Intelligent integrated testing method and system for power generation capacity prediction

Abstract

The present disclosure provides an intelligent integrated testing method and system for power generation capacity prediction, which relates to the technical field of power generation capacity prediction. The method includes: obtaining multiple distributed photovoltaic information of a target area; constructing a photovoltaic twin model; generating an environmental data deviation matching library; obtaining multiple matching environmental prediction data; inputting the multiple matching environmental prediction data into a photovoltaic power generation prediction model for power generation prediction, wherein the photovoltaic power generation prediction model is constructed based on the principle of integrated learning and the photovoltaic twin model; respectively predicting the power generation of multiple matching environmental prediction data through multiple photovoltaic power generation prediction sub-channels, and obtaining the photovoltaic power generation prediction power of the target area within a preset time window through weighted calculation. The present disclosure can solve the technical problem in the prior art that the accuracy of power generation capacity prediction is low, resulting in reduced system stability, and achieve the technical effect of improving the accuracy of power generation capacity prediction.

Description

Technical Field

[0001] The present disclosure relates to the technical field of power generation capacity prediction, and in particular to an intelligent integrated testing method and system for power generation capacity prediction. Background Art

[0002] The accuracy of the prediction of the power generation capacity of distributed photovoltaic power generation equipment affects the stable operation and economic benefits of the power system. At present, distributed photovoltaic power generation equipment is greatly affected by the environment, and because some distributed photovoltaic power generation equipment may be distributed in remote mountainous areas, semi-mountainous areas and hilly areas, the environmental parameters of the location of the distributed photovoltaic power generation equipment and the environmental data collected by the operation and management center have a large deviation, which affects the prediction of power generation capacity. At the same time, predicting power generation capacity based on a single algorithm results in a large error in the prediction of power generation capacity, and even causes energy waste and reduced system stability.

[0003] To sum up, there are technical problems in the prior art that the environmental parameters of distributed power generation equipment have a large deviation from the environmental data collected by the operation and management center, resulting in low data collection accuracy and a relatively simple prediction method, which leads to low accuracy in the prediction of the overall power generation capacity of distributed photovoltaic power generation equipment. Summary of the invention

[0004] The present disclosure provides an intelligent integrated testing method and system for power generation capacity prediction, which is used to solve the technical problems in the prior art that the environmental parameters of distributed power generation equipment have a large deviation from the environmental data collected by the operation and management center, resulting in low data collection accuracy and a relatively simple prediction method, resulting in low accuracy in the overall power generation capacity prediction of distributed photovoltaic power generation equipment.

[0005] According to the first aspect of the present disclosure, an intelligent integrated testing method for power generation capacity prediction is provided, including: connecting a distributed photovoltaic operation and management system to obtain multiple distributed photovoltaic information of a target area, wherein the distributed photovoltaic information includes device coordinates and device operation data; based on digital twins, simulation modeling is performed based on the device operation data, and the device coordinates are used as model positioning identifiers to construct a photovoltaic twin model of the target area; based on the location coordinates of a photovoltaic operation and management center, light radiation deviation, temperature deviation and weather deviation analysis of multiple distributed photovoltaic devices are performed based on the photovoltaic twin model to generate an environmental data deviation matching library; environmental prediction data within a preset time window of the photovoltaic operation and management center is collected, and Based on the environmental data deviation matching library, deviation analysis is performed on the environmental prediction data to obtain multiple matching environmental prediction data, wherein the matching environmental prediction data correspond to distributed photovoltaic equipment one by one; the multiple matching environmental prediction data are input into a photovoltaic power generation prediction model to predict power generation, wherein the photovoltaic power generation prediction model is constructed based on the ensemble learning principle and the photovoltaic twin model, and comprises multiple photovoltaic power generation prediction channels, and each photovoltaic power generation prediction channel comprises multiple photovoltaic power generation prediction sub-channels; power generation prediction is performed on the multiple matching environmental prediction data respectively through multiple photovoltaic power generation prediction sub-channels, and weighted calculation is performed to obtain the predicted photovoltaic power generation power of the target area within the preset time window.

[0006] According to a second aspect of the present disclosure, an intelligent integrated test system for power generation capacity prediction is provided, including: a distributed photovoltaic information acquisition module, the distributed photovoltaic information acquisition module is used to connect to a distributed photovoltaic operation and management system to obtain multiple distributed photovoltaic information of a target area, the distributed photovoltaic information includes device coordinates and device operation data; a photovoltaic twin model construction module, the photovoltaic twin model construction module is used to perform simulation modeling based on digital twins and the device operation data, and use the device coordinates as a model positioning identifier to construct a photovoltaic twin model of the target area; an environmental data deviation matching library acquisition module, the environmental data deviation matching library acquisition module is used to perform light radiation deviation, temperature deviation and weather deviation analysis of multiple distributed photovoltaic devices based on the location coordinates of a photovoltaic operation and management center and the photovoltaic twin model, and generate an environmental data deviation matching library; a matching environment prediction data acquisition module, the matching environment prediction data acquisition module The module is used to collect environmental prediction data within a preset time window of the photovoltaic operation and management center, and perform deviation analysis on the environmental prediction data based on the environmental data deviation matching library to obtain multiple matching environmental prediction data, wherein the matching environmental prediction data corresponds to distributed photovoltaic equipment one by one; a photovoltaic power generation prediction channel acquisition module, the photovoltaic power generation prediction channel acquisition module is used to input the multiple matching environmental prediction data into a photovoltaic power generation prediction model for power generation prediction, wherein the photovoltaic power generation prediction model is constructed based on the integrated learning principle and the photovoltaic twin model, and contains multiple photovoltaic power generation prediction channels, and each photovoltaic power generation prediction channel includes multiple photovoltaic power generation prediction sub-channels; a power generation prediction module, the power generation prediction module is used to perform power generation prediction on the multiple matching environmental prediction data respectively through multiple photovoltaic power generation prediction sub-channels, and obtain the photovoltaic power generation prediction power of the target area within the preset time window by weighted calculation.

[0007] One or more technical solutions provided in the present disclosure have at least the following technical effects or advantages: according to the method adopted in the present disclosure, by connecting a distributed photovoltaic operation and management system, multiple distributed photovoltaic information of the target area is obtained, and the distributed photovoltaic information includes device coordinates and device operation data; based on the digital twin, simulation modeling is performed based on the device operation data, and the device coordinates are used as the model positioning identifier to construct a photovoltaic twin model of the target area; based on the location coordinates of the photovoltaic operation and management center, the light radiation deviation, temperature deviation and weather deviation analysis of multiple distributed photovoltaic devices are performed based on the photovoltaic twin model to generate an environmental data deviation matching library; the environmental prediction data within a preset time window of the photovoltaic operation and management center is collected, and the environmental prediction data is subjected to deviation analysis based on the environmental data deviation matching library to obtain multiple matching environmental prediction data, wherein the matching environmental prediction data and the distributed photovoltaic equipment correspond one to one; the multiple matching environmental prediction data are input into A photovoltaic power generation prediction model is used to predict power generation, wherein the photovoltaic power generation prediction model is constructed based on the integrated learning principle and the photovoltaic twin model, and includes multiple photovoltaic power generation prediction channels, and each photovoltaic power generation prediction channel includes multiple photovoltaic power generation prediction sub-channels; the power generation prediction is performed on the multiple matching environmental prediction data respectively through multiple photovoltaic power generation prediction sub-channels, and the photovoltaic power generation prediction power of the target area within the preset time window is obtained by weighted calculation, which solves the technical problems in the prior art that the environmental parameters of the distributed power generation equipment have a large deviation from the environmental data collected by the operation and management center, resulting in low data collection accuracy and a relatively single prediction method, resulting in low accuracy in the prediction of the overall power generation capacity of the distributed photovoltaic power generation equipment, and realizes multi-dimensional targeted prediction of the distributed photovoltaic equipment, reduces the power generation prediction error, and improves the accuracy of the overall power generation prediction of the distributed photovoltaic, thereby improving the technical effect of improving the stability of the distributed photovoltaic grid-connected power supply.

[0008] It should be understood that the contents described in this section are not intended to indicate the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the present disclosure or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0010] Figure 1 A schematic diagram of a flow chart of an intelligent integrated testing method for power generation capacity prediction provided by an embodiment of the present disclosure;

[0011] Figure 2 A schematic diagram of the structure of an intelligent integrated test system for power generation capacity prediction provided in an embodiment of the present disclosure.

[0012] Explanation of the accompanying drawings: distributed photovoltaic information acquisition module 11, photovoltaic twin model construction module 12, environmental data deviation matching library acquisition module 13, matching environment prediction data acquisition module 14, photovoltaic power generation prediction channel acquisition module 15, power generation power prediction module 16. DETAILED DESCRIPTION

[0013] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0014] Embodiment 1

[0015] The intelligent integrated test method for power generation capacity prediction provided by the embodiment of the present disclosure is referred to Figure 1 Illustratively, the method comprises:

[0016] Connecting to the distributed photovoltaic operation and management system to obtain multiple distributed photovoltaic information of the target area, wherein the distributed photovoltaic information includes device coordinates and device operation data;

[0017] Specifically, the distributed photovoltaic operation and management system is used to manage and maintain distributed photovoltaic equipment. Connect the distributed photovoltaic operation and management system to obtain multiple distributed photovoltaic information of the target area. The target area is the area where the power generation capacity is to be predicted. The distributed photovoltaic information includes the device coordinates and device operation data of the distributed photovoltaic equipment. In the target area, a coordinate system is constructed with a random point as the coordinate origin to obtain the device coordinates of the distributed photovoltaic equipment. The equipment operation data includes the operating parameter data and basic equipment information of the distributed photovoltaic equipment. For example, the equipment operation data includes installed capacity, equipment type, and photoelectric conversion rate.

[0018] Based on the digital twin, simulation modeling is performed based on the equipment operation data, and the equipment coordinates are used as model positioning identifiers to construct a photovoltaic twin model of the target area;

[0019] Specifically, a digital twin model is constructed based on digital twin technology, and the equipment operation data and equipment coordinates are input into the digital twin model to obtain a simulation model of the distributed photovoltaic equipment as a photovoltaic twin model within the target area. Among them, the photovoltaic twin model has a model positioning identifier, and the parameter data of the model positioning identifier is the equipment coordinates.

[0020] Based on the location coordinates of the photovoltaic operation and management center, the photovoltaic twin model is used to analyze the light radiation deviation, temperature deviation and weather deviation of multiple distributed photovoltaic devices to generate an environmental data deviation matching library;

[0021] Specifically, the distributed photovoltaic operation and management system has a photovoltaic operation and management center, and the location coordinates of the photovoltaic operation and management center are obtained. Among them, the photovoltaic operation and management center is obtained by a technician in this field according to the actual situation through custom settings. For example, according to the device coordinates of the distributed photovoltaic equipment, the area where the distributed photovoltaic equipment is concentrated is obtained, and the photovoltaic operation and management center is obtained by taking the center point value. Through the photovoltaic twin model, the light radiation, temperature, weather and other data at the locations of multiple distributed photovoltaic devices are respectively calculated for deviations, and the light radiation deviation, temperature deviation and weather deviation of multiple distributed photovoltaic devices are obtained with the photovoltaic operation and management center as the standard light radiation, temperature and weather, and an environmental data deviation matching library is generated, thereby obtaining the light radiation deviation, temperature deviation and weather deviation between any distributed photovoltaic devices in the target area.

[0022] Collecting environmental prediction data within a preset time window of the photovoltaic operation and management center, and performing deviation analysis on the environmental prediction data based on the environmental data deviation matching library to obtain a plurality of matching environmental prediction data, wherein the matching environmental prediction data corresponds to distributed photovoltaic equipment one by one;

[0023] Specifically, the environmental prediction data within the preset time window of the photovoltaic operation and management center is collected, wherein the preset time window is obtained by a technician in this field according to the actual situation through custom settings, for example, the preset time window is the next half hour. For example, the environmental prediction data is obtained by collecting and monitoring through corresponding monitoring equipment, such as a thermometer. Based on the environmental prediction data of the photovoltaic operation and management center, the environmental prediction data is analyzed for deviations according to the environmental data deviation matching library, and the prediction data of multiple distributed photovoltaic devices are obtained as multiple matching environmental prediction data. Among them, multiple matching environmental prediction data correspond to distributed photovoltaic devices one by one.

[0024] Inputting the plurality of matching environment prediction data into a photovoltaic power generation prediction model to perform power generation prediction, wherein the photovoltaic power generation prediction model is constructed based on an integrated learning principle and the photovoltaic twin model, and comprises a plurality of photovoltaic power generation prediction channels, and each photovoltaic power generation prediction channel comprises a plurality of photovoltaic power generation prediction sub-channels;

[0025] Specifically, a photovoltaic power generation prediction model is constructed, and the photovoltaic power generation prediction model is trained based on the principle of integrated learning and the data in the photovoltaic twin model. The photovoltaic power generation prediction model has multiple photovoltaic power generation prediction channels, which correspond to the weather data and each photovoltaic power generation prediction channel includes multiple photovoltaic power generation prediction sub-channels, each photovoltaic power generation prediction sub-channel corresponds to a power generation prediction operator, and the power generation prediction operator is an algorithm for predicting environmental data, such as a convolutional neural network.

[0026] The power generation power of the plurality of matching environment prediction data is predicted respectively through a plurality of photovoltaic power generation prediction sub-channels, and the photovoltaic power generation prediction power of the target area within the preset time window is obtained by weighted calculation.

[0027] Specifically, multiple photovoltaic power generation prediction sub-channels are used to predict power generation for multiple matching environmental prediction data. The higher the correlation between weather data and power generation prediction operator, the higher the prediction accuracy of the corresponding matching photovoltaic power generation prediction sub-channel. Conversely, the lower the prediction accuracy, the higher the weight is assigned. Conversely, the weighted power generation prediction power is used to calculate the photovoltaic power generation prediction power of the target area within the preset time window.

[0028] Among them, this embodiment can solve the technical problems in the prior art that the environmental parameters of distributed power generation equipment have a large deviation from the environmental data collected by the operation and management center, resulting in low data collection accuracy and a relatively single prediction method, resulting in low accuracy in the prediction of the overall power generation capacity of distributed photovoltaic power generation equipment. It can realize multi-dimensional targeted prediction of distributed photovoltaic equipment, reduce the error in power generation prediction, and improve the accuracy of distributed photovoltaic overall power generation prediction, thereby improving the technical effect of improving the stability of distributed photovoltaic grid-connected power supply.

[0029] The method provided in the embodiment of the present disclosure also includes:

[0030] Based on the distributed photovoltaic operation and management system, multiple historical environmental data of the photovoltaic operation and management center and a historical environmental data set of multiple distributed photovoltaic devices are retrieved, wherein the historical environmental data and the historical environmental data set have the same data monitoring node;

[0031] Selecting a first distributed photovoltaic device from the plurality of distributed photovoltaic devices, and acquiring a first historical environmental data set of the first distributed photovoltaic device;

[0032] Performing data anomaly analysis on the first historical environment data set according to the plurality of historical environment data, and removing the abnormal data to obtain a first standard environment data set;

[0033] Taking the data monitoring node as a constraint, performing light radiation deviation, temperature deviation and weather deviation analysis based on the first standard environment data set and the corresponding multiple historical environment data to obtain first deviation data;

[0034] The environmental data deviation matching library is constructed based on the first deviation data.

[0035] Specifically, multiple historical environmental data of the photovoltaic operation and management center in the distributed photovoltaic operation and management system and a historical environmental data set of multiple distributed photovoltaic devices in the historical time are retrieved. For example, the historical environmental data and the historical environmental data set are environmental data such as meteorological data. Furthermore, the historical environmental data and the historical environmental data set have the same data monitoring node. Among them, the data monitoring node refers to the time node of the monitoring environment within the historical time period. For example, a data monitoring node is set every hour.

[0036] Furthermore, a distributed photovoltaic device is randomly selected from a plurality of distributed photovoltaic devices, and a first distributed photovoltaic device is randomly obtained as a first distributed photovoltaic device, and a first historical environmental data set corresponding to the first distributed photovoltaic device is obtained from the historical environmental data set. The first historical environmental data set includes a set of historical environmental data of a plurality of data monitoring nodes.

[0037] Furthermore, a correlation analysis is performed on the first historical environment data set based on multiple historical environment data to obtain data change trends and data deviation thresholds, and data anomaly analysis is performed based on the data change trends and data deviation thresholds to obtain abnormal data, and the abnormal data is eliminated from the first historical environment data set to obtain a first standard environment data set.

[0038] Furthermore, taking the data monitoring node as the time constraint for deviation analysis, the first standard environmental data set and the historical environmental data under the same data monitoring node are analyzed for light radiation deviation, temperature deviation and weather deviation, and the deviation data obtained respectively are combined to obtain the first deviation data.

[0039] Furthermore, the first deviation data is stored, and an environmental data deviation matching library is constructed. The environmental data deviation matching library uses the device coordinates of the distributed photovoltaic equipment as child nodes for deviation matching, and outputs the deviation matching results to merge into a deviation matching result set. For example, fixed environmental deviation data of the area where multiple random photovoltaic power generation equipment are located in the target area can be obtained according to the environmental data deviation matching library, such as obtaining the fixed light radiation deviation of the photovoltaic operation and management center and the random photovoltaic power generation equipment under the same monitoring time. Generating an environmental data deviation matching library can improve the accuracy of obtaining any random environmental data deviation through local environmental data deviation, thereby improving the accuracy of power generation prediction.

[0040] The method provided in the embodiment of the present disclosure also includes:

[0041] Performing correlation analysis on the first historical environment data set according to the multiple historical environment data to determine multiple data change trends and multiple data deviation thresholds;

[0042] The first historical environment data in the first historical environment data set that does not meet the data change trend and does not meet the data deviation threshold is marked as abnormal data.

[0043] Specifically, multiple historical environmental data and the first historical environmental data set are analyzed for association according to the time sequence of the data monitoring node. During the analysis, the data change trend of the first historical environmental data set at the data monitoring node is determined, and the data comparison of the historical environmental data of the same data monitoring node and the first historical environmental data set is determined to obtain the threshold of data deviation. Among them, the change trend refers to the environmental data that changes with the time of the data monitoring node, for example, the data change trend generated by the change of rainfall at 7 o'clock, 8 o'clock and 9 o'clock. The data deviation threshold refers to the data deviation threshold of the historical environmental data and the first historical environmental data set at the same monitoring node, for example, at 7 o'clock, the data deviation of the rainfall of the photovoltaic operation and management center and the rainfall of the distributed photovoltaic equipment, and the data deviation of the rainfall at 8 o'clock, 9 o'clock and 10 o'clock is obtained, and the mean of the data deviation at multiple times is obtained as the data deviation threshold. For example, the mean of radiation deviation, the mean of temperature deviation and the mean of weather deviation are obtained as the data deviation threshold.

[0044] Furthermore, the first historical environmental data in the first historical environmental data set are extracted in sequence, and the data change trend and the data deviation threshold are judged to be satisfied. When the first historical environmental data set does not satisfy the data change trend and does not satisfy the data deviation threshold at the same time, it means that the first historical environmental data set has a large change and the data deviation from the historical environmental data of the photovoltaic operation and management center is large. Then, the first historical environmental data in the first historical environmental data set that does not satisfy the data change trend and does not satisfy the data deviation threshold at the same time is marked to obtain abnormal data. Among them, by performing correlation analysis on the first historical environmental data set and the historical environmental data, and performing abnormal analysis, the accuracy of obtaining abnormal data can be improved.

[0045] The method provided in the embodiment of the present disclosure also includes:

[0046] Among them, the types of weather monitoring data analyzed through weather deviation include at least rainfall scenes, cloudy scenes and haze scenes;

[0047] Under the same data monitoring node, deviation calculation is performed based on the first standard environment data set and the corresponding multiple historical environment data to obtain multiple light radiation deviations, multiple temperature deviations and multiple weather deviations;

[0048] Mean values ​​are calculated for the multiple light radiation deviations, the multiple temperature deviations, and the multiple weather deviations, respectively, and the mean calculation result is used as the first deviation data.

[0049] Specifically, the weather deviation analysis has a weather monitoring data type, which refers to a scene indicator for weather deviation analysis, including at least a rainy scene, a cloudy scene, and a haze scene. The weather monitoring data type can be added or deleted according to actual conditions.

[0050] Furthermore, the first standard environmental data set of the first distributed photovoltaic device and the historical environmental data of the photovoltaic operation and management center under the same data monitoring node are calculated for deviation, and multiple light radiation deviations, multiple temperature deviations and multiple weather deviations under multiple data monitoring nodes are obtained. Among them, the weather deviation includes at least rainfall deviation, cloud cover deviation and haze content.

[0051] Furthermore, the mean values ​​of the multiple light radiation deviations, the multiple temperature deviations and the multiple weather deviations are calculated respectively to obtain the first radiation deviation mean value, the first temperature deviation mean value and the first weather deviation mean value, and the first radiation deviation mean value, the first temperature deviation mean value and the first weather deviation mean value are used as the first deviation data. The accuracy of the first deviation data can be improved by analyzing the light radiation deviation, the temperature deviation and the weather deviation.

[0052] The method provided in the embodiment of the present disclosure also includes:

[0053] Setting N integrated learning operators, wherein the integrated learning operators include at least a BP neural network, a support vector machine, and a convolutional neural network, and N is an integer greater than or equal to 3;

[0054] Randomly selecting a first photovoltaic model in the photovoltaic twin model, and obtaining first photovoltaic device information of the first photovoltaic model;

[0055] Constructing N first photovoltaic power generation prediction sub-channels based on the N integrated learning operators and the first photovoltaic device information;

[0056] Taking the weather monitoring data type as a constraint, obtaining a training data set under multiple weather scenarios;

[0057] Randomly selecting a first weather scene from the multiple weather scenes, obtaining a training data set under the first weather scene, dividing the training data set into N equal parts, and performing supervised learning on the N first photovoltaic power generation prediction sub-channels respectively to obtain N first photovoltaic power generation prediction sub-channels that meet the expected convergence index and the corresponding prediction accuracy;

[0058] Constructing a first photovoltaic power generation prediction channel under the first weather scenario according to the N first photovoltaic power generation prediction sub-channels and the corresponding prediction accuracy rates;

[0059] A photovoltaic power generation prediction model is generated by taking the first weather scenario as a first child node and the first photovoltaic power generation prediction channel under the first weather scenario as a leaf node of the first child node.

[0060] Specifically, N integrated learning operators are set to perform environmental data deviation calculations, wherein the N integrated learning operators at least include learning operators of a BP neural network, a support vector machine, and a convolutional neural network, and therefore, N is an integer greater than or equal to 3.

[0061] Furthermore, a photovoltaic power generation device is randomly selected in the photovoltaic twin model as the first photovoltaic device, and the photovoltaic twin model corresponding to the first photovoltaic device is obtained as the first photovoltaic model. The first photovoltaic device information of the first photovoltaic model is obtained. The first photovoltaic device information refers to the device parameter information of the first photovoltaic device.

[0062] Furthermore, N first photovoltaic power generation prediction sub-channels are constructed based on the N integrated learning operators and the first photovoltaic device information, respectively. The N first photovoltaic power generation prediction sub-channels are used for calculating the environmental data deviation of the first photovoltaic device, and the operators of the N first photovoltaic power generation prediction sub-channels are different.

[0063] Furthermore, the light radiation data and temperature data in the historical environmental data are extracted from the light radiation data and temperature data in the historical environmental data set, and are respectively combined with the power generation records of the data monitoring nodes corresponding to the historical environmental data and the power generation records of the data monitoring nodes corresponding to the historical environmental data set, that is, the light radiation data and temperature data in the historical environmental data are combined with the power generation records of the corresponding data monitoring nodes, and the light radiation data and temperature data in the historical environmental data set are combined with the power generation records of the corresponding data monitoring nodes to obtain combined data. Among them, taking the weather monitoring data type as a constraint, that is, taking the rain scene, cloudy scene and haze scene as the training scene constraint, the combined data is trained in multiple training scene constraints to obtain a training data set. For example, the combined data is trained in the rain scene, cloudy scene and haze scene respectively to obtain a training data set.

[0064] Furthermore, a weather scene is randomly selected from multiple weather scenes as the first weather scene, for example, the first weather scene is a rainfall scene. A training data set under the first weather scene is obtained. The training data set under the first weather scene is evenly divided into N parts, and the N training data sets are respectively input into the first photovoltaic power generation prediction sub-channels corresponding to the N integrated learning operators for supervised learning. When the expected convergence index is met, N first photovoltaic power generation prediction sub-channels and the corresponding prediction accuracy are obtained. Among them, the expected convergence index is obtained by a technician in this field according to the actual situation. For example, the expected convergence index is qualified after 10,000 training times.

[0065] Furthermore, the N prediction accuracy rates are compared with the preset accuracy threshold of the first photovoltaic power generation prediction sub-channel corresponding to each integrated learning operator. When the prediction accuracy of the first photovoltaic power generation prediction sub-channel is greater than or equal to the preset accuracy threshold corresponding to the integrated learning operator, it means that in the first weather scene, the prediction accuracy of the established integrated learning operator is high, and the prediction accuracy of other integrated learning operators is low. The correlation between the weather scene and the integrated learning operator is obtained, and then the first photovoltaic power generation prediction channel in the first weather scene is constructed. The preset accuracy threshold is obtained by a technician in this field according to the actual situation. For example, the preset accuracy threshold is 95%.

[0066] Furthermore, the first weather scene is taken as the first child node, and the first photovoltaic power generation prediction channel under the first weather scene is taken as the leaf node of the first child node, so that the weather scene is matched first, and then the corresponding integrated learning operator is selected. The second weather scene is taken as the second child node, and the second photovoltaic power generation prediction channel under the second weather scene is taken as the leaf node of the second child node, until N integrated learning operators are associated with the weather scenes in turn to generate a photovoltaic power generation prediction model. Among them, the prediction accuracy corresponding to different integrated learning operators is different in different scenes. By associating the integrated learning operator with the weather scene, the accuracy of the result of the power generation capacity prediction is improved.

[0067] The method provided in the embodiment of the present disclosure also includes:

[0068] Randomly selecting first matching environment prediction data from the plurality of matching environment prediction data, and extracting first weather monitoring data from the first matching environment prediction data;

[0069] According to the first weather monitoring data, a plurality of corresponding first matching photovoltaic power generation prediction sub-channels are obtained by matching in the photovoltaic power generation prediction model;

[0070] Performing power generation prediction on the first matching environment prediction data through the multiple first matching photovoltaic power generation prediction sub-channels to obtain multiple first power generation prediction powers;

[0071] Based on the coefficient of variation method, according to the prediction accuracy rates corresponding to the multiple first matching photovoltaic power generation prediction sub-channels, credible weights are set for the multiple first matching photovoltaic power generation prediction sub-channels to obtain multiple credible weights;

[0072] Performing weighted calculation on the plurality of first power generation prediction powers according to the plurality of credible weights to obtain a first power generation prediction power;

[0073] Based on the first power generation prediction power, multiple power generation prediction powers corresponding to multiple matching environment prediction data are obtained in sequence, and the multiple power generation prediction powers are summed up to obtain the photovoltaic power generation prediction power.

[0074] Specifically, a piece of matching environment prediction data is randomly selected from a plurality of matching environment prediction data as the first matching environment prediction data, and a piece of weather monitoring data randomly selected from the first matching environment prediction data is extracted as the first weather monitoring data.

[0075] Furthermore, a plurality of corresponding first matching photovoltaic power generation prediction sub-channels are obtained by matching the first weather monitoring data in the photovoltaic power generation prediction model, which are used to perform environmental prediction according to the correlation between the weather data and the integrated learning operator, and then perform power generation prediction.

[0076] Furthermore, the power generation power of the first matching environment prediction data is predicted through multiple first matching photovoltaic power generation prediction sub-channels to obtain multiple first power generation prediction powers, wherein the prediction accuracy of the multiple first power generation prediction powers is different. When the correlation between the weather data and the integrated learning operator is higher, the prediction accuracy of the first power generation prediction power of the corresponding first matching photovoltaic power generation prediction sub-channel is higher, and vice versa.

[0077] Furthermore, based on the coefficient of variation method, a higher credible weight is set for the first matching photovoltaic power generation prediction sub-channel with a higher prediction accuracy, and vice versa, a lower credible weight is set, thereby obtaining multiple credible weights.

[0078] Furthermore, a plurality of first power generation prediction powers are weightedly calculated according to a plurality of credible weights to obtain the first power generation prediction power.

[0079] Further, according to the method for obtaining the first power generation prediction power, the second power generation prediction power corresponding to the second matching environment prediction data is obtained until multiple matching environment prediction data are accessed in sequence to obtain multiple corresponding power generation prediction powers, and the multiple power generation prediction powers are summed up to obtain the photovoltaic power generation prediction power. Among them, the accuracy of obtaining multiple power generation prediction powers can be improved by weight allocation.

[0080] The method provided in the embodiment of the present disclosure also includes:

[0081] Performing a temperature rise trend analysis on the first matching environment prediction data within the preset time window, and judging whether the first temperature prediction data meets a preset high temperature warning threshold according to the temperature rise trend analysis result;

[0082] If satisfied, analyze and obtain the warning time node and terminal temperature, and determine the warning time period based on the warning time node;

[0083] determining a power compensation coefficient based on the terminal temperature and the warning time period;

[0084] The first predicted power generation power is compensated according to the power compensation coefficient.

[0085] Specifically, if the temperature data collected in the first matching environment prediction data within the preset time window increases, the first matching environment prediction data has a heating trend, otherwise, it does not. The result of the first matching environment prediction data having a heating trend in the heating trend analysis result is extracted to determine whether the first temperature prediction data meets the preset high temperature warning threshold. Among them, the preset high temperature warning threshold is obtained by custom settings by technical personnel in this field according to actual conditions. For example, the preset high temperature warning threshold is 25 degrees Celsius. Among them, the increase in temperature will cause the output power of the photovoltaic module to decrease. That is, for every degree increase in temperature, the power generation of the photovoltaic power station will decrease by about 0.44%. Attenuation coefficient. Usually exceeds 25 degrees Celsius.

[0086] Furthermore, if it is determined that the first temperature prediction data meets the preset high temperature warning threshold, the warning time node and the terminal temperature are analyzed and obtained, wherein the warning time node refers to the time when the temperature data reaches 25 degrees Celsius, and the terminal temperature refers to the highest temperature rise within the preset time window. The time from the warning time node to the end time node of the preset time window is used as the warning time period.

[0087] Furthermore, the terminal temperature and the preset high temperature warning threshold of the warning time node are averaged to obtain the average temperature exceeding the preset high temperature warning threshold, and the average temperature is multiplied by the power generation attenuation coefficient to obtain the power generation loss coefficient of the warning time period. The warning time coefficient of the warning time period accounting for the preset time window is calculated, and the result calculated based on the product of the warning time coefficient and the power generation loss coefficient is used as the power compensation coefficient.

[0088] Furthermore, the first predicted power generation power is compensated according to the power compensation coefficient, that is, the power generation power is increased or decreased, etc., so as to improve the accuracy of obtaining the first predicted power generation power.

[0089] Embodiment 2

[0090] Based on the same inventive concept as the intelligent integrated testing method for power generation capacity prediction in the aforementioned embodiment, Figure 2 As shown, the present disclosure also provides an intelligent integrated test system for power generation capacity prediction, the system comprising:

[0091] A distributed photovoltaic information acquisition module 11, which is used to connect to a distributed photovoltaic operation and management system to obtain a plurality of distributed photovoltaic information of a target area, wherein the distributed photovoltaic information includes device coordinates and device operation data;

[0092] A photovoltaic twin model construction module 12, wherein the photovoltaic twin model construction module 12 is used to perform simulation modeling based on the digital twin and the device operation data, and to construct a photovoltaic twin model of the target area using the device coordinates as a model positioning identifier;

[0093] An environmental data deviation matching library acquisition module 13, wherein the environmental data deviation matching library acquisition module 13 is used to analyze the light radiation deviation, temperature deviation and weather deviation of multiple distributed photovoltaic devices based on the photovoltaic twin model based on the location coordinates of the photovoltaic operation and management center, and generate an environmental data deviation matching library;

[0094] A matching environment prediction data acquisition module 14, the matching environment prediction data acquisition module 14 is used to collect the environment prediction data within a preset time window of the photovoltaic operation and management center, and perform deviation analysis on the environment prediction data based on the environment data deviation matching library to obtain a plurality of matching environment prediction data, wherein the matching environment prediction data corresponds to the distributed photovoltaic equipment one by one;

[0095] A photovoltaic power generation prediction channel acquisition module 15, wherein the photovoltaic power generation prediction channel acquisition module 15 is used to input the plurality of matching environment prediction data into a photovoltaic power generation prediction model for power generation prediction, wherein the photovoltaic power generation prediction model is constructed based on an integrated learning principle and the photovoltaic twin model, and comprises a plurality of photovoltaic power generation prediction channels, and each photovoltaic power generation prediction channel comprises a plurality of photovoltaic power generation prediction sub-channels;

[0096] The power generation prediction module 16 is used to perform power generation prediction on the multiple matching environment prediction data respectively through multiple photovoltaic power generation prediction sub-channels, and obtain the photovoltaic power generation prediction power of the target area within the preset time window through weighted calculation.

[0097] Furthermore, the environment data deviation matching library acquisition module 13 also includes:

[0098] Based on the distributed photovoltaic operation and management system, multiple historical environmental data of the photovoltaic operation and management center and a historical environmental data set of multiple distributed photovoltaic devices are retrieved, wherein the historical environmental data and the historical environmental data set have the same data monitoring node;

[0099] Selecting a first distributed photovoltaic device from the plurality of distributed photovoltaic devices, and acquiring a first historical environmental data set of the first distributed photovoltaic device;

[0100] Performing data anomaly analysis on the first historical environment data set according to the plurality of historical environment data, and removing the abnormal data to obtain a first standard environment data set;

[0101] Taking the data monitoring node as a constraint, performing light radiation deviation, temperature deviation and weather deviation analysis based on the first standard environment data set and the corresponding multiple historical environment data to obtain first deviation data;

[0102] The environmental data deviation matching library is constructed based on the first deviation data.

[0103] Furthermore, the environment data deviation matching library acquisition module 13 also includes:

[0104] Performing correlation analysis on the first historical environment data set according to the multiple historical environment data to determine multiple data change trends and multiple data deviation thresholds;

[0105] The first historical environment data in the first historical environment data set that does not meet the data change trend and does not meet the data deviation threshold is marked as abnormal data.

[0106] Furthermore, the environment data deviation matching library acquisition module 13 also includes:

[0107] Among them, the types of weather monitoring data analyzed through weather deviation include at least rainfall scenes, cloudy scenes and haze scenes;

[0108] Under the same data monitoring node, deviation calculation is performed based on the first standard environment data set and the corresponding multiple historical environment data to obtain multiple light radiation deviations, multiple temperature deviations and multiple weather deviations;

[0109] Mean values ​​are calculated for the multiple light radiation deviations, the multiple temperature deviations, and the multiple weather deviations, respectively, and the mean calculation result is used as the first deviation data.

[0110] Furthermore, the environment data deviation matching library acquisition module 13 also includes:

[0111] Setting N integrated learning operators, wherein the integrated learning operators include at least a BP neural network, a support vector machine, and a convolutional neural network, and N is an integer greater than or equal to 3;

[0112] Randomly selecting a first photovoltaic model in the photovoltaic twin model, and obtaining first photovoltaic device information of the first photovoltaic model;

[0113] Constructing N first photovoltaic power generation prediction sub-channels based on the N integrated learning operators and the first photovoltaic device information;

[0114] Taking the weather monitoring data type as a constraint, obtaining a training data set under multiple weather scenarios;

[0115] Randomly selecting a first weather scene from the multiple weather scenes, obtaining a training data set under the first weather scene, dividing the training data set into N equal parts, and performing supervised learning on the N first photovoltaic power generation prediction sub-channels respectively to obtain N first photovoltaic power generation prediction sub-channels that meet the expected convergence index and the corresponding prediction accuracy;

[0116] Constructing a first photovoltaic power generation prediction channel under the first weather scenario according to the N first photovoltaic power generation prediction sub-channels and the corresponding prediction accuracy rates;

[0117] A photovoltaic power generation prediction model is generated by taking the first weather scenario as a first child node and the first photovoltaic power generation prediction channel under the first weather scenario as a leaf node of the first child node.

[0118] Furthermore, the environment data deviation matching library acquisition module 13 also includes:

[0119] Randomly selecting first matching environment prediction data from the plurality of matching environment prediction data, and extracting first weather monitoring data from the first matching environment prediction data;

[0120] According to the first weather monitoring data, a plurality of corresponding first matching photovoltaic power generation prediction sub-channels are obtained by matching in the photovoltaic power generation prediction model;

[0121] Performing power generation prediction on the first matching environment prediction data through the multiple first matching photovoltaic power generation prediction sub-channels to obtain multiple first power generation prediction powers;

[0122] Based on the coefficient of variation method, according to the prediction accuracy rates corresponding to the multiple first matching photovoltaic power generation prediction sub-channels, credible weights are set for the multiple first matching photovoltaic power generation prediction sub-channels to obtain multiple credible weights;

[0123] Performing weighted calculation on the plurality of first power generation prediction powers according to the plurality of credible weights to obtain a first power generation prediction power;

[0124] Based on the first power generation prediction power, multiple power generation prediction powers corresponding to multiple matching environment prediction data are obtained in sequence, and the multiple power generation prediction powers are summed up to obtain the photovoltaic power generation prediction power.

[0125] Furthermore, the environment data deviation matching library acquisition module 13 also includes:

[0126] Performing a temperature rise trend analysis on the first matching environment prediction data within the preset time window, and judging whether the first temperature prediction data meets a preset high temperature warning threshold according to the temperature rise trend analysis result;

[0127] If satisfied, analyze and obtain the warning time node and terminal temperature, and determine the warning time period based on the warning time node;

[0128] determining a power compensation coefficient based on the terminal temperature and the warning time period;

[0129] The first predicted power generation power is compensated according to the power compensation coefficient.

[0130] The specific example of the intelligent integrated test method for power generation capacity prediction in the aforementioned embodiment 1 is also applicable to the intelligent integrated test system for power generation capacity prediction in this embodiment. Through the aforementioned detailed description of the intelligent integrated test method for power generation capacity prediction, those skilled in the art can clearly understand the intelligent integrated test system for power generation capacity prediction in this embodiment, so for the sake of brevity of the specification, it will not be described in detail here. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0131] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of this disclosure can be achieved, and this document is not limited here.

[0132] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. An intelligent integrated test method for power generation capacity prediction, characterized in that: The method comprises: Connecting to the distributed photovoltaic operation and management system to obtain multiple distributed photovoltaic information of the target area, wherein the distributed photovoltaic information includes device coordinates and device operation data; Based on the digital twin, simulation modeling is performed based on the equipment operation data, and the equipment coordinates are used as model positioning identifiers to construct a photovoltaic twin model of the target area; Based on the location coordinates of the photovoltaic operation and management center, the photovoltaic twin model is used to analyze the light radiation deviation, temperature deviation and weather deviation of multiple distributed photovoltaic devices to generate an environmental data deviation matching library; Collecting environmental prediction data within a preset time window of the photovoltaic operation and management center, and performing deviation analysis on the environmental prediction data based on the environmental data deviation matching library to obtain a plurality of matching environmental prediction data, wherein the matching environmental prediction data corresponds to distributed photovoltaic equipment one by one; Inputting the plurality of matching environment prediction data into a photovoltaic power generation prediction model to perform power generation prediction, wherein the photovoltaic power generation prediction model is constructed based on an integrated learning principle and the photovoltaic twin model, and comprises a plurality of photovoltaic power generation prediction channels, and each photovoltaic power generation prediction channel comprises a plurality of photovoltaic power generation prediction sub-channels; The photovoltaic power generation prediction sub-channels are used to respectively predict the power generation of the plurality of matching environment prediction data, and a weighted calculation is performed to obtain the photovoltaic power generation prediction power of the target area within the preset time window; Wherein, the generation of the environment data deviation matching library includes: Based on the distributed photovoltaic operation and management system, multiple historical environmental data of the photovoltaic operation and management center and a historical environmental data set of multiple distributed photovoltaic devices are retrieved, wherein the historical environmental data and the historical environmental data set have the same data monitoring node; Selecting a first distributed photovoltaic device from the plurality of distributed photovoltaic devices, and acquiring a first historical environmental data set of the first distributed photovoltaic device; Performing data anomaly analysis on the first historical environment data set according to the plurality of historical environment data, and removing the abnormal data to obtain a first standard environment data set; Taking the data monitoring node as a constraint, performing light radiation deviation, temperature deviation and weather deviation analysis based on the first standard environment data set and the corresponding multiple historical environment data to obtain first deviation data; Building the environmental data deviation matching library based on the first deviation data; The performing data anomaly analysis on the first historical environment data set according to the plurality of historical environment data includes: Performing correlation analysis on the first historical environment data set according to the multiple historical environment data to determine multiple data change trends and multiple data deviation thresholds; The first historical environment data in the first historical environment data set that does not meet the data change trend and does not meet the data deviation threshold is marked as abnormal data.

2. The method according to claim 1, characterized in that The first standard environment data set and the corresponding multiple historical environment data are used to perform light radiation deviation, temperature deviation and weather deviation analysis to obtain first deviation data, include: Among them, the types of weather monitoring data analyzed through weather deviation include at least rainfall scenes, cloudy scenes and haze scenes; Under the same data monitoring node, deviation calculation is performed based on the first standard environment data set and the corresponding multiple historical environment data to obtain multiple light radiation deviations, multiple temperature deviations and multiple weather deviations; Mean values ​​are calculated for the multiple light radiation deviations, the multiple temperature deviations, and the multiple weather deviations, respectively, and the mean calculation result is used as the first deviation data.

3. The method according to claim 2, characterized in that The method further comprises: Setting N integrated learning operators, wherein the integrated learning operators include at least a BP neural network, a support vector machine, and a convolutional neural network, and N is an integer greater than or equal to 3; Randomly selecting a first photovoltaic model in the photovoltaic twin model, and obtaining first photovoltaic device information of the first photovoltaic model; Constructing N first photovoltaic power generation prediction sub-channels based on the N integrated learning operators and the first photovoltaic device information; Taking the weather monitoring data type as a constraint, obtaining a training data set under multiple weather scenarios; Randomly selecting a first weather scene from the multiple weather scenes, obtaining a training data set under the first weather scene, dividing the training data set into N equal parts, and performing supervised learning on the N first photovoltaic power generation prediction sub-channels respectively to obtain N first photovoltaic power generation prediction sub-channels that meet the expected convergence index and the corresponding prediction accuracy; Constructing a first photovoltaic power generation prediction channel under the first weather scenario according to the N first photovoltaic power generation prediction sub-channels and the corresponding prediction accuracy rates; A photovoltaic power generation prediction model is generated by taking the first weather scenario as a first child node and the first photovoltaic power generation prediction channel under the first weather scenario as a leaf node of the first child node.

4. The method according to claim 3, characterized in that The method of respectively predicting the power generation of the plurality of matching environment prediction data through a plurality of photovoltaic power generation prediction sub-channels, and obtaining the photovoltaic power generation prediction power of the target area within the preset time window through weighted calculation, further includes: Randomly selecting first matching environment prediction data from the plurality of matching environment prediction data, and extracting first weather monitoring data from the first matching environment prediction data; According to the first weather monitoring data, a plurality of corresponding first matching photovoltaic power generation prediction sub-channels are obtained by matching in the photovoltaic power generation prediction model; Performing power generation prediction on the first matching environment prediction data through the multiple first matching photovoltaic power generation prediction sub-channels to obtain multiple first power generation prediction powers; Based on the coefficient of variation method, according to the prediction accuracy rates corresponding to the multiple first matching photovoltaic power generation prediction sub-channels, credible weights are set for the multiple first matching photovoltaic power generation prediction sub-channels to obtain multiple credible weights; Performing weighted calculation on the plurality of first power generation prediction powers according to the plurality of credible weights to obtain a first power generation prediction power; Based on the first power generation prediction power, multiple power generation prediction powers corresponding to multiple matching environment prediction data are obtained in sequence, and the multiple power generation prediction powers are summed up to obtain the photovoltaic power generation prediction power.

5. The method according to claim 4, characterized in that The obtaining of the first predicted power generation comprises: Performing a temperature rise trend analysis on the first matching environment prediction data within the preset time window, and judging whether the first temperature prediction data meets a preset high temperature warning threshold according to the temperature rise trend analysis result; If satisfied, analyze and obtain the warning time node and terminal temperature, and determine the warning time period based on the warning time node; determining a power compensation coefficient based on the terminal temperature and the warning time period; The first predicted power generation power is compensated according to the power compensation coefficient.

6. An intelligent integrated test system for power generation capacity prediction, characterized in that: The system comprises: A distributed photovoltaic information acquisition module, which is used to connect to a distributed photovoltaic operation and management system to obtain a plurality of distributed photovoltaic information of a target area, wherein the distributed photovoltaic information includes device coordinates and device operation data; A photovoltaic twin model construction module, wherein the photovoltaic twin model construction module is used to perform simulation modeling based on the digital twin and the device operation data, and to construct a photovoltaic twin model of the target area using the device coordinates as a model positioning identifier; An environmental data deviation matching library acquisition module, which is used to analyze the light radiation deviation, temperature deviation and weather deviation of multiple distributed photovoltaic devices based on the photovoltaic twin model based on the location coordinates of the photovoltaic operation and management center, and generate an environmental data deviation matching library; A matching environment prediction data acquisition module, the matching environment prediction data acquisition module is used to collect the environment prediction data within a preset time window of the photovoltaic operation and management center, and perform deviation analysis on the environment prediction data based on the environment data deviation matching library to obtain a plurality of matching environment prediction data, wherein the matching environment prediction data corresponds to the distributed photovoltaic equipment one by one; A photovoltaic power generation prediction channel acquisition module, the photovoltaic power generation prediction channel acquisition module is used to input the multiple matching environment prediction data into a photovoltaic power generation prediction model to perform power generation prediction, wherein the photovoltaic power generation prediction model is constructed based on an integrated learning principle and the photovoltaic twin model, and includes multiple photovoltaic power generation prediction channels, and each photovoltaic power generation prediction channel includes multiple photovoltaic power generation prediction sub-channels; A power generation prediction module, the power generation prediction module is used to respectively predict the power generation of the multiple matching environment prediction data through multiple photovoltaic power generation prediction sub-channels, and obtain the photovoltaic power generation prediction power of the target area within the preset time window through weighted calculation; The environment data deviation matching library acquisition module is also used for: Based on the distributed photovoltaic operation and management system, multiple historical environmental data of the photovoltaic operation and management center and a historical environmental data set of multiple distributed photovoltaic devices are retrieved, wherein the historical environmental data and the historical environmental data set have the same data monitoring node; Selecting a first distributed photovoltaic device from the plurality of distributed photovoltaic devices, and acquiring a first historical environmental data set of the first distributed photovoltaic device; Performing data anomaly analysis on the first historical environment data set according to the plurality of historical environment data, and removing the abnormal data to obtain a first standard environment data set; Taking the data monitoring node as a constraint, performing light radiation deviation, temperature deviation and weather deviation analysis based on the first standard environment data set and the corresponding multiple historical environment data to obtain first deviation data; Building the environmental data deviation matching library based on the first deviation data; The performing data anomaly analysis on the first historical environment data set according to the plurality of historical environment data includes: Performing correlation analysis on the first historical environment data set according to the multiple historical environment data to determine multiple data change trends and multiple data deviation thresholds; The first historical environment data in the first historical environment data set that does not meet the data change trend and does not meet the data deviation threshold is marked as abnormal data.

Images