of photovoltaic power generation technology. The method obtains the current climate and environmental data of the photovoltaic power station in real time, and the climate and environmental data include altitude data, slope data, time data, temperature data, humidity data, wind speed data, and light intensity data; the environmental data are grouped according to slope climate and environmental data and flat climate and environmental data; the slope climate and environmental data and the flat climate and environmental data are preprocessed to obtain a data group 1 and a data group 2 aligned in time, the data group 1 including slope environmental data; the data group 2 including flat environmental data; and the light power of the photovoltaic power station is predicted based on the data group 1 and the data group 2. The present application performs fine division, grouping, and training on the power generation units of the photovoltaic power station, thereby improving the prediction accuracy of the generated power.

Description

Technical Field

[0001] The present disclosure belongs to the field of photovoltaic power generation technology, and in particular relates to a light power prediction system, method, and electronic equipment applicable to a photovoltaic power station. Background Art

[0002] Existing optical power prediction systems predict the power generation of photovoltaic power stations based on factors such as ambient temperature, light radiation, air pressure, relative humidity, and meteorological data. Photovoltaic power station power prediction requires a large amount of data and includes numerical weather forecast data, real-time meteorological data, real-time power data, operating status and planned maintenance information, but ignores regional differences in meteorological data.

[0003] In addition, the existing photovoltaic power prediction system focuses on the correlation between the output power of the photovoltaic power station and the solar radiation intensity, while considering the influence of temperature, relative humidity and cloud cover. Combined with the correlation analysis of the relationship between various meteorological factors and photovoltaic output, the actual solar radiation intensity, temperature, relative humidity of the historical measured meteorological conditions and the active power generation power in the historical output data of the photovoltaic power station are determined as input layer sample data for training to obtain the predicted photovoltaic power generation power.

[0004] However, the solar radiation intensity, temperature, and relative humidity in historical meteorological data do not take into account the regional differences in the distribution of photovoltaic panels. The solar radiation intensity, temperature, and relative humidity in the historical meteorological data of the same region are used as standard data for training photovoltaic power stations. Since the meteorological data corresponding to the actual operation of multiple photovoltaic panels in distributed photovoltaic power stations deviate from the standard data, there are deviations in the correlation analysis of the relationship between various meteorological factors and photovoltaic output, resulting in inaccurate photovoltaic power prediction values.

[0005] In addition, the existing photovoltaic power prediction system does not take into account that the multiple photovoltaic panels of the photovoltaic power station are actually arranged in a hillside environment. When the multiple photovoltaic panels of the actual photovoltaic power station are in operation, the wind forces on different slopes are different. For example, the photovoltaic panels on the left slope are in a windy environment and the photovoltaic panels on the right slope are in a windless environment. On the one hand, the wind on the left slope will take away some of the heat on the photovoltaic panels, thereby making the operating temperature of the photovoltaic panels on the left slope lower than the operating temperature of the photovoltaic panels on the right slope; on the other hand, the wind on the left slope will take away some of the dust on the photovoltaic panels, thereby making the surface cleanliness of the photovoltaic panels on the left slope higher than the surface cleanliness of the photovoltaic panels on the right slope; the difference in operating temperature and surface cleanliness of the photovoltaic panels on the left and right slopes makes the power generation of different photovoltaic panels different. There will be deviations in prediction using a unified training model, resulting in a decrease in the accuracy of the predicted value.

[0006] Public content

[0007] The present disclosure aims to provide a system, method, and electronic device for predicting optical power in photovoltaic power plants, addressing the problem of inaccurate predictions caused by deviations resulting from predictions using a unified training model. The method, based on slopes, flat land, and varying wind speeds, meticulously divides and groups the power generation units of a photovoltaic power plant using photovoltaic strings as units, and trains each group separately. This balances the effects of differences in photovoltaic panel heating, cleanliness, heat dissipation, and light radiation intensity on power generation, improving power generation prediction accuracy.

[0008] A method for predicting light power for a photovoltaic power station, applied in the field of green electricity hydrogen production, comprises:

[0009] Acquire climate and environmental data of the photovoltaic power station in real time, including altitude data, slope data, time data, temperature data, humidity data, wind speed data, and light intensity data;

[0010] Grouping the climate and environment data according to slope climate and environment data and flat land climate and environment data;

[0011] Preprocessing the slope climate environment data and the flat land climate environment data to obtain a first data set and a second data set aligned in time, wherein the first data set includes the slope environment data and the second data set includes the flat land environment data;

[0012] The power generation power is predicted based on the data set 1 and the data set 2.

[0013] The method further comprises:

[0014] The data cleaning and data alignment of the slope climate environment data and the flat land climate environment data includes:

[0015] Data cleaning, processing missing values ​​and outliers;

[0016] Data standardization, standardize or normalize the cleaned data;

[0017] Feature extraction, extracting target feature data from standardized or normalized environmental data, the target is Shanghai 2024-11487 (M240487CNI)

[0018] Characteristic data include time characteristics, meteorological characteristics and geographical characteristics;

[0019] Data alignment: align the target feature data of the slope and the target feature data of the flat land by time.

[0020] The method further comprises:

[0021] The performing of optical power prediction of the photovoltaic power station based on the data group 1 and the data group 2 includes:

[0022] Inputting the slope optical power prediction model and the flat land optical power prediction model corresponding to the data group 1 and the data group 2 respectively obtains the slope optical power prediction value and the flat land optical power prediction value;

[0023] Based on the slope land optical power prediction value and the flat land optical power prediction value, a total optical power theoretical prediction value is calculated.

[0024] The method further comprises:

[0025] The input of the slope optical power prediction model also includes the optical power prediction value of the flat land optical power prediction model, and the output of the slope optical power prediction model also includes a first deviation value, which is the difference between the slope optical power prediction value and the flat land optical power prediction value;

[0026] The input parameters of the flat land optical power prediction model also include the optical power prediction value of the slope land optical power prediction model, and the output parameters of the flat land optical power prediction model also include a second deviation value, which is the difference between the flat land optical power prediction value and the optical power prediction value of the slope land optical power prediction model;

[0027] The optical power prediction value of the photovoltaic power station obtained based on the slope optical power prediction value and the flat land optical power prediction value further includes:

[0028] Correcting the slope optical power prediction value according to the first deviation value to obtain a first corrected slope optical power prediction value;

[0029] Correcting the predicted value of the flat-ground optical power according to the second deviation value to obtain a first corrected value of the predicted value of the flat-ground optical power;

[0030] The first revised value of the predicted light power of the photovoltaic power station is calculated based on the first revised value of the predicted light power of the slope and the first revised value of the predicted light power of the flat.

[0031] The method further comprises:

[0032] The first deviation value is used to correct the slope optical power prediction value to obtain a first correction value of the slope optical power prediction value, including: the first correction value of the slope optical power prediction value = the slope optical power prediction value + K1 * Shanghai 2024-11487 (M240487CNI)

[0033] The first deviation value, K1 first correction coefficient;

[0034] The method of correcting the predicted value of the flat ground optical power according to the second deviation value to obtain a first corrected value of the predicted value of the flat ground optical power includes: the first corrected value of the predicted value of the flat ground optical power = the predicted value of the flat ground optical power + K2*

[0035] The second deviation value, K2 second correction coefficient;

[0036] The first corrected value of the predicted light power of the photovoltaic power station is calculated based on the first corrected value of the predicted light power of the slope and the first corrected value of the predicted light power of the flat land, including: summing the first corrected value of the predicted light power of the slope and the first corrected value of the predicted light power of the flat land to obtain the first corrected value of the predicted light power of the photovoltaic power station.

[0037] The method further comprises:

[0038] Obtaining historical forecast data of flat land optical power, historical forecast data of sloped land optical power, and historical forecast data of photovoltaic power station optical power having the same time characteristics, meteorological characteristics, and geographical characteristics as the data group one and the data group two, wherein the historical forecast data of flat land optical power, historical forecast data of sloped land optical power, and historical forecast data of photovoltaic power station optical power having the same time characteristics, meteorological characteristics, and geographical characteristics means that the deviation is within a preset range, and the historical forecast data of optical power is theoretical forecast value data of optical power whose deviation between the theoretical forecast value of optical power and the actual value of optical power is within the preset range;

[0039] Comparing the first corrected value of the slope land optical power prediction value of the photovoltaic power station with the slope land optical power historical prediction data to obtain a third deviation value, and comparing the first corrected value of the flat land optical power prediction value of the photovoltaic power station with the flat land optical power historical prediction data to obtain a fourth deviation value;

[0040] Determine the second correction value of the slope land optical power prediction value according to the third deviation value, and determine the second correction value of the flat land optical power prediction value according to the fourth deviation value,

[0041] The second correction value of the optical power prediction value is calculated based on the second correction value of the optical power prediction value on the slope and the second correction value of the optical power prediction value on the flat land;

[0042] Comparing the second correction value of the optical power prediction value with the historical optical power prediction data of the photovoltaic power station to obtain a fifth deviation value;

[0043] When the fifth deviation value is less than or equal to the set deviation value, outputting a theoretical predicted value of the total optical power of the photovoltaic power station;

[0044] If the fifth deviation value is greater than the set deviation value, the second correction value of the optical power prediction value is Shanghai 2024-11487 (M240487CNI)

[0045] Correction is performed to obtain a third corrected value of the optical power prediction value, and the third corrected value of the optical power prediction value is output as a theoretical predicted value of the total optical power of the photovoltaic power station.

[0046] The method further comprises:

[0047] Obtain actual optical power data during the forecast period;

[0048] The actual optical power data is compared with the theoretical predicted value of the total optical power of the photovoltaic power station to obtain a sixth deviation value. If the sixth deviation value is less than or equal to the set deviation value, the third correction value of the optical power prediction value is stored as historical data; if the sixth deviation value is greater than the set deviation value, the slope optical power prediction model is optimized according to the first deviation value and the third deviation value; and the flat land optical power prediction model is optimized according to the second deviation value and the fourth deviation value.

[0049] Optionally, the photovoltaic power station is provided with a plurality of photovoltaic strings, each photovoltaic string comprising a plurality of photovoltaic panels arranged in series, and the plurality of photovoltaic panels are connected in parallel to a combiner box;

[0050] The obtaining of the current climate and environmental data of the photovoltaic power station includes obtaining environmental data of each photovoltaic string;

[0051] Grouping the environmental data according to slope climate environmental data and flat land climate environmental data includes:

[0052] The photovoltaic strings with a wind speed less than or equal to a first preset value in the photovoltaic power station are divided into a slope low wind group within the station;

[0053] The photovoltaic strings in the photovoltaic power station with a wind speed of the first preset value < ≦ the second preset value are divided into the slope and moderate wind group within the station;

[0054] The photovoltaic strings with wind speed greater than the second preset value in the photovoltaic power station are divided into the high wind group on the slope within the station;

[0055] The data of each photovoltaic string in the station slope low wind group, the station slope medium wind group and the station slope high wind group include photovoltaic data and environmental data;

[0056] Grouping the photovoltaic data and environmental data according to slope data and flat land data includes:

[0057] The photovoltaic strings with wind speed less than or equal to the fourth preset value in the photovoltaic power station are divided into the flat and gentle wind group within the station;

[0058] The photovoltaic strings with wind speed greater than the third preset value in the photovoltaic power station are divided into the flat ground and wind groups within the station.

[0059] The slope light power prediction model and the flat land light power prediction model each include a plurality of parallel DNN neural network modules, the station slope low wind group, the station slope medium wind group and the station slope high wind group of each photovoltaic string within the statistical time are respectively input into the independent DNN neural network module of the slope light power prediction model, the station flat land slow wind group and the station flat land and wind group data are respectively input into the independent DNN neural network module of the flat land light power prediction model, and the slope light power prediction model and the flat land light power prediction model are respectively connected. Shanghai 2024-11487 (M240487CNI)

[0060] The output power generation of each DNN neural network module of the power prediction model is accumulated to obtain the slope light power prediction value and the flat land light power prediction value.

[0061] A light power prediction system for a photovoltaic power station includes a photovoltaic power station control room controller and multiple photovoltaic strings. The photovoltaic power station control room controller communicates with each photovoltaic string in the photovoltaic power station and exchanges data.

[0062] An electronic device stores a computer program, which, when executed by a processor, implements a light power prediction method applicable to a photovoltaic power station.

[0063] Beneficial technical effects:

[0064] By finely dividing and grouping the power generation units of the photovoltaic power station with photovoltaic strings as units according to slopes, flat land and different wind speeds, and training each group separately, the impact of differences in photovoltaic panel heating, cleanliness, heat dissipation and light radiation intensity on the power generation power is balanced, thereby improving the prediction accuracy of the power generation power. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present disclosure. The same reference symbols are used throughout the drawings to represent the same components. In the drawings:

[0066] Figure 1 This is a schematic diagram of the arrangement of photovoltaic panels on a slope in a photovoltaic power station according to an embodiment of the present disclosure.

[0067] Figure 2 Schematic diagram of the photovoltaic panel structure according to an embodiment of the present disclosure.

[0068] Figure 3 Schematic diagram of a photovoltaic string and combiner box according to an embodiment of the present disclosure.

[0069] Figure 4 Schematic diagram of the data measurement arrangement of the photovoltaic string and combiner box according to an embodiment of the present disclosure.

[0070] Figure 5 This is a schematic diagram of the parallel connection of the method model of the embodiment of the present disclosure. DETAILED DESCRIPTION

[0071] In order to make the purpose, technical solutions and advantages of the present disclosure more clear, the present disclosure is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present disclosure and are not intended to limit the present disclosure. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work are the property of this company. Shanghai 2024-11487 (M240487CNI)

[0072] Open the protection range.

[0073] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or apparatus comprising the element.

[0074] Green Hydrogen via Electrolysis refers to the use of renewable energy sources such as solar energy and wind energy to drive the water electrolysis process to produce hydrogen. In this process, it is crucial to accurately predict the power generation of renewable energy, because this directly affects the working efficiency of the electrolyzer and the hydrogen production. The concept of the present disclosure is that, unlike the traditional photovoltaic station data-training-prediction method, this application subdivides and groups the photovoltaic station data, thereby depicting smaller and more accurate dimensions that affect power generation, and then conducts training, and accumulates the results to obtain predicted values. For a photovoltaic power station system based on solar energy, a light power prediction model is required to estimate the solar radiation intensity in the future, and predict the output power of the photovoltaic power generation system based on this.

[0075] Example 1

[0076] Photovoltaic power generation is a green and clean energy. Using photovoltaic green energy to produce hydrogen not only improves the utilization rate of green energy but also improves the local balance of photovoltaic power generation, thereby avoiding the loss of long-distance transmission. Using photovoltaic green energy to produce hydrogen involves the prediction of photovoltaic power generation. In the existing photovoltaic power prediction system, the output power of the photovoltaic power station is focused on the correlation between the intensity of solar radiation and the intensity of solar radiation. At the same time, the influence of temperature, relative humidity and cloud cover is considered. Combined with the correlation analysis of the relationship between various meteorological factors and photovoltaic output, the actual solar radiation intensity, temperature, relative humidity of historical measured meteorological conditions and the active power generation power in the historical output data of photovoltaic power stations are determined as input layer sample data for training to obtain the predicted photovoltaic power generation power.

[0077] However, the solar radiation intensity, temperature and relative humidity in the historical meteorological data do not take into account the regional differences in the distribution of photovoltaic panels. The solar radiation intensity, temperature and relative humidity in the historical meteorological data of the same region are used as the standard data for photovoltaic power stations for training. Since the meteorological data corresponding to the actual operation of multiple photovoltaic panels in distributed photovoltaic power stations deviate from the standard data, the relationship between various meteorological factors and photovoltaic output is not well understood.

[0078] There is a deviation in the correlation analysis, which results in inaccurate photovoltaic power prediction values.

[0079] like Figure 1 As shown, the existing photovoltaic power prediction system does not take into account that the multiple photovoltaic panels of the photovoltaic power station are actually arranged in a hillside environment. When the multiple photovoltaic panels of the actual photovoltaic power station are in operation, the wind forces on different slopes are different. For example, the photovoltaic panels on the left slope are in a windy environment and the photovoltaic panels on the right slope are in a windless environment. On the one hand, the wind on the left slope will take away part of the heat on the photovoltaic panels, so that the operating temperature of the photovoltaic panels on the left slope is lower than that of the photovoltaic panels on the right slope; on the other hand, the wind on the left slope will take away part of the dust on the photovoltaic panels, so that the surface cleanliness of the photovoltaic panels on the left slope is higher than that of the photovoltaic panels on the right slope; the difference in operating temperature and surface cleanliness of the photovoltaic panels on the left and right slopes makes the power generation of different photovoltaic panels different. There will be deviations in prediction using a unified training model, resulting in a decrease in the accuracy of the predicted value.

[0080] like Figure 2As shown, each photovoltaic panel of a photovoltaic power station contains multiple photoelectric conversion units. For example, each photovoltaic panel includes 60 photoelectric conversion units. A photovoltaic power station has 1,000 photovoltaic panels, which means that a photovoltaic power station contains a total of 60,000 photoelectric conversion units. If temperature detection is performed on each photoelectric conversion unit, a photovoltaic power station will be equipped with 60,000 temperature detection modules. First, the amount of temperature detection is large; second, the cost of temperature detection modules and wiring is high; third, the sampling data of 60,000 temperature detection modules is used for training, resulting in excessively large parameters, difficult model convergence, and a large amount of calculation, which leads to a geometric increase in computing cost.

[0081] This disclosure proposes a method for predicting light power for a photovoltaic power station, which is applied to the field of green electricity hydrogen production. The method includes:

[0082] Acquire the current climate and environmental data of the photovoltaic power station in real time, including altitude data, slope data, time data, temperature data, humidity data, wind speed data, and light intensity data;

[0083] The environmental data are grouped according to the climate and environmental data of slope land and the climate and environmental data of flat land;

[0084] Preprocessing the slope climate environment data and the flat land climate environment data to obtain a first data set and a second data set aligned in time, wherein the first data set includes the slope environment data and the second data set includes the flat land environment data;

[0085] The optical power of the photovoltaic power station is predicted based on the data group 1 and the data group 2.

[0086] Since the heat dissipation and cleanliness of photovoltaic power stations on slopes and flat land are different, the data of slopes and flat lands are divided into Shanghai 2024-11487 (M240487CNI)

[0087] The photovoltaic stations are grouped and light power predictions are performed separately, which improves the prediction accuracy.

[0088] Specifically, in the embodiments of this application, by distinguishing climate and environmental data for sloped and flat areas, and performing data preprocessing and power generation prediction separately, it is possible to more accurately reflect the power generation performance of photovoltaic power stations under different terrain conditions. This is because the differences in topography and landforms between sloped and flat areas lead to different climatic conditions such as light and wind speed, which in turn affect the power generation efficiency of photovoltaic panels. By grouping the data, it is possible to more finely analyze and predict the power generation under different terrain conditions, improving the accuracy of the prediction.

[0089] Furthermore, by acquiring, grouping, and preprocessing climate and environmental data in real time, the present invention enables the prediction model to adapt to varying environmental conditions. This adaptability is crucial for maintaining stable operation of photovoltaic power plants in changing climates, helping to improve the overall power generation efficiency and economic benefits of the power plant.

[0090] It’s worth noting that accurate solar power forecasting can help power plant operators better plan and dispatch resources. For example, they can use forecast results to rationally schedule maintenance and upgrades for PV plants, minimizing losses from unexpected power outages or power generation shortfalls.

[0091] For example, if a forecast indicates that high wind speeds on a sloped area may result in lower power generation from photovoltaic panels, while flat areas will experience good sunlight, maintenance personnel can adjust the panels' angles or clean them in advance to improve power generation efficiency. Furthermore, based on these forecasts, power plants can communicate with the grid company in advance to optimize power generation plans and ensure a stable power supply. This approach allows power plants to more effectively utilize limited resources, improving power generation efficiency and economic benefits.

[0092] Furthermore, in some possible implementations, the preprocessing of the slope climate environment data and the flat land climate environment data includes:

[0093] Data cleaning, processing missing values ​​and outliers;

[0094] Data standardization, standardize or normalize the cleaned data;

[0095] Feature extraction, extracting target feature data from the standardized or normalized photovoltaic data and environmental data, wherein the target feature data includes time features, meteorological features, and geographical features;

[0096] Data alignment: align the target feature data of the slope and the target feature data of the flat land by time.

[0097] Process the data to facilitate data calculation.

[0098] Specifically, in the embodiment of the present application, missing values ​​and outliers are processed by data cleaning, which can eliminate the data Shanghai 2024-11487 (M240487CNI)

[0099] Errors and inconsistencies in the data set are eliminated, thereby improving the accuracy and reliability of the data. This is crucial for subsequent data analysis and model training, as low-quality data can lead to deviations in analysis results and degraded model performance.

[0100] Specifically, standardizing or normalizing the data can ensure that the scales of different features are consistent. Standardized data can speed up the convergence of the algorithm and improve the predictive ability of the model.

[0101] Specifically, feature extraction is the process of converting raw data into a form more useful for the model. By extracting temporal, meteorological, and geographic features, we can better capture the key factors affecting photovoltaic power generation, thereby improving the accuracy and generalization ability of the prediction model.

[0102] More specifically, aligning the target feature data of slopes and flat lands by time can ensure the consistency of data in different terrains in the time series. This is very important for time series analysis and prediction models because it can ensure that the model can obtain consistent training and prediction results under different terrain conditions.

[0103] For example, consider a photovoltaic power station located in a mountainous area. Its dataset contains climate and environmental data for the past year, including altitude, slope, time of day, temperature, humidity, wind speed, and light intensity. Before employing the aforementioned method, the dataset might contain missing values, outliers, and temporal inconsistencies in the data across different terrains, resulting in poor training results for the power generation prediction model.

[0104] After adopting the above method, the data set can be cleaned first, and missing values ​​and outliers can be processed, such as using mean filling or statistical-based methods to identify and process outliers. Then, the cleaned data can be standardized or normalized to make the scales of different features consistent, such as using minimum and maximum scaling or Z-score standardization. Next, the target feature data is extracted from the standardized or normalized data, including time features (such as hours of the day), meteorological features (such as temperature, humidity, wind speed), and geographical features (such as altitude, slope). Finally, the target feature data of slopes and flat lands are aligned in time to ensure the consistency of data of different terrains in the time series. Through this series of data preprocessing steps, the accuracy and reliability of photovoltaic power station power generation prediction can be significantly improved.

[0105] Furthermore, in some possible implementations, performing optical power prediction of a photovoltaic power station based on the first and second data sets includes:

[0106] Inputting the slope optical power prediction model and the flat land optical power prediction model corresponding to the data group 1 and the data group 2 respectively to obtain the slope optical power prediction value and the flat land optical power prediction value;

[0107] Shanghai 2024-11487(M240487CNI)

[0108] The theoretical predicted value of the total optical power of the photovoltaic power station is calculated based on the predicted value of the optical power on the slope and the predicted value of the optical power on the flat land.

[0109] That is, different training models are used to predict grouped data, which improves the prediction accuracy of input parameters for slopes and flat lands.

[0110] Specifically, by establishing separate solar power prediction models for slopes and flat areas, we can more accurately capture the impact of different terrain types on photovoltaic power generation. This approach considers the impact of terrain changes on climatic conditions such as light and temperature, thereby improving prediction accuracy.

[0111] Different terrains have varying effects on the power generation efficiency of photovoltaic panels. Therefore, by performing separate predictions for both sloping and flat land, the model can better adapt to photovoltaic power plants in diverse terrain conditions, improving the generalization of the prediction model. Accurate solar power predictions help power plant operators more effectively plan and dispatch resources. For example, based on the prediction results, maintenance and upgrades of photovoltaic power plants can be rationally scheduled, reducing losses caused by unexpected power outages or power generation shortages.

[0112] It is worth noting that photovoltaic power generation is intermittent and fluctuating, and large-scale photovoltaic power generation connected to the power grid will affect the stable operation of the power grid. However, in the embodiments of this application, accurate power prediction can reasonably arrange the operation mode and response measures of the power grid, thereby improving the safety and reliability of the power grid.

[0113] For example, power plant operators can create separate optical power prediction models for sloped and flat areas. First, they collect climate and environmental data for both slopes and flat areas, including altitude, slope gradient, time of day, temperature, humidity, wind speed, and light intensity. They then input this data into the corresponding prediction models to obtain optical power predictions for both slopes and flat areas. Finally, they calculate the theoretical total optical power prediction based on these two predictions.

[0114] Furthermore, in some possible implementations, the method further includes:

[0115] Inputting the optical power prediction value of the flat land optical power prediction model into the slope land optical power prediction model, and based on the input optical power prediction value of the flat land optical power prediction model, the slope land optical power prediction model outputs a first deviation value, where the first deviation value is the difference between the slope land optical power prediction value and the flat land optical power prediction value;

[0116] The optical power prediction value of the slope optical power prediction model is input into the flat land optical power prediction model. Based on the input optical power prediction value of the slope optical power prediction model, the flat land optical power prediction model is

[0117] The prediction model outputs a second deviation value, where the second deviation value is the difference between the flat land optical power prediction value and the optical power prediction value of the slope land optical power prediction model;

[0118] Correcting the slope optical power prediction value according to the first deviation value to obtain a first corrected slope optical power prediction value;

[0119] Correcting the predicted value of the flat-ground optical power according to the second deviation value to obtain a first corrected value of the predicted value of the flat-ground optical power;

[0120] The first revised value of the predicted light power of the photovoltaic power station is calculated based on the first revised value of the predicted light power of the slope and the first revised value of the predicted light power of the flat.

[0121] Specifically, in this embodiment, by inputting the predicted optical power values ​​for flat land into the optical power prediction model for sloped land, and vice versa, the deviation between the two values ​​can be calculated. This two-way correction method can more accurately capture the difference in power generation between sloped and flat land, thereby improving overall prediction accuracy.

[0122] By calculating the first and second deviation values ​​and adjusting the optical power predictions for slopes and flat areas accordingly, the prediction model can be better adapted to different terrain conditions. This approach helps improve the model's robustness in complex terrain, enabling it to provide reliable predictions under diverse environmental conditions.

[0123] It’s worth noting that accurate optical power prediction is crucial for the operation of photovoltaic power plants and the scheduling of power grids. This method can more effectively plan power plant generation plans, rationally arrange power grid operations, and improve grid security and reliability.

[0124] For example, power plant operators can establish separate optical power prediction models for sloped and flat areas. They then input the predicted values ​​for the flat areas into the sloped area model and calculate a first deviation value, which reflects the prediction error of the sloped area model when considering the flat area conditions. Similarly, they also input the predicted values ​​for the sloped area into the flat area model and calculate a second deviation value, which reflects the prediction error of the flat area model when considering the sloped area conditions. Next, the operators correct the predicted values ​​for the sloped and flat areas based on these two deviation values ​​to obtain the corrected predicted values. Finally, they calculate the total optical power prediction value for the entire power plant based on these corrected predicted values. For example, if the first deviation value shows that the actual generated power on the sloped areas is generally higher than predicted by the flat area model, then they can adjust the predicted value for the sloped areas accordingly to reflect this difference. In this way, the power plant can more accurately predict the generated power under different terrain conditions, thereby more effectively Shanghai 2024-11487(M240487CNI)

[0125] Local planning of power generation and grid dispatching.

[0126] Furthermore, in some possible implementations, the slope optical power prediction value is corrected according to the first deviation value to obtain a first correction value of the slope optical power prediction value, including: the first correction value of the slope optical power prediction value = the slope optical power prediction value + K1*first deviation value, K1 is the first correction coefficient; optionally, K1 is a value between 0.1 and 0.2, that is, K1 is a positive value, and the heat dissipation condition of the slope is better than that of the flat land, so it is considered to make a positive correction to its prediction value.

[0127] The first correction value of the flat ground optical power prediction value is obtained by correcting the flat ground optical power prediction value according to the second deviation value, including: the first correction value of the flat ground optical power prediction value = the flat ground optical power prediction value + K2*the second deviation value, K2 is the second correction coefficient; optionally, K2 is a value between -0.1 and -0.2, that is, K2 is a negative value, and the heat dissipation condition of the flat ground is worse than that of the slope, so it is considered to make a negative correction to its predicted value.

[0128] The first corrected value of the predicted light power of the photovoltaic power station is calculated based on the first corrected value of the predicted light power of the slope and the first corrected value of the predicted light power of the flat land, including: summing the first corrected value of the predicted light power of the slope and the first corrected value of the predicted light power of the flat land to obtain the first corrected value of the predicted light power of the photovoltaic power station.

[0129] In this embodiment, correction coefficients K1 and K2 are introduced to modify the predicted optical power values ​​for slopes and flat areas, which can more accurately reflect the actual generated power. This method allows the model to adjust the predicted values ​​based on historical data and actual conditions, thereby improving the accuracy of the prediction. At the same time, the correction coefficients K1 and K2 can be adjusted according to different time periods, climate conditions, and terrain characteristics, making the prediction model adaptable to various environmental changes and improving the model's generalization ability.

[0130] For example, power plant operators create optical power prediction models for sloped and flat land. They then input the predicted values ​​for the flat land into the sloped land model and calculate a first deviation, which reflects the prediction error of the sloped land model when considering the flat land conditions. Similarly, they input the predicted values ​​for the sloped land into the flat land model and calculate a second deviation, which reflects the prediction error of the flat land model when considering the sloped land conditions.

[0131] The method further comprises:

[0132] Obtain the historical forecast data of flat land light power, historical forecast data of slope land light power, and photovoltaic power station light power Shanghai 2024-11487 (M240487CNI) with the same time characteristics, meteorological characteristics and geographical characteristics as the data group 1 and data group 2

[0133] Historical prediction data, wherein the same means that the deviation is within a preset range, and the optical power historical prediction data is the optical power theoretical prediction value data whose deviation between the optical power theoretical prediction value and the optical power actual value is within the preset range;

[0134] Comparing the first corrected value of the slope land optical power prediction value of the photovoltaic power station with the slope land optical power historical prediction data to obtain a third deviation value, and comparing the first corrected value of the flat land optical power prediction value of the photovoltaic power station with the flat land optical power historical prediction data to obtain a fourth deviation value;

[0135] Determine the second correction value of the slope land optical power prediction value according to the third deviation value, and determine the second correction value of the flat land optical power prediction value according to the fourth deviation value,

[0136] The second revised value of the predicted light power value of the photovoltaic power station is calculated according to the second revised value of the predicted light power value on the slope and the second revised value of the predicted light power value on the flat land;

[0137] Comparing the second correction value of the predicted light power value of the photovoltaic power station with the historical predicted light power data of the photovoltaic power station to obtain a fifth deviation value;

[0138] When the fifth deviation value is less than or equal to the set deviation value, the theoretical predicted value of the total optical power of the photovoltaic power station is output;

[0139] If the fifth deviation value is greater than the set deviation value, the second correction value of the optical power prediction value is corrected to obtain a third correction value of the optical power prediction value, and the third correction value of the optical power prediction value is output as the theoretical prediction value of the total optical power of the photovoltaic power station.

[0140] In this embodiment, by incorporating historical prediction data to calculate deviation values, the performance of the current prediction model can be more accurately evaluated. This method leverages the statistical properties of historical data, helping to improve prediction accuracy. By calculating the third and fourth deviation values ​​and adjusting the optical power prediction values ​​for slopes and flat areas accordingly, the prediction model can better adapt to the changing trends of historical data and enhance the model's self-correction capabilities.

[0141] For example: Power station operation and maintenance personnel can establish optical power prediction models for slopes and flat lands respectively. Then, they collect historical prediction data with the same time characteristics, meteorological characteristics and geographical characteristics as these data groups, and ensure that the deviation of these historical data is within the preset range. Next, they compare the current slope optical power prediction value with the historical slope optical power prediction data to obtain a third deviation value, and correct the slope optical power prediction value accordingly. Similarly, they also correct the flat land optical power prediction value. Finally, they calculate the total optical power prediction value of the entire power station based on the corrected slope and flat land optical power prediction values, and compare it with the historical power station optical power prediction data to obtain the fifth deviation value. If the fifth deviation value is within the preset Shanghai 2024-11487 (M240487CNI)

[0142] If the deviation is outside the preset range, they will further revise the prediction value until a satisfactory accuracy is achieved.

[0143] Furthermore, in some possible implementations, the above method further includes:

[0144] Obtain actual optical power data during the forecast period;

[0145] The actual optical power data is compared with the theoretical predicted value of the total optical power of the photovoltaic power station to obtain a sixth deviation value. If the sixth deviation value is less than or equal to the set deviation value, the third correction value of the optical power prediction value is stored as historical data; if the sixth deviation value is greater than the set deviation value, the slope optical power prediction model is optimized according to the first deviation value and the third deviation value; and the flat land optical power prediction model is optimized according to the second deviation value and the fourth deviation value.

[0146] By coupling and correcting the slope prediction value and the flat land prediction value, the mutual influence between photovoltaic power generation power at different locations is taken into account, and the overall prediction accuracy of the photovoltaic power station is improved.

[0147] In the embodiment of the present application, the accuracy of the prediction model can be quantified by obtaining actual optical power data and comparing it with the theoretical prediction value. If the sixth deviation value is within the preset range, it means that the prediction accuracy of the model is acceptable and it can be stored as historical data for calibration and improvement of future models. If the sixth deviation value is greater than the set deviation value, it means that the prediction accuracy of the current model is insufficient and needs further optimization. By comparing the first deviation value with the third deviation value, and the second deviation value with the fourth deviation value, the deficiencies of the model in the prediction of slopes and flat areas can be identified, thereby optimizing the optical power prediction models of slopes and flat areas in a targeted manner.

[0148] It is worth noting that the model optimization process takes actual power generation data into account, enabling the model to better adapt to changes in actual environmental conditions, improving the model's adaptability and robustness. By continuously optimizing the prediction model, the accuracy of photovoltaic power station power generation prediction can be improved, thereby helping power grid companies to more effectively dispatch and allocate resources, and improving the safety and reliability of the power grid.

[0149] For example: The power station operation and maintenance personnel first establish optical power prediction models for slopes and flat lands respectively and perform predictions. Then, they collect actual optical power data and compare it with the theoretical prediction value to obtain the sixth deviation value. If this deviation value is within the preset range, it means that the prediction accuracy of the model is acceptable, and these prediction values ​​can be stored as historical data. If the deviation value exceeds the preset range, they will optimize the slope optical power prediction model based on the first and third deviation values, and optimize the slope optical power prediction model based on the second and third deviation values.

[0150] Four deviation values ​​are used to optimize the flatland optical power prediction model. For example, if analysis reveals that the slope model has low prediction accuracy in cloudy weather, they might introduce additional meteorological factors, such as cloud cover and wind speed, to improve the model. This approach allows power plants to continuously improve the accuracy of power generation forecasts, leading to more efficient power generation planning and grid scheduling.

[0151] Furthermore, in some possible implementations, the photovoltaic power station is provided with a plurality of photovoltaic strings, each of the photovoltaic strings includes a plurality of photovoltaic panels arranged in series, and the plurality of photovoltaic panels are connected in parallel to a combiner box;

[0152] The obtaining of the current climate and environmental data of the photovoltaic power station includes obtaining environmental data of each photovoltaic string;

[0153] Grouping the environmental data according to slope climate environmental data and flat land climate environmental data includes:

[0154] The photovoltaic strings with wind speed less than or equal to the first preset value in the photovoltaic power station are divided into the slope low wind group within the station;

[0155] The photovoltaic strings in the photovoltaic power station with a wind speed of the first preset value < ≦ the second preset value are divided into the slope and moderate wind group within the station;

[0156] The photovoltaic strings with wind speed greater than the second preset value in the photovoltaic power station are divided into the high wind group on the slope within the station;

[0157] The data of each photovoltaic string in the station slope low wind group, the station slope medium wind group and the station slope high wind group include photovoltaic data and environmental data;

[0158] Grouping the photovoltaic data and environmental data according to slope data and flat land data includes:

[0159] The photovoltaic strings with wind speed less than or equal to the fourth preset value in the photovoltaic power station are divided into the flat and gentle wind group within the station;

[0160] The photovoltaic strings with wind speed greater than the third preset value in the photovoltaic power station are divided into the flat ground and wind groups within the station.

[0161] In the embodiment of the present application, by dividing the photovoltaic strings of the photovoltaic power station into different groups according to the wind speed, more refined environmental data management and power generation prediction can be performed on the photovoltaic strings of each group. This grouping method helps to identify and analyze the performance of photovoltaic panels under different wind speed conditions, thereby improving the management efficiency and power generation efficiency of the power station. By grouping the photovoltaic strings under different wind speed conditions, the power generation under specific wind speed conditions can be more accurately predicted. This is because wind speed is one of the important factors affecting the efficiency of photovoltaic power generation, especially under different terrain conditions of slopes and flat lands, the impact of wind speed on photovoltaic panels may be different. Based on this, through grouping and refined management, power station operators can allocate resources more specifically, such as taking different maintenance and optimization measures for slope areas with higher wind speeds to improve power generation efficiency.

[0162] For example, if the forecast shows that in a slope area with high wind speed, the power generation efficiency of the photovoltaic string may be Shanghai 2024-11487 (M240487CNI)

[0163] If wind speed affects power generation efficiency, operations and maintenance personnel can take proactive measures, such as adjusting the angle of the photovoltaic panels or adding support structures, to reduce the impact of wind speed on power generation efficiency. Furthermore, based on these forecasts, power plants can communicate with the grid company in advance to rationalize power generation plans and ensure a stable power supply. This approach allows power plants to more effectively utilize limited resources, improving power generation efficiency and economic benefits.

[0164] Furthermore, in some possible implementations, the slope optical power prediction model and the flat land optical power prediction model respectively include multiple parallel DNN neural network modules, and the data of the slope low wind group, slope medium wind group and slope high wind group of each photovoltaic string within the statistical time are respectively input into the independent DNN neural network module of the slope optical power prediction model, and the data of the flat land slow wind group and the flat land and wind group within the station are respectively input into the independent DNN neural network module of the flat land optical power prediction model, and the output power generation of each DNN neural network module of the slope optical power prediction model and the flat land optical power prediction model are respectively accumulated to obtain the slope optical power prediction value and the flat land optical power prediction value.

[0165] In the embodiments of the present application, by using a deep neural network (DNN) model, the complex nonlinear relationships in the climate and environmental data of photovoltaic power stations can be captured, thereby improving the accuracy of light power prediction. Due to its deep structure, the DNN model can learn high-level features in the data, which is particularly important for predicting time series data such as photovoltaic power generation that is affected by multiple factors. Multiple parallel DNN neural network modules can independently learn and extract data features of different photovoltaic strings, and then output the power generation power by accumulating them. Such a structure helps to improve the model's generalization ability for photovoltaic power generation behavior under different conditions.

[0166] For example, power plant operators can first obtain environmental data for each photovoltaic string, including wind speed, temperature, and humidity. Then, based on wind speed, they categorize the photovoltaic strings into groups: low wind on slopes within the station, moderate wind on slopes within the station, high wind on slopes within the station, slow wind on flat land within the station, and moderate wind on flat land within the station. Each group of photovoltaic strings includes both photovoltaic and environmental data. These data are then fed into corresponding DNN neural network modules. For example, data from the low wind on slopes within the station is fed into a DNN module in a sloped land solar power prediction model, while data from the slow wind on flat land within the station is fed into a DNN module in a flat land solar power prediction model. This allows each DNN module to specifically process data under specific wind speed conditions, resulting in more accurate predictions of generated power under those conditions. Finally, the outputs of all DNN modules are summed to obtain the predicted solar power value for the entire photovoltaic power station. This approach significantly improves the power generation prediction accuracy of photovoltaic power stations by leveraging the powerful learning and multi-level feature extraction capabilities of DNNs.

[0167] Accuracy and reliability.

[0168] Furthermore, an embodiment of the present application also provides an optical power prediction system suitable for a photovoltaic power station, including a photovoltaic power station control room controller and multiple photovoltaic strings. The photovoltaic power station control room controller establishes a data interaction link with each photovoltaic string in the photovoltaic power station through wired or wireless communication methods.

[0169] like Figure 3 As shown, multiple photovoltaic panels of a photovoltaic power station are connected in series to form a photovoltaic string, and multiple photovoltaic strings are connected in parallel to a junction box. For example, 20 photovoltaic panels are connected in series to form a photovoltaic string. A photovoltaic power station has 1,000 photovoltaic panels, so a photovoltaic power station has 50 photovoltaic strings.

[0170] Optionally, a photovoltaic power station has one or more combiner boxes.

[0171] like Figure 4 and Figure 5 As shown, a method for predicting optical power of a photovoltaic power station includes the following steps:

[0172] S1. The controller in the centralized control room of the photovoltaic power station obtains data of multiple photovoltaic strings of each photovoltaic power station;

[0173] It can be understood that multiple photovoltaic panels of the photovoltaic power station are connected in series in sequence to form a photovoltaic string; multiple photovoltaic strings are connected in parallel to the combiner box 4.

[0174] For slopes:

[0175] Preferably, a wind speed meter 1 is provided on the head photovoltaic panel of at least one photovoltaic string, and a thermometer 2 and a light radiation intensity meter 3 are provided on the tail photovoltaic panel;

[0176] Preferably, a thermometer 2 and a light radiation intensity meter 3 are provided on the head photovoltaic panel of at least one photovoltaic string, and a wind speed meter 1 is provided on the tail photovoltaic panel;

[0177] Preferably, a wind speed meter 1, a thermometer 2 and a light radiation intensity meter 3 are provided on the head end photovoltaic panel of at least one photovoltaic string;

[0178] Preferably, a wind speed meter 1, a thermometer 2 and a light radiation intensity meter 3 are provided on the tail photovoltaic panel of at least one photovoltaic string;

[0179] Preferably, an anemometer 1 is provided on an intermediate photovoltaic panel of at least one photovoltaic string, and a thermometer 2 and a light radiation intensity meter 3 are provided on another intermediate photovoltaic panel; that is, an anemometer 1, a thermometer 2 and a light radiation intensity meter 3 are provided on adjacent or non-adjacent intermediate photovoltaic panels of the photovoltaic string, respectively, that is, an anemometer 1 is provided on one intermediate photovoltaic panel, and a thermometer 2 and a light radiation intensity meter 3 are provided on another intermediate photovoltaic panel at the same time;

[0180] Shanghai 2024-11487(M240487CNI)

[0181] The slope data includes the wind speed measured by the anemometer 1 of each photovoltaic string, the temperature measured by the thermometer 2, the light radiation intensity measured by the light radiation intensity meter 3, and the power generated by the photovoltaic string;

[0182] For slopes, wind speed, solar radiation, heat dissipation, and temperature vary greatly at different altitudes and slopes. Therefore, a unified meteorological temperature is not used. Instead, each photovoltaic string independently measures wind speed, temperature, and solar radiation intensity. In addition, for slopes, by collecting wind speed and temperature for each photovoltaic string, different wind speeds have already comprehensively reflected the cleanliness and heat dissipation conditions of the photovoltaic panels, so there is no need to measure the cleanliness of the photovoltaic panels or photovoltaic strings on the slope.

[0183] For flat land:

[0184] Preferably, a wind speed meter 1 is provided on the head photovoltaic panel of at least one photovoltaic string, and a light radiation intensity meter 3 is provided on the tail photovoltaic panel;

[0185] Preferably, a light radiation intensity meter 3 is provided on the head photovoltaic panel of at least one photovoltaic string, and a wind speed meter 1 is provided on the tail photovoltaic panel;

[0186] Preferably, a wind speed meter 1 and a light radiation intensity meter 3 are provided on the head end photovoltaic panel of at least one photovoltaic string;

[0187] Preferably, a wind speed meter 1 and a light radiation intensity meter 3 are provided on the tail photovoltaic panel of at least one photovoltaic string;

[0188] Preferably, a wind speed measuring instrument 1 is provided on one middle photovoltaic panel of at least one photovoltaic string, and a light radiation intensity meter 3 is provided on another middle photovoltaic panel; that is, a wind speed measuring instrument 1 and a light radiation intensity meter 3 are provided on adjacent or non-adjacent middle photovoltaic panels of the photovoltaic string, respectively, that is, the wind speed measuring instrument 1 is provided on one middle photovoltaic panel, and the light radiation intensity meter 3 is provided on another middle photovoltaic panel;

[0189] Optionally, a weighing unit 5 is provided on the top of the combiner box 4 on the flat ground, and the weighing unit 5 is used to measure the weight of dust and impurities to centrally map the cleanliness of the photovoltaic panels on the flat ground;

[0190] The mapping is:

[0191] If the weight measured by weighing unit 5 is less than 10 grams, the cleanliness factor of the photovoltaic panel on flat ground is in the range of 0.9-1;

[0192] The weight measured by the weighing unit 5 is between 10 and 30 grams, and the cleanliness factor of the photovoltaic panel on flat ground is between 0.7 and 0.9.

[0193] Shanghai 2024-11487(M240487CNI)

[0194] The weight measured by weighing unit 5 is between 30-50 grams, and the cleanliness factor of the photovoltaic panel on flat ground is in the range of 0.6-0.7;

[0195] If the weight measured by weighing unit 5 is greater than 50 grams, the cleanliness factor of the photovoltaic panel on flat ground is in the range of 0.4-0.6;

[0196] Optionally, the cleanliness coefficient obtained by mapping the weighing unit 5 is used as the cleanliness coefficient of each photovoltaic string connected to the combiner box 4 on the flat ground, and the cleanliness is only measured at the combiner box 4 on the flat ground, avoiding the equipment complexity and data complexity of multiple measurements.

[0197] Optionally, on flat land, the controller in the photovoltaic power station control room obtains the meteorological temperature from the meteorological department. Because there is no difference between the windward and leeward sides of the slope on flat land, the heat dissipation and cleanliness conditions on flat land are the same. Therefore, there is no need to measure the temperature of each photovoltaic string independently, but only need to use the same meteorological temperature.

[0198] Optionally, the flat land data includes the wind speed measured by the anemometer 1 of each photovoltaic string, the meteorological temperature, the light radiation intensity measured by the light radiation intensity meter 3, the cleanliness factor measured by the junction box 4 at the flat land where the photovoltaic string is connected, and the photovoltaic string power generation power.

[0199] It can be understood that slopes are hills, sand dunes or hills, and flat lands are plains or flat areas with small elevation differences.

[0200] S2, the controller in the photovoltaic power station control room groups the acquired data;

[0201] For slopes:

[0202] The photovoltaic strings with wind speed ≤ 1 m / s in the photovoltaic power station are divided into the low-wind group on the slope within the station;

[0203] The photovoltaic strings in the photovoltaic power station with wind speed of 1 m / s < ≦ 5 m / s are divided into the slope and moderate wind group within the station;

[0204] The photovoltaic strings in the photovoltaic power station with wind speed greater than 5 m / s are divided into the high wind group on the slope within the station;

[0205] The data of each PV string in the low-wind group, medium-wind group, and high-wind group on the slopes within the station include wind speed, temperature, solar radiation intensity, and PV string power generation;

[0206] For flat land:

[0207] The photovoltaic strings with wind speeds of ≤5 m / s in the photovoltaic power station are divided into the flat and gentle wind group within the station;

[0208] The photovoltaic strings in the photovoltaic power station with wind speed greater than 5 m / s are divided into flat ground and wind groups within the station;

[0209] The data of each PV string in the flat and gentle wind group within the station includes wind speed, meteorological temperature, light radiation intensity, cleanliness coefficient and PV string power generation;

[0210] Shanghai 2024-11487(M240487CNI)

[0211] The data of each PV string in the flat land and wind group within the station includes wind speed, meteorological temperature, solar radiation intensity and PV string power generation;

[0212] For the flat and slow-wind group within the station with a wind speed of ≤5 m / s, the dust and impurity diffusion conditions are poor, so the cleanliness of its photovoltaic panels has a greater impact on the power generation of the photovoltaic panels and their photovoltaic strings. Therefore, the cleanliness coefficient of the flat and slow-wind group within the station is collected to consider the impact of dust and impurities on the power generation of the photovoltaic panels and improve the accuracy of the photovoltaic power generation.

[0213] S3, the controller in the photovoltaic power station control room trains the grouped data to obtain the optical power prediction value;

[0214] The controller in the centralized control room of the photovoltaic power station contains multiple parallel DNN neural network modules (which can be DNN neural network modules one, two, three, four and five). The data of the station slope low wind group, station slope medium wind group, station slope high wind group, station flat land slow wind group and station flat land moderate wind group within the statistical time are input into independent DNN neural network modules respectively, and the output power generation of each DNN neural network module is accumulated to obtain the optical power prediction value.

[0215] The method of the present application avoids the existing differences in operating temperature and surface cleanliness of photovoltaic panels on slopes and flat lands, which results in differences in the power generation of different photovoltaic panels, and the deviation caused by prediction using a unified training model, resulting in inaccurate prediction values. The method of the present application finely divides and groups the power generation units of the photovoltaic power station based on slopes, flat lands and different wind speeds, with photovoltaic strings as units, and trains each group separately, thereby balancing the effects of differences in photovoltaic panel heating, cleanliness, heat dissipation and light radiation intensity on the power generation, and improving the prediction accuracy of the power generation.

[0216] Optionally, the controller in the photovoltaic power station control room is connected to each photovoltaic string in the photovoltaic power station for communication and data exchange, such as network communication or optical communication.

[0217] Optionally, the optical power prediction value may be 10 days or 4 hours.

[0218] Optionally, the above-mentioned DNN neural network can also be replaced by other machine learning models.

[0219] A light power prediction system for a photovoltaic power station includes a photovoltaic power station control room controller and multiple photovoltaic strings. The photovoltaic power station control room controller communicates with each photovoltaic string in the photovoltaic power station and exchanges data.

[0220] Multiple photovoltaic panels in a photovoltaic power station are connected in series to form a photovoltaic string; multiple photovoltaic strings are connected in parallel to a combiner box.

[0221] For slopes:

[0222] Shanghai 2024-11487(M240487CNI)

[0223] A wind speed meter 1 is provided on the head photovoltaic panel of at least one photovoltaic string, and a thermometer 2 and a light radiation intensity meter 3 are provided on the tail photovoltaic panel;

[0224] A thermometer 2 and a light radiation intensity meter 3 are provided on the head photovoltaic panel of at least one photovoltaic string, and a wind speed meter 1 is provided on the tail photovoltaic panel;

[0225] At least one photovoltaic string is provided with a wind speed measuring instrument 1, a thermometer 2 and a light radiation intensity meter 3 at the head end photovoltaic panel;

[0226] At least one photovoltaic string is provided with a wind speed measuring instrument 1, a thermometer 2 and a light radiation intensity meter 3 at the tail photovoltaic panel;

[0227] At least one intermediate photovoltaic panel of a photovoltaic string is provided with an anemometer 1, and another intermediate photovoltaic panel is provided with a thermometer 2 and a light radiation intensity meter 3; that is, the anemometer 1, the thermometer 2 and the light radiation intensity meter 3 are respectively provided on adjacent or non-adjacent intermediate photovoltaic panels of the photovoltaic string, that is, the anemometer 1 is provided on one intermediate photovoltaic panel, and the thermometer 2 and the light radiation intensity meter 3 are provided on the other intermediate photovoltaic panel at the same time;

[0228] The slope data includes the wind speed measured by the anemometer 1 of each photovoltaic string, the temperature measured by the thermometer 2, the light radiation intensity measured by the light radiation intensity meter 3, and the power generated by the photovoltaic string.

[0229] For flat land:

[0230] A wind speed meter 1 is provided on the photovoltaic panel at the head end of at least one photovoltaic string, and a light radiation intensity meter 3 is provided on the photovoltaic panel at the tail end;

[0231] A light radiation intensity meter 3 is provided on the photovoltaic panel at the head end of at least one photovoltaic string, and a wind speed meter 1 is provided on the photovoltaic panel at the tail end;

[0232] A wind speed meter 1 and a light radiation intensity meter 3 are provided on the head end photovoltaic panel of at least one photovoltaic string;

[0233] At least one photovoltaic string is provided with a wind speed meter 1 and a light radiation intensity meter 3 at the tail photovoltaic panel;

[0234] At least one middle photovoltaic panel of a photovoltaic string is provided with a wind speed measuring instrument 1, and another middle photovoltaic panel is provided with a light radiation intensity meter 3; that is, the wind speed measuring instrument 1 and the light radiation intensity meter 3 are provided on adjacent or non-adjacent middle photovoltaic panels of the photovoltaic string, that is, the wind speed measuring instrument 1 is provided in a middle Shanghai 2024-11487 (M240487CNI)

[0235] On the middle photovoltaic panel, the light radiation intensity meter 3 is set on another middle photovoltaic panel;

[0236] A weighing unit 5 is provided on the top of the combiner box 4 on the flat ground. The weighing unit 5 is used to measure the weight of dust and impurities to centrally map the cleanliness of the photovoltaic panels on the flat ground.

[0237] The mapping is:

[0238] If the weight measured by weighing unit 5 is less than 10 grams, the cleanliness factor of the photovoltaic panel on flat ground is in the range of 0.9-1;

[0239] The weight measured by the weighing unit 5 is between 10 and 30 grams, and the cleanliness factor of the photovoltaic panel on flat ground is between 0.7 and 0.9.

[0240] The weight measured by weighing unit 5 is between 30-50 grams, and the cleanliness factor of the photovoltaic panel on flat ground is in the range of 0.6-0.7;

[0241] If the weight measured by weighing unit 5 is greater than 50 grams, the cleanliness factor of the photovoltaic panel on flat ground is in the range of 0.4-0.6;

[0242] On flat land, the controller in the photovoltaic power station's centralized control room obtains meteorological temperature from the meteorological department. Because there is no difference between the windward and leeward sides of a slope on flat land, the heat dissipation and cleanliness conditions on flat land are the same. Therefore, there is no need to measure the temperature of each photovoltaic string independently, but only need to use the same meteorological temperature.

[0243] The flat land data includes the wind speed measured by the anemometer 1 of each photovoltaic string, the meteorological temperature, the light radiation intensity measured by the light radiation intensity meter 3, the cleanliness factor measured by the junction box 4 at the flat land where the photovoltaic string is connected, and the photovoltaic string power generation.

[0244] Model selection:

[0245] A variety of machine learning algorithms can be used for prediction, such as Support Vector Machine (SVM), Random Forest, Gradient Boosting Trees, or deep learning methods such as Long Short-Term Memory (LSTM) networks.

[0246] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can make their own decisions for each specific application. Shanghai 2024-11487 (M240487CNI)

[0247] Different methods may be used to implement the described functionality, but such implementation should not be considered beyond the scope of the present disclosure.

[0248] It should be noted that the serial numbers of the embodiments of the present disclosure are only for description and do not represent the advantages or disadvantages of the embodiments.

[0249] The above are only preferred embodiments of the present disclosure and are not intended to limit the patent scope of the present disclosure. Any equivalent structure or equivalent process transformation made using the contents of the present disclosure and the drawings, or directly or indirectly applied in other related technical fields, and any combination of various embodiments or schemes, are also included in the patent protection scope of the present disclosure.

Claims

1. A method for predicting light power of photovoltaic power stations, applied in the field of green electricity hydrogen production, characterized in that: The method comprises: Acquire climate and environmental data of the photovoltaic power station in real time, including altitude data, slope data, time data, temperature data, humidity data, wind speed data, and light intensity data; Grouping the climate and environment data according to slope climate and environment data and flat land climate and environment data; Preprocessing the slope climate environment data and the flat land climate environment data to obtain a first data set and a second data set aligned in time, wherein the first data set includes the slope environment data and the second data set includes the flat land environment data; Performing power generation prediction based on the data group 1 and the data group 2; The method further comprises: The performing of optical power prediction of the photovoltaic power station based on the data group 1 and the data group 2 includes: Inputting the slope optical power prediction model and the flat land optical power prediction model corresponding to the data group 1 and the data group 2 respectively obtains the slope optical power prediction value and the flat land optical power prediction value; Calculating a theoretical predicted value of total optical power based on the predicted value of optical power on the slope and the predicted value of optical power on the flat land; The method further comprises: Inputting the optical power prediction value of the flat land optical power prediction model into the slope land optical power prediction model, and based on the input optical power prediction value of the flat land optical power prediction model, the slope land optical power prediction model outputs a first deviation value, where the first deviation value is the difference between the slope land optical power prediction value and the flat land optical power prediction value; Inputting the optical power prediction value of the slope optical power prediction model into the flat land optical power prediction model, and based on the input optical power prediction value of the slope optical power prediction model, the flat land optical power prediction model outputs a second deviation value, where the second deviation value is the difference between the flat land optical power prediction value and the optical power prediction value of the slope optical power prediction model; Correcting the slope optical power prediction value according to the first deviation value to obtain a first corrected slope optical power prediction value; Correcting the predicted value of the flat-ground optical power according to the second deviation value to obtain a first corrected value of the predicted value of the flat-ground optical power; The first revised value of the predicted light power of the photovoltaic power station is calculated based on the first revised value of the predicted light power of the slope and the first revised value of the predicted light power of the flat.

2. The method according to claim 1, wherein The data preprocessing of the slope land climate environment data and the flat land climate environment data includes: Data cleaning, processing missing values ​​and outliers; Data standardization, standardize or normalize the cleaned data; Feature extraction, extracting target feature data from the standardized or normalized environmental data, wherein the target feature data includes time features, meteorological features, and geographical features; Data alignment: align the target feature data of the slope and the target feature data of the flat land by time.

3. The method according to claim 1, wherein The method of correcting the slope optical power prediction value according to the first deviation value to obtain a first correction value of the slope optical power prediction value includes: the first correction value of the slope optical power prediction value=the slope optical power prediction value+K1*the first deviation value, K1 being a first correction coefficient; The method of correcting the predicted value of the flat ground optical power according to the second deviation value to obtain a first corrected value of the predicted value of the flat ground optical power includes: the first corrected value of the predicted value of the flat ground optical power = the predicted value of the flat ground optical power + K2*the second deviation value, K2 being a second correction coefficient; The first corrected value of the predicted light power of the photovoltaic power station is calculated based on the first corrected value of the predicted light power of the slope and the first corrected value of the predicted light power of the flat land, comprising: summing the first corrected value of the predicted light power of the slope and the first corrected value of the predicted light power of the flat land to obtain the first corrected value of the predicted light power of the photovoltaic power station.

4. The method according to claim 3, wherein The method further comprises: Obtaining historical forecast data of flat land optical power, historical forecast data of sloped land optical power, and historical forecast data of photovoltaic power station optical power having the same time characteristics, meteorological characteristics, and geographical characteristics as the data group one and the data group two, wherein the historical forecast data of flat land optical power, historical forecast data of sloped land optical power, and historical forecast data of photovoltaic power station optical power having the same time characteristics, meteorological characteristics, and geographical characteristics means that the deviation is within a preset range, and the historical forecast data of optical power is theoretical forecast value data of optical power whose deviation between the theoretical forecast value of optical power and the actual value of optical power is within the preset range; Comparing the first corrected value of the slope land optical power prediction value of the photovoltaic power station with the slope land optical power historical prediction data to obtain a third deviation value, and comparing the first corrected value of the flat land optical power prediction value of the photovoltaic power station with the flat land optical power historical prediction data to obtain a fourth deviation value; Determine the second correction value of the slope land optical power prediction value according to the third deviation value, and determine the second correction value of the flat land optical power prediction value according to the fourth deviation value, The second correction value of the optical power prediction value is calculated based on the second correction value of the optical power prediction value on the slope and the second correction value of the optical power prediction value on the flat land; Comparing the second correction value of the optical power prediction value with the historical optical power prediction data of the photovoltaic power station to obtain a fifth deviation value; When the fifth deviation value is less than or equal to the set deviation value, outputting a theoretical predicted value of the total optical power of the photovoltaic power station; If the fifth deviation value is greater than the set deviation value, the second correction value of the optical power prediction value is corrected to obtain a third correction value of the optical power prediction value, and the third correction value of the optical power prediction value is output as the theoretical prediction value of the total optical power of the photovoltaic power station.

5. The method according to claim 4, wherein The method further comprises: Obtain actual optical power data during the forecast period; The actual optical power data is compared with the theoretical predicted value of the total optical power of the photovoltaic power station to obtain a sixth deviation value. If the sixth deviation value is less than or equal to the set deviation value, the third correction value of the optical power prediction value is stored as historical data; if the sixth deviation value is greater than the set deviation value, the slope optical power prediction model is optimized according to the first deviation value and the third deviation value; and the flat land optical power prediction model is optimized according to the second deviation value and the fourth deviation value.

6. The method according to any one of claims 1 to 5, wherein: The photovoltaic power station is provided with a plurality of photovoltaic strings, each of the photovoltaic strings comprises a plurality of photovoltaic panels arranged in series, and the plurality of photovoltaic panels are connected in parallel to a combiner box; The obtaining of the current climate and environmental data of the photovoltaic power station includes obtaining environmental data of each photovoltaic string; Grouping the environmental data according to slope climate environmental data and flat land climate environmental data includes: The photovoltaic strings with a wind speed less than or equal to a first preset value in the photovoltaic power station are divided into a slope low wind group within the station; The photovoltaic strings in the photovoltaic power station with a wind speed of the first preset value < ≦ the second preset value are divided into the slope and moderate wind group within the station; The photovoltaic strings with wind speed greater than the second preset value in the photovoltaic power station are divided into the high wind group on the slope within the station; The data of each photovoltaic string in the station slope low wind group, the station slope medium wind group and the station slope high wind group include photovoltaic data and environmental data; Grouping the photovoltaic data and environmental data according to slope data and flat land data includes: The photovoltaic strings with wind speed less than or equal to the fourth preset value in the photovoltaic power station are divided into the flat and slow wind group within the station; The photovoltaic strings with wind speed greater than the third preset value in the photovoltaic power station are divided into the flat ground and wind groups within the station.

7. The method according to claim 6, wherein The slope light power prediction model and the flat land light power prediction model respectively include multiple parallel DNN neural network modules. The data of the slope low wind group, the slope medium wind group and the slope high wind group of each photovoltaic string within the statistical time are respectively input into the independent DNN neural network module of the slope light power prediction model. The data of the flat land slow wind group and the flat land and wind group are respectively input into the independent DNN neural network module of the flat land light power prediction model. The output power generation of each DNN neural network module of the slope light power prediction model and the flat land light power prediction model are respectively accumulated to obtain the slope light power prediction value and the flat land light power prediction value.

8. A light power prediction system applicable to a photovoltaic power station, wherein the system executes the method according to any one of claims 1 to 7 when in operation, characterized in that: It includes a photovoltaic power station control room controller and multiple photovoltaic strings. The photovoltaic power station control room controller communicates and exchanges data with each photovoltaic string in the photovoltaic power station.

Images