Intelligent regulation and control method based on distributed new energy power station

Abstract

The invention relates to the technical field of new energy, in particular to an intelligent regulation and control method based on a distributed new energy power station. The method comprises the following steps: acquiring a target quantity characteristic value of an environmental data type at a future monitoring moment, and acquiring a data window of the environmental data type at the future monitoring moment according to the target quantity characteristic value of the environmental data type at the future monitoring moment, inputting the data window into a prediction model of the environmental data type, and outputting predicted environmental data of the environmental data type at a future monitoring moment; and according to the predicted environment data of the different environment data types and the generating capacity prediction model, obtaining the predicted generating capacity of the distributed new energy power station, and according to the predicted generating capacity, regulating and controlling the distributed new energy power station. And the accuracy of predicting the generating capacity of the distributed new energy power station can be improved, so that the effect of subsequently regulating and controlling the electric energy output to the power grid by the distributed new energy power station can be ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of new energy, and particularly to a smart regulation method based on a distributed new energy power station. Background Art

[0002] Since scheduling and controlling a distributed new energy power station is crucial for improving the stability of power grid operation, enhancing the new energy consumption efficiency, optimizing the energy utilization efficiency, reducing the operation cost, etc., it is of great importance to schedule and control the distributed new energy power station currently, that is to say, it is crucial to regulate and control the distributed new energy power station currently.

[0003] In the prior art, usually, the environmental data at the location of the distributed new energy power station is first predicted, and then the power generation of the distributed new energy power station at a future moment is predicted based on the predicted environmental data, and the power output from the distributed new energy power station to the power grid is regulated based on the predicted power generation of the distributed new energy power station; however, currently, when predicting the environmental data, generally, historical environmental data with a fixed length or historical environmental data with a random length is selected to predict the future environment at the location of the distributed new energy power station as the input during prediction. However, when using historical environmental data with a fixed length or random length as the input to predict the environmental data, the influence of environmental changes on the prediction of future environmental data will be ignored. That is, when using historical environmental data with a fixed length or random length as the input during prediction to predict future environmental data, the response ability of the model for predicting environmental data to environmental changes or the ability to capture environmental changes will be weak. As a result, when predicting the future environment at the location of the distributed new energy power station, problems of unreliable and inaccurate environmental prediction will occur. And the problems of unreliable and inaccurate environmental prediction will lead to unreliable and inaccurate prediction when predicting the power generation of the distributed new energy power station based on the predicted environmental data subsequently. And when there are problems with inaccurate prediction of the power generation of the distributed new energy power station, it will lead to poor regulation effect of the electric energy output from the distributed new energy power station to the power grid subsequently. Therefore, how to improve the accuracy of predicting the power generation of the distributed new energy power station to ensure the regulation effect of the electric energy output from the distributed new energy power station to the power grid subsequently has become an urgent problem to be solved. Summary of the Invention

[0004] To solve the above problems, the present invention provides a smart regulation method based on a distributed new energy power station, and the specific technical solution adopted is as follows: An embodiment of the present invention provides a smart regulation method based on a distributed new energy power station, including the following steps: Obtain the environmental data sequences corresponding to different environmental data types at the current monitoring moment, and the data in the environmental data sequences are environmental data collected at the location of the distributed new energy power station; For any environmental data type, obtain the seasonal component curve and the residual component curve of the environmental data sequence corresponding to the environmental data type, and obtain the extreme moment combination corresponding to the environmental data type at the current monitoring moment according to the extreme points on the seasonal component curve. According to the difference between the two extreme moments in the extreme moment combination and the positions on the residual component curve where the abscissa belongs to the extreme moment, obtain the target quantity characteristic value of the environmental data type at the future monitoring moment, and obtain the data window of the environmental data type at the future monitoring moment according to the target quantity characteristic value of the environmental data type at the future monitoring moment. Input the data window into the prediction model of the environmental data type to output the predicted environmental data of the environmental data type at the future monitoring moment; Obtain the predicted power generation of the distributed new energy power station according to the predicted environmental data of different environmental data types and the power generation prediction model, and regulate the distributed new energy power station according to the predicted power generation.

[0005] Beneficial effects: The present invention first obtains the environmental data sequences corresponding to different environmental data types at the current monitoring moment; then obtains the extreme points on the seasonal component curve of the environmental data sequence to obtain the extreme moment combination corresponding to the environmental data type at the current monitoring moment. According to the difference between the two extreme moments in the extreme moment combination and the positions on the residual component curve of the environmental data sequence where the abscissa belongs to the extreme moment, obtain the target quantity characteristic value of the environmental data type at the future monitoring moment, and obtain the data window of the environmental data type at the future monitoring moment according to the target quantity characteristic value of the environmental data type at the future monitoring moment. Input the data window into the prediction model of the environmental data type to output the predicted environmental data of the environmental data type at the future monitoring moment; obtain the predicted power generation of the distributed new energy power station according to the predicted environmental data of different environmental data types and the power generation prediction model, and regulate the distributed new energy power station according to the predicted power generation. Moreover, by adaptively obtaining the data window, the present invention can improve the accuracy of predicting the power generation of the distributed new energy power station, thereby ensuring the effect of subsequent regulation of the electric energy output from the distributed new energy power station to the power grid, and further avoiding the phenomenon of damage to the operation stability of the power grid and losses in the economy and market as much as possible. Description of the Drawings

[0006] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0007] Figure 1 It is a flowchart of a smart regulation method for a distributed new energy power station according to the present invention. Specific embodiments

[0008] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art belong to the scope protected by the embodiments of the present invention.

[0009] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0010] This embodiment provides a smart regulation method for a distributed new energy power station, which is described in detail as follows: As Figure 1 shown, the smart regulation method for a distributed new energy power station includes the following steps: Step S001: Obtain the environmental data sequence corresponding to different environmental data types at the current monitoring moment.

[0011] In this embodiment, by improving the prediction accuracy of the environmental data at the location of the distributed new energy power station, the prediction accuracy of the power generation of the distributed new energy power station is improved, so as to ensure the effect of subsequent regulation of the electric energy output from the distributed new energy power station to the power grid; in addition, for the convenience of analysis, the subsequent analysis and description of this embodiment will be carried out on the process of regulating the electric energy output from any distributed new energy power station connected to the power grid to the power grid.

[0012] First, obtain the location of the distributed new energy power station and denote it as the monitoring location. Obtain all the environmental data types related to the power generation of the distributed new energy power station, and denote the set constructed by all the environmental data types related to the power generation of the distributed new energy power station as the environmental data type set corresponding to the distributed new energy power station. And if the change of a certain type of environmental data will affect the power generation of the distributed new energy power station, it indicates that the corresponding environmental data type is an environmental data type related to the power generation of the distributed new energy power station. For example, for a photovoltaic distributed new energy power station, an increase in light intensity usually increases the photovoltaic power generation. Although the temperature rises along with strong light, if it is too high, it will reduce the conversion efficiency of component power generation, thus suppressing power generation. And an increase in humidity will cause clouds or haze, reducing the solar radiation reaching the surface of the components, thus reducing the power generation. Therefore, light intensity, environmental temperature, and environmental humidity all belong to the environmental data types related to the power generation of the distributed new energy power station. Then, in the current monitoring moment, obtain the environmental data sequences corresponding to each environmental data type in the environmental data type set. And in the current monitoring moment, the specific obtaining process of the environmental data sequence corresponding to any environmental data type is as follows: First, obtain the moment when the collection of environmental data belonging to this environmental data type starts at the monitoring location and denote it as the starting monitoring moment of this environmental data type. Then, denote the time series sequence composed of all the environmental data belonging to this environmental data type collected by the environmental data collection device during the time period from the starting monitoring moment of this environmental data type to the current monitoring moment as the environmental data sequence corresponding to this environmental data type at the current monitoring moment. The environmental data sequence corresponding to this environmental data type at the current monitoring moment includes the environmental data belonging to this environmental data type collected at the current monitoring moment.

[0013] In this embodiment, the collected or predicted environmental data is mainly meteorological data, and all the environmental data types in the above environmental data type set are collected synchronously, that is, the type and number of types of environmental data collected at any moment are the same as those in the environmental data type set. In addition, in specific applications, implementers usually set the time interval between adjacent collection moments in combination with the power station scale, hardware conditions, grid connection agreements, etc. For example, the time interval between adjacent monitoring moments can be set between 5 minutes and 1 hour. And it should also be noted that generally, monitoring devices are used to monitor and collect the environmental data at the location of the distributed new energy power station, such as using temperature and humidity sensors to monitor and collect the environmental temperature, environmental humidity, etc. at the location of the distributed new energy power station.

[0014] Therefore, through the above process, the environmental data sequences corresponding to different environmental data types at the current monitoring moment are obtained in this embodiment. The data in the environmental data sequences are all actual collected data. Moreover, the main reason why environmental factors affect the power generation of a distributed power station is that the distributed power station needs to convert external energy into electrical energy during the power generation process.

[0015] Step S002: For any environmental data type, obtain the seasonal component curve and the residual component curve of the environmental data sequence corresponding to the environmental data type. Obtain the extreme moment combination corresponding to the environmental data type at the current monitoring moment according to the extreme points on the seasonal component curve. Obtain the target quantity eigenvalue of the environmental data type at the future monitoring moment according to the difference between two extreme moments in the extreme moment combination and the positions of the abscissas belonging to the extreme moments on the residual component curve. Obtain the data window of the environmental data type at the future monitoring moment according to the target quantity eigenvalue of the environmental data type at the future monitoring moment, and input the data window into the prediction model of the environmental data type to output the predicted environmental data of the environmental data type at the future monitoring moment.

[0016] When regulating a distributed new energy power station currently, the power generation of the distributed new energy power station is usually predicted based on the predicted environmental data of the location where the distributed new energy power station is located, and then the power transmission of the distributed new energy power station to the power grid is regulated based on the prediction result; however, when predicting the environmental data of the location where the distributed new energy power station is located currently, historical environmental data of a fixed length or historical environmental data of a random length are generally selected to predict the future environmental data at the monitoring location. However, this process of predicting future environmental data with historical environmental data of a fixed length or historical environmental data of a random length will not only ignore the impact of environmental change characteristics on future environmental prediction, but also be easily interfered by early invalidated data, resulting in prediction lag or increased error. That is, when using historical environmental data of a fixed length or historical environmental data of a random length as the input for prediction to predict future environmental data, the model for predicting environmental data will have a weak response ability to environmental changes, a weak ability to capture environmental changes, and a high possibility that the prediction result of the model for predicting environmental data is interfered by early invalidated data. As a result, when predicting the future environment of the location where the distributed new energy power station is located, problems of unreliable and inaccurate environmental prediction will occur. For example, when predicting the environmental data at the monitoring location at a certain moment, and the environmental data close to this moment is more volatile than the historical environmental data at a distance from this moment, that is, the environment at the monitoring location at this moment has a high volatility trend. Then, if historical environmental data of a longer length is still selected as the input of the prediction model when predicting the environmental data at the monitoring location at this moment, the change characteristics of the environment at the monitoring location at this moment cannot be captured during prediction, resulting in difficulty in capturing the immediate impact of sudden environmental changes on power generation when predicting the power generation. As a result, when predicting the power generation of the distributed new energy power station based on the predicted environmental data subsequently, problems of unreliable and inaccurate prediction will occur. When there are problems with inaccurate prediction of the power generation of the distributed new energy power station, the subsequent regulation effect of the electric energy output by the distributed new energy power station to the power grid will be poor. And when the regulation effect of the electric energy output by the distributed new energy power station to the power grid is poor, a series of consequences such as damage to the stability of the power grid operation and losses in the economy and market will be caused. That is, the problem of inaccurate power generation prediction will affect the power grid stability, economic benefits, etc.

[0017] In order to ensure the effect of regulating the electric energy output from the subsequent distributed new energy power station to the power grid, it is necessary to improve the prediction accuracy and reliability of the power generation of the distributed new energy power station in the subsequent process. The prediction accuracy and reliability of the power generation of the distributed new energy power station are related to the prediction reliability and accuracy of the environmental data at the monitoring location. Therefore, in the subsequent process of this embodiment, the prediction reliability and accuracy of the environmental data at the monitoring location will be improved to enhance the prediction accuracy of the power generation of the distributed new energy power station. And this embodiment mainly adaptively determines the length of the historical data used for prediction by analyzing the characteristics of the environmental data close to the prediction moment, so as to improve the prediction accuracy of the power generation of the distributed new energy power station. That is to say, this embodiment improves the response ability of the prediction model for environmental data prediction to environmental changes, the ability to capture environmental changes, and the ability to capture the periodic change characteristics of data by adaptively regulating the data window used for prediction, so as to improve the prediction reliability and accuracy of the environmental data at the monitoring location.

[0018] Based on the above analysis, it can be seen that before obtaining the predicted environmental data in this embodiment, it is necessary to adaptively obtain the data window for environmental data prediction, that is, the data window at the future monitoring moment. The data window at the future monitoring moment is mainly used for predicting the environmental data at the monitoring location at the future monitoring moment. Also, since the types of environmental data affecting the power generation of the distributed new energy power station are diverse, and different types of environmental data will have different characteristics at the same moment, there will also be differences in the sizes of the data windows used for predicting environmental data belonging to different environmental data types. Therefore, in the subsequent process of this embodiment, it is necessary to obtain the data windows of each environmental data type at the future monitoring moment; in addition, since the acquisition methods of the data windows of all environmental data types at any monitoring moment are the same, for the convenience of understanding and analysis in this embodiment, the subsequent description will take the acquisition process of the data window of any environmental data type W at the future monitoring moment as an example. And the future monitoring moment in this embodiment mainly refers to the next monitoring moment of the current monitoring moment. Then the specific acquisition process of the data window of the environmental data type W at the future monitoring moment is as follows: First, perform STL decomposition on the environmental data sequence corresponding to environmental data type W at the current monitoring moment to obtain the seasonal component and residual component of each environmental data in the environmental data sequence corresponding to environmental data type W at the previous monitoring moment. Then, construct the seasonal mapping space and residual mapping space corresponding to environmental data type W. After that, map the seasonal components of all environmental data in the environmental data sequence corresponding to environmental data type W at the current monitoring moment and the acquisition time of the corresponding environmental data into the seasonal mapping space to obtain the seasonal component data points in the seasonal mapping space, and connect all the seasonal component data points in the seasonal mapping space in chronological order. The connected curve is denoted as the seasonal component curve of the environmental data sequence corresponding to environmental data type W at the current monitoring moment. Map the residual components of all environmental data in the environmental data sequence corresponding to environmental data type W at the current monitoring moment and the acquisition time of the corresponding environmental data into the residual mapping space to obtain the residual component data points in the residual mapping space, and connect all the residual component data points in the residual mapping space in chronological order. The connected curve is denoted as the residual component curve of the environmental data sequence corresponding to environmental data type W at the current monitoring moment. That is, the horizontal axis of the seasonal mapping space corresponding to environmental data type W is time, and the vertical axis is the seasonal component; the horizontal axis of the residual mapping space is time, and the vertical axis is the residual component. And the process of performing STL decomposition on the data is a well-known technology, so it will not be described in detail here. After obtaining the residual component curve and the seasonal component curve, the extreme points on the seasonal component curve of the environmental data sequence corresponding to environmental data type W at the current monitoring moment are obtained. The extreme points are obtained by using the derivative method. According to the extreme points on the seasonal component curve of the environmental data sequence corresponding to environmental data type W at the current monitoring moment, the extreme moment combination corresponding to environmental data type W at the current monitoring moment is obtained. Then, according to the difference between the two extreme moments in the extreme moment combination corresponding to environmental data type W at the current monitoring moment and the position on the residual component curve of the environmental data sequence corresponding to environmental data type W at the current monitoring moment where the abscissa belongs to the extreme moment, the target quantity characteristic value of environmental data type W at the future monitoring moment is obtained. And according to the target quantity characteristic value of environmental data type W at the future monitoring moment, the data window of environmental data type W at the future monitoring moment is obtained. The target quantity characteristic value of environmental data type W at the future monitoring moment is positively correlated with the data window of environmental data type W at the future monitoring moment. Therefore, the target quantity characteristic value can be used as the basis for determining the data window.

[0019] In this embodiment, the specific process of obtaining the extreme value moment combination corresponding to the environmental data type W at the current monitoring moment according to the extreme value points on the seasonal component curve of the environmental data sequence corresponding to the environmental data type W at the current monitoring moment is as follows: First, sort the abscissa values of all the extreme value points on the seasonal component curve of the environmental data sequence corresponding to the environmental data type W at the current monitoring moment in chronological order, and denote the sorted sequence as the extreme value moment sequence. The extreme value points include maximum value points and minimum value points. If a certain extreme value moment in the extreme value moment sequence is the abscissa value of a certain extreme value point on the seasonal component curve, and this extreme value point is mapped according to the seasonal component of a certain environmental data in the environmental data sequence corresponding to the environmental data type W at the current monitoring moment and the acquisition time of this environmental data, then the value of this extreme value moment is the acquisition time of this environmental data. Then, construct combinations of any two adjacent extreme value moments in the extreme value moment sequence, and denote all the combinations constructed by two adjacent extreme value moments in the extreme value moment sequence as the extreme value moment combination corresponding to the environmental data type W at the current monitoring moment. Moreover, the number of the extreme value moment combinations corresponding to the environmental data type W at the current monitoring moment is the number of extreme value moments in the extreme value moment sequence minus 1. The construction of the extreme value moment combination is for subsequent analysis of data characteristics.

[0020] In this embodiment, the specific process of obtaining the target quantity characteristic value of environmental data type W at the future monitoring time is as follows: First, according to the time interval between the two extreme value times in the extreme value time combination corresponding to environmental data type W at the current monitoring time and the time interval between the two extreme value times in any extreme value time combination, the initial quantity characteristic value of environmental data type W at the future monitoring time is obtained; Since there may be unpredictable weather changes or sudden weather changes, and when such unpredictable weather changes or sudden weather changes occur, the data window length should be further reduced so that the prediction model for environmental data prediction pays more attention to local historical data. Therefore, in order to further improve the prediction accuracy of environmental data in this embodiment, this embodiment then needs to analyze the random fluctuation characterization value of environmental data type W at the future monitoring time, and further adjust the initial quantity characteristic value based on the obtained random fluctuation characterization value. That is, in this embodiment, first, on the residual component curve of the environmental data sequence corresponding to environmental data type W at the current monitoring time, the residual component curve segment corresponding to each extreme value time combination of environmental data type W at the current monitoring time is obtained, and the combination formed by the abscissa values of the two endpoints of the residual component curve segment corresponding to the extreme value time combination is the corresponding extreme value time combination. That is, if the two extreme value times in an extreme value time combination are t1 and t2 respectively, then the residual component curve segment from the data point with abscissa t1 to the data point with abscissa t2 on the residual component curve is the residual component curve segment corresponding to this extreme value time combination; Then, according to the mean value of all absolute values of the residual components on the residual component curve segment corresponding to each extreme value time combination of environmental data type W at the current monitoring time and the standard deviation of all residual components on the residual component curve segment corresponding to the extreme value time combination, the random fluctuation characterization value of environmental data type W at the future monitoring time is obtained, and the larger the random fluctuation characterization value, the greater the degree of adjustment to the initial quantity characteristic value, that is, the larger the random fluctuation characterization value, the shorter the data window length obtained subsequently; Finally, the result of multiplying the preset first constant minus the random fluctuation characterization value of environmental data type W at the future monitoring time by the initial quantity characteristic value of environmental data type W at the future monitoring time is obtained and used as the target quantity characteristic value of environmental data type W at the future monitoring time. That is, the target quantity characteristic value of environmental data type W at the future monitoring time is , where is the initial quantity characteristic value of environmental data type W at the future monitoring time, C1 is the preset first constant. Since ranges from 0 to 1, C1 is set to 1 in this embodiment. It is the random fluctuation characterization value of environmental data type W at a future monitoring time.

[0021] In this embodiment, the specific process of obtaining the initial quantity characteristic value of environmental data type W at a future monitoring time is as follows: First, obtain the time interval between two extreme times in the extreme time combination and denote it as the distance value corresponding to the extreme time combination. Obtain the middle time of the time period formed by the two extreme times in the extreme time combination and denote it as the characterization time corresponding to the extreme time combination. That is, if the two extreme times in a certain extreme time combination are t1 and t2 respectively, the middle time between t1 and t2 is the characterization time of this extreme time combination. Then calculate the normalized value of the reciprocal of the time interval between the characterization time of the extreme time combination and the current time, and use it as the distance weight value corresponding to the extreme time combination. And if the reciprocal of the time interval between the characterization time of the extreme time combination and the current time is denoted as the initial weight value of the extreme time combination, then the result of normalizing the initial weight value of the extreme time combination is the cumulative result of the initial weight value of the extreme time combination and the initial weight values of all extreme time combinations corresponding to environmental data type W at the current monitoring time. That is, since in the subsequent calculation of the stability index value and the random fluctuation characterization value, the closer the extreme time combination is to the current monitoring time, the greater its reference value, so the greater the distance weight value of the extreme time combination closer to the current monitoring time. Then, among all the extreme time combinations corresponding to environmental data type W at the current monitoring time, obtain the extreme time combination that is closest to the current monitoring time in terms of time and denote it as the nearest neighbor combination corresponding to environmental data type W at the current monitoring time. Immediately afterwards, obtain the weighted difference value of the extreme time combination. And the weighted difference value corresponding to any extreme time combination is the result of multiplying the absolute value of the difference between the distance value corresponding to this extreme time combination and the distance value corresponding to the nearest neighbor combination by the distance weight value of this extreme time. That is, the weighted difference value corresponding to the s-th extreme time combination corresponding to environmental data type W at the current monitoring time is where T0 is the distance value corresponding to the nearest neighbor combination corresponding to environmental data type W at the current monitoring time, is the distance value corresponding to the s-th extreme time combination corresponding to environmental data type W at the current monitoring time, is the distance weight value of the s-th extreme time combination corresponding to environmental data type W at the current monitoring time. Then obtain the negative correlation mapping value of the result obtained by accumulating the weighted difference values of all extreme time combinations corresponding to environmental data type W at the current monitoring time, and denote it as the stability index value corresponding to environmental data type W at the current monitoring time. That is, the stability index value corresponding to environmental data type W at the current monitoring time is , exp() is the exponential function with the constant e as the base. And when the stability index value corresponding to environmental data type W at the current monitoring time is smaller, that is The larger it is, it indicates that for the environmental data belonging to the environmental data type W, the environmental data with a collection time closer to the current monitoring moment has a higher degree of fluctuation or a lower degree of stability relative to the overall historical data. Then, the data window length of the environmental data type W at the subsequent future monitoring moment should be shorter to avoid relying on too much historical data and improve the local prediction response ability. When the stability index value corresponding to the environmental data type W at the current monitoring moment is larger, that is The smaller it is, it indicates that for the environmental data belonging to the environmental data type W, the stability of the overall historical data is higher. Then, the data window length of the environmental data type W at the subsequent future monitoring moment should be longer to improve the reference to historical data and thus improve the accuracy of prediction; then obtain the mean value of the distance values corresponding to all extreme moment combinations corresponding to the environmental data type W at the current monitoring moment and denote it as the periodic index value corresponding to the environmental data type W at the current monitoring moment. And since the environmental data has a certain periodic change, in order to be able to more completely capture the periodic characteristics of the data during subsequent prediction and improve the accuracy of the prediction result, in this embodiment, the window of the data with a faster periodic change should be shorter, and the window of the data with a slower periodic change should be shorter. And when the mean value of the distance values corresponding to all extreme moment combinations corresponding to the environmental data type W at the current monitoring moment is larger, it indicates that the change period of the environmental data belonging to the environmental data type W is slower. When the mean value of the distance values corresponding to all extreme moment combinations corresponding to the environmental data type W at the current monitoring moment is smaller, it indicates that the change period of the environmental data belonging to the environmental data type W is faster; finally, obtain the product of the stability index value corresponding to the environmental data type W at the current monitoring moment and the periodic index value corresponding to the environmental data type W at the current monitoring moment and denote it as the initial quantity characteristic value of the environmental data type W at the future monitoring moment.

[0022] Based on the above, when the initial quantity eigenvalue is smaller, it indicates that for the data closer to the future monitoring moment compared to the data farther from the future monitoring moment, the change of the data closer to the future monitoring moment is more unstable. It also indicates that the data change period of the data belonging to environmental data type W at the monitoring position is faster. Then, when predicting the data belonging to environmental data type W at the future monitoring moment, only by focusing on using the data closer to the future monitoring moment can the recent change trend and periodic characteristics of environmental data type W be captured more accurately. That is, only by focusing on using the data closer to the future monitoring moment can the prediction accuracy of the data belonging to environmental data type W at the future monitoring moment be improved. It also shows that at this time, a shorter data window should be used to better capture the recent change characteristics of environmental data type W at the monitoring position, or at this time, a shorter data window should be used to predict the data belonging to environmental data type W at the future monitoring moment. When the initial quantity eigenvalue is larger, it indicates that from the start of monitoring environmental data type W to the current monitoring moment, the overall change of the data belonging to environmental data type W is relatively stable and the data change period is longer. Then, when predicting the data belonging to environmental data type W at the future monitoring moment, a longer data window should be used to ensure the prediction accuracy of environmental data type W at the future monitoring moment. And when the data change period is longer, using a longer data window for prediction can more completely capture the data periodic characteristics, making the prediction result more accurate.

[0023] In this embodiment, the specific process of obtaining the random fluctuation characterization value of environmental data type W at the future monitoring moment is as follows: First, obtain the mean value of the absolute values of all residual components on the residual term curve segment corresponding to the extreme value moment combination, and record it as the residual mean corresponding to the extreme value moment combination. Obtain the standard deviation of all residual components on the residual term curve segment corresponding to the extreme value moment combination, and record it as the residual standard deviation corresponding to the extreme value moment combination. And the residual component on the residual term curve segment corresponding to the extreme value moment combination refers to the ordinate value of the residual component data point on the residual term curve segment. Then, obtain the product of the standard deviation corresponding to the extreme value moment combination and the residual mean corresponding to the extreme value moment combination, and record it as the residual eigenvalue corresponding to the extreme value moment combination. After that, obtain the result of multiplying the result of comparing the residual eigenvalue corresponding to the extreme value moment combination with the residual eigenvalue corresponding to the nearest neighbor combination by the distance weight value corresponding to the extreme value moment combination, and record it as the random eigenvalue corresponding to the extreme value moment combination. That is, the random eigenvalue of the s-th extreme value moment combination corresponding to environmental data type W at the current monitoring moment is , where is the residual standard deviation of the nearest neighbor combination corresponding to environmental data type W at the current monitoring moment, F0 is the residual mean of the nearest neighbor combination corresponding to environmental data type W at the current monitoring moment, is the residual standard deviation of the s-th extreme moment combination corresponding to the environmental data type W at the current monitoring moment. is the residual mean of the s-th extreme moment combination corresponding to the environmental data type W at the current monitoring moment; finally, the normalized value of the sum of the random eigenvalues of all extreme moment combinations corresponding to the environmental data type W at the current monitoring moment is used as the random fluctuation characterization value of the environmental data type W at the current monitoring moment. At this time, the normalization function Norm() is used to normalize the sum of the random eigenvalues of all extreme moment combinations corresponding to the environmental data type W at the current monitoring moment.

[0024] Moreover, when the random fluctuation characterization value of the environmental data type W at the future monitoring moment is larger, it indicates that there is a greater possibility of an unpredictable weather change or a sudden weather change in the local time range where the current monitoring moment is located, such as sandstorms, rain, etc. Then, when predicting the environmental data belonging to the environmental data type W at the next monitoring moment of the current monitoring moment, it is necessary to focus on using the data closer to the future monitoring moment to accurately capture this change. That is, only by focusing on using the data closer to the future monitoring moment can the prediction accuracy of the data belonging to the environmental data type W at the future monitoring moment be improved. It also means that at this time, a shorter data window should be used to better capture the recent change characteristics of the data belonging to the environmental data type W at the monitoring position or, in other words, at this time, a shorter data window should be used to predict the data belonging to the environmental data type W at the future monitoring moment. Therefore, when the random fluctuation characterization value of the environmental data type W at the future monitoring moment is larger, the adjustment degree of the initial quantity eigenvalue of the environmental data type W at the future monitoring moment is larger; when the random fluctuation characterization value of the environmental data type W at the future monitoring moment is smaller, it indicates that there is a smaller possibility of an unpredictable weather change or a sudden weather change in the local time range where the current monitoring moment is located. Then, in order to ensure the prediction accuracy at this time, the adjustment degree of the initial quantity eigenvalue of the environmental data type W at the future monitoring moment should be smaller, that is, at this time, the target quantity eigenvalue close to the initial quantity eigenvalue can be used to obtain the data window; in this embodiment, the characteristics shown by the extreme moment combination closest to the current monitoring moment can better reflect the recent change trend of the environment.

[0025] In this embodiment, the specific process of obtaining the data window of environmental data type W at the future monitoring moment according to the target quantity characteristic value of environmental data type W at the future monitoring moment is as follows: First, a preset number of monitoring moments are selected as sample moments before the future monitoring moment, and the target quantity characteristic value of environmental data type W at each sample moment is obtained. In this embodiment, the implementer can determine the value of the preset number according to the total number of monitoring moments before the future monitoring moment. For example, in this embodiment, the preset number can be set to 70% of the total number of monitoring moments before the future monitoring moment, and when selecting sample moments, they are selected according to the distance from the future monitoring moment, that is, the monitoring data closer to the future monitoring moment is selected first; in addition, the method for obtaining the target quantity characteristic value of environmental data type W at the sample moment is the same as the method for obtaining the target quantity characteristic value of environmental data type W at the future monitoring moment, so it will not be described in detail in this embodiment; then, according to the target quantity characteristic value of environmental data type W at the sample moment, the target intercept and target slope corresponding to environmental data type W are obtained; then, according to the target intercept and target slope corresponding to environmental data type W, a function is constructed with the intercept being the target intercept corresponding to environmental data type W, the slope being the target slope corresponding to environmental data type W, the input being the target quantity characteristic value of environmental data type W, and the output being the data length of environmental data type W, and it is denoted as the data length function corresponding to environmental data type W; then, according to the data length function corresponding to environmental data type W and the target quantity characteristic value of environmental data type W at the future monitoring moment, the data window of environmental data type W at the future monitoring moment is obtained.

[0026] In this embodiment, the specific process of obtaining the data window of environmental data type W at the future monitoring time according to the data length function corresponding to the environmental data type W and the target quantity characteristic value of the environmental data type W at the future monitoring time is as follows: Input the target quantity characteristic value of the environmental data type W at the future monitoring time into the data length function corresponding to the environmental data type W, and denote the ceiling value of the result output by the data length function corresponding to the environmental data type W at this time as the data length of the environmental data type W at the future monitoring time; then start traversing forward from the current monitoring time until the traversed environmental data length belonging to the environmental data type W is equal to the data length of the environmental data type W at the future monitoring time, and stop traversing, and denote the window composed of all the environmental data belonging to the environmental data type W traversed as the data window of the environmental data type W at the future monitoring time, that is, if the current monitoring time is the Z-th monitoring time and the data length of the environmental data type W at the future monitoring time is L, then the data window composed of the environmental data belonging to the environmental data type W collected at all monitoring times from the current monitoring time to the (Z - L)-th monitoring time is the data window of the environmental data type W at the future monitoring time, including the data collected at the current monitoring time and the (Z - L)-th monitoring time.

[0027] In this embodiment, the specific process of obtaining the target intercept and target slope corresponding to the environmental data type W is as follows: First, a random intercept and a random slope are selected. Since the obtained target quantity eigenvalue is positively correlated with the data length in this embodiment, the slope value in this embodiment is required to be a positive number, and the intercept value is a natural number. Then, the combination of the randomly selected intercept and slope is denoted as the parameter combination corresponding to the first iteration under the environmental data type W, and the optimal index value corresponding to the parameter combination corresponding to the first iteration under the environmental data type W is obtained. After that, the intercept and slope in the parameter combination corresponding to the first iteration under the environmental data type W are updated by using the parameter update method in the Bayesian optimization algorithm, and the updated result is denoted as the parameter combination corresponding to the second iteration under the environmental data type W. Then, the optimal index value corresponding to the parameter combination corresponding to the second iteration under the environmental data type W is continuously obtained, and then the intercept and slope in the parameter combination corresponding to the second iteration under the environmental data type W are continuously updated by using the parameter update method in the Bayesian optimization algorithm, and the updated result is denoted as the parameter combination corresponding to the third iteration under the environmental data type W. Then, the optimal index value corresponding to the parameter combination corresponding to the third iteration under the environmental data type W is continuously obtained, and so on, until the iteration number reaches the preset iteration number and stops. Then, the parameter combination corresponding to each iteration under the environmental data type W and the optimal index value corresponding to the parameter combination corresponding to each iteration under the environmental data type W are counted. Then, among the parameter combinations corresponding to all iterations under the environmental data type W, the slope and intercept in the parameter combination corresponding to the maximum optimal index value are selected as the target intercept and target slope corresponding to the environmental data type W. And since the process of parameter optimization using the Bayesian optimization algorithm is a well-known technology, no specific description is given in this embodiment. In addition, the preset iteration number of the iteration stop condition needs to be set according to the size of the parameter value range, for example, it can be set to 50 times.

[0028] In this embodiment, the specific process of the optimal index value of the parameter combination corresponding to the i-th iteration under the environmental data type W is as follows: First, the slope and intercept in the parameter combination corresponding to the i-th iteration under the environmental data type W are respectively denoted as the experimental slope and the experimental intercept, and the actual data at each sample time and the predicted data at each sample time under the parameter combination corresponding to the i-th iteration are obtained. And at this time, the actual data at any sample time t is the data in the environmental data sequence corresponding to the environmental data type W at the current monitoring time whose acquisition time is equal to the sample time t, that is, the actual data at any sample time t mentioned here is the environmental data belonging to the environmental data type W collected at the sample time t. Then, the mean square error is calculated according to the actual data at all sample times and the predicted data at the sample time t under the parameter combination corresponding to the i-th iteration, and is denoted as the optimal index value of the parameter combination corresponding to the i-th iteration under the environmental data type W. That is, the optimal index value of the parameter combination corresponding to the i-th iteration under the environmental data type W is , where J is the number of sample times, is the actual data at the j-th sample time among all sample times, and is also the environmental data belonging to the environmental data type W collected at the j-th sample time, is the predicted data at the j-th sample time under the parameter combination corresponding to the i-th iteration, and is also the prediction result of the data belonging to the environmental data type W at the j-th sample time when the slope and intercept belong to the parameters in the parameter combination corresponding to the i-th iteration; and the target intercept and target slope determined by parameter optimization through the Bayesian optimization algorithm can make the subsequent obtained prediction results more reliable and accurate.

[0029] In this embodiment, the specific process for obtaining the predicted data at each sample time under the parameter combination corresponding to the i-th iteration is as follows: The ceiling value of the result obtained by multiplying the target quantity feature value of the environmental data type W at the sample time t by the experimental slope and then adding the experimental intercept is denoted as T; in the environmental data sequence corresponding to the environmental data type W at the previous and current monitoring times of the sample time t, a continuous data segment with a collection time equal to the sample time t and a length of T is obtained, and is denoted as the experimental data window of the sample time t under the parameter combination corresponding to the i-th iteration, that is, if the sample time t is the m-th monitoring time in the time period from the start of monitoring the environmental data type W to the current monitoring time, and the collection time of the m-th data in the environmental data sequence corresponding to the environmental data type W at the current monitoring time is the m-th monitoring time, and the collection time of the m-T-th data in the environmental data sequence corresponding to the environmental data type W at the current monitoring time is the m-T-th monitoring time, then the data in the experimental data window of the sample time t under the parameter combination corresponding to the i-th iteration is composed of the data from the m-th data to the m-T-th data in the environmental data sequence corresponding to the environmental data type W at the current monitoring time; the experimental data window of the sample time t under the parameter combination corresponding to the i-th iteration is input into the prediction model of the environmental data type W, and the output result of the prediction model of the environmental data type W at this time is the predicted data of the sample time t under the parameter combination corresponding to the i-th iteration.

[0030] And it should be noted that the prediction model of the environmental data type W is a trained LSTM model, and the input of the prediction model of the environmental data type W is the historical data belonging to the environmental data type W, and the output is the predicted data belonging to the environmental data type W, that is, the data predicted by the prediction model of the environmental data type W belongs to the environmental data type W.

[0031] After obtaining the data window of environmental data type W at the future monitoring moment, the data window of environmental data type W at the future monitoring moment is input into the prediction model of environmental data type W, and the result output by the prediction model of environmental data type W at this time is recorded as the predicted environmental data of environmental data type W at the future monitoring moment. That is, the predicted environmental data of environmental data type W at the future monitoring moment refers to the predicted environmental data at the location of the distributed new energy power station at the future monitoring moment. And in this embodiment, the future monitoring moments all refer to the monitoring moment adjacent to the current monitoring moment and located behind the current monitoring moment in time, that is, the next monitoring moment of the current monitoring moment.

[0032] Therefore, through the above process, this embodiment obtains the predicted environmental data of environmental data type W at the future monitoring moment. And since the acquisition methods of the predicted environmental data of different environmental data types at the future monitoring moment are the same as the acquisition method of the predicted environmental data of environmental data type W at the future monitoring moment, this embodiment can also obtain the predicted environmental data of each environmental data type in the environmental data type set at the future monitoring moment through the above process.

[0033] Step S003: Obtain the predicted power generation of the distributed new energy power station according to the predicted environmental data of different environmental data types and the power generation prediction model, and regulate the distributed new energy power station according to the predicted power generation.

[0034] After obtaining the predicted environmental data of each environmental data type in the environmental data type set at the future monitoring moment, the predicted environmental data of each environmental data type in the environmental data type set at the future monitoring moment is input into the power generation prediction model, and the result output by the power generation prediction model at this time is recorded as the predicted power generation of the distributed new energy power station at the future monitoring moment. And the power generation prediction model in this embodiment is also a trained LSTM model. However, there are differences in the input and output between the power generation prediction model and the above prediction model of environmental data types. The input of the power generation prediction model is the predicted environmental data and the output is the predicted power generation, while the input of the prediction model of environmental data type W is the historical environmental data and the output is the predicted environmental data. And since the training process of the LSTM model is a well-known technology, this embodiment will not describe the training process of the model anymore.

[0035] Subsequently, the distributed new energy power station is regulated according to the predicted power generation of the distributed new energy power station at the future monitoring moment. Specifically, the power output from the distributed new energy power station to the power grid is regulated mainly based on the predicted power generation and the load required in the power grid. Since the method for regulating the power output from the distributed new energy power station connected to the power grid when the predicted power generation of the distributed new energy power station and the load required in the power grid are known is a well-known technology, it will not be described in this embodiment.

[0036] Thus, this embodiment has completed the regulation of the distributed new energy power station.

[0037] In summary, this embodiment first obtains the environmental data sequences corresponding to different environmental data types at the current monitoring moment; then obtains the extreme point on the seasonal component curve of the environmental data sequence to get the extreme moment combination corresponding to the environmental data type at the current monitoring moment. Based on the difference between the two extreme moments in the extreme moment combination and the position of the abscissa belonging to the extreme moment on the residual component curve of the environmental data sequence, the target quantity feature value of the environmental data type at the future monitoring moment is obtained. And according to the target quantity feature value of the environmental data type at the future monitoring moment, the data window of the environmental data type at the future monitoring moment is obtained. The data window is input into the prediction model of the environmental data type to output the predicted environmental data of the environmental data type at the future monitoring moment; according to the predicted environmental data of different environmental data types and the power generation prediction model, the predicted power generation of the distributed new energy power station is obtained, and the distributed new energy power station is regulated according to the predicted power generation. By adaptively obtaining the data window, this embodiment can improve the accuracy of predicting the power generation of the distributed new energy power station, thus ensuring the effect of subsequent regulation of the electric energy output from the distributed new energy power station to the power grid, and further avoiding the phenomenon of damage to the power grid operation stability and losses in the economy and market as much as possible.

[0038] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A smart regulation method based on a distributed new energy power station, characterized in that, The method includes the following steps: Obtain environmental data sequences corresponding to different environmental data types at the current monitoring moment, and the data in the environmental data sequences are environmental data collected at the location of the distributed new energy power station; For any environmental data type, obtain the seasonal component curve and the residual component curve of the environmental data sequence corresponding to the environmental data type, and obtain the extreme moment combination corresponding to the environmental data type at the current monitoring moment according to the extreme points on the seasonal component curve. According to the difference between two extreme moments in the extreme moment combination and the positions of the abscissas belonging to the extreme moments on the residual component curve, obtain the target quantity characteristic value of the environmental data type at the future monitoring moment, and obtain the data window of the environmental data type at the future monitoring moment according to the target quantity characteristic value of the environmental data type at the future monitoring moment. Input the data window into the prediction model of the environmental data type to output the predicted environmental data of the environmental data type at the future monitoring moment; Obtain the predicted power generation of the distributed new energy power station according to the predicted environmental data of different environmental data types and the power generation prediction model, and regulate the distributed new energy power station according to the predicted power generation.

2. The intelligent control method based on a distributed new energy power station according to claim 1, wherein, The method for obtaining the extreme moment combination includes: Denote the sequence of the abscissa values of all extreme points on the seasonal component curve arranged in chronological order as the extreme moment sequence, and denote all combinations constructed by two adjacent extreme moments in the extreme moment sequence as the extreme moment combination corresponding to the environmental data type at the current monitoring moment.

3. A smart regulation method based on a distributed new energy power station according to claim 1, characterized in that The method for obtaining the target quantity characteristic value of the environmental data type at the future monitoring moment includes: According to the time interval between two extreme moments in the extreme moment combination closest to the current monitoring moment in time and the time interval between two extreme moments in any extreme moment combination, obtain the initial quantity characteristic value of the environmental data type at the future monitoring moment; on the residual component curve, obtain the residual component curve segment corresponding to each extreme moment combination, and the combination formed by the endpoint abscissas of the residual component curve segment corresponding to the extreme moment combination is the corresponding extreme moment combination. According to the mean value of all absolute values of the residual components on the residual component curve segment corresponding to the extreme moment combination and the standard deviation of all residual components on the residual component curve segment corresponding to the extreme moment combination, obtain the random fluctuation characterization value of the environmental data type at the future monitoring moment; use the result of multiplying the preset first constant minus the random fluctuation characterization value by the initial quantity characteristic value of the environmental data type at the future monitoring moment as the target quantity characteristic value of the environmental data type at the future monitoring moment.

4. The intelligent control method based on a distributed new energy power station according to claim 3, characterized in that, The method for obtaining the initial quantity characteristic value of the environmental data type at the future monitoring moment includes: Denote the time interval between two extreme moments in the extreme moment combination as the distance value corresponding to the extreme moment combination, and take the normalized value of the reciprocal of the time interval between the intermediate moment between the two extreme moments in the extreme moment combination and the current monitoring moment as the distance weight value of the corresponding extreme moment combination; among all the extreme moment combinations, denote the extreme moment combination that is closest to the current monitoring moment in time as the nearest neighbor combination; obtain the weighted difference value of the extreme moment combination, where the weighted difference value corresponding to the extreme moment combination is the result of multiplying the absolute value of the difference between the distance value corresponding to the extreme moment combination and the distance value corresponding to the nearest neighbor combination by the distance weight value of the corresponding extreme moment; denote the negative correlation mapping value of the sum of the weighted difference values of all the extreme moment combinations corresponding to the environmental data type at the current monitoring moment as the stability index value corresponding to the environmental data type at the current monitoring moment; take the mean value of the distance values corresponding to all the extreme moment combinations corresponding to the environmental data type at the current monitoring moment as the periodicity index value corresponding to the environmental data type at the current monitoring moment; denote the product of the stability index value and the periodicity index value as the initial quantity characteristic value of the environmental data type at the future monitoring moment.

5. The intelligent control method based on a distributed new energy power station according to claim 4, wherein, A method for obtaining the random fluctuation characterization value of the environmental data type at the future monitoring moment, comprising: Take the mean value of the absolute values of all the residual components on the residual term curve segment corresponding to the extreme moment combination as the residual mean of the corresponding extreme moment combination, take the standard deviation of all the residual components on the residual term curve segment corresponding to the extreme moment combination as the residual standard deviation of the corresponding extreme moment combination, denote the product of the residual standard deviation of the extreme moment combination and the residual mean of the corresponding extreme moment combination as the residual characteristic value of the corresponding extreme moment combination, take the result of multiplying the result of comparing the residual characteristic value of the extreme moment combination with the residual characteristic value of the nearest neighbor combination by the distance weight value of the corresponding extreme moment combination as the random characteristic value of the corresponding extreme moment combination, and take the normalized value of the sum of the random characteristic values of all the extreme moment combinations corresponding to the environmental data type at the current monitoring moment as the random fluctuation characterization value of the environmental data type at the current monitoring moment.

6. The intelligent control method based on a distributed new energy power station according to claim 1, characterized in that, A method for obtaining the data window of the environmental data type at the future monitoring moment according to the target quantity characteristic value of the environmental data type at the future monitoring moment, comprising: Select a preset number of monitoring moments as sample moments before the future monitoring moment, and obtain the target quantity feature values of the environmental data type at the sample moments; according to the target quantity feature values of the environmental data type at the sample moments, obtain the target intercept and target slope corresponding to the environmental data type; according to the target intercept and target slope corresponding to the environmental data type, construct a function with the intercept being the target intercept corresponding to the environmental data type, the slope being the target slope corresponding to the environmental data type, the input being the target quantity feature values of the environmental data type, and the output being the data length, and denote it as the data length function corresponding to the environmental data type; according to the data length function corresponding to the environmental data type and the target quantity feature values of the environmental data type at the future monitoring moment, obtain the data window of the environmental data type at the future monitoring moment.

7. The intelligent control method based on a distributed new energy power station according to claim 6, wherein, The method for obtaining the data window of the environmental data type at the future monitoring moment according to the data length function corresponding to the environmental data type and the target quantity feature values of the environmental data type at the future monitoring moment includes: Input the target quantity feature values of the environmental data type at the current monitoring moment into the data length function corresponding to the environmental data type, and denote the integer value of the output result of the data length function as the data length of the environmental data type at the future monitoring moment; traverse forward from the current monitoring moment until the environmental data length belonging to the environmental data type being traversed is equal to the data length of the environmental data type at the future monitoring moment, and then stop traversing, and denote the window formed by all the environmental data belonging to the environmental data type being traversed as the data window of the environmental data type at the future monitoring moment.

8. A smart regulation method for a distributed new energy power station according to claim 7, characterized in that, The method for obtaining the target intercept and target slope corresponding to the environmental data type includes: Denote the combination constructed by a randomly selected intercept and a slope as the parameter combination corresponding to the first iteration, and obtain the preferred index value of the parameter combination corresponding to the first iteration under the environmental data type. Use the Bayesian optimization algorithm to update the intercept and slope in the parameter combination corresponding to the first iteration to obtain the parameter combination corresponding to the second iteration and the preferred index value of the parameter combination corresponding to the second iteration under the environmental data type. Continue to use the Bayesian optimization algorithm to update the intercept and slope in the parameter combination corresponding to the second iteration to obtain the parameter combination corresponding to the third iteration and the determination index value of the parameter combination corresponding to the third iteration under the environmental data type, and so on, until the iteration times reach the preset iteration times and then stop. Statistically obtain the parameter combination corresponding to each iteration under the environmental data type and the preferred index value of the parameter combination corresponding to each iteration under the environmental data type, and select the slope and intercept in the parameter combination corresponding to the minimum preferred index value as the target intercept and target slope corresponding to the environmental data type.

9. A smart regulation method based on a distributed new energy power station according to claim 8, characterized in that, The method for obtaining the preferred index value of the parameter combination corresponding to any iteration includes: For the i-th iteration under the environmental data type, denote the slope and intercept in the parameter combination corresponding to the i-th iteration under the environmental data type as the experimental slope and the experimental intercept respectively, and obtain the actual data at each sample time and the predicted data at each sample time under the parameter combination corresponding to the i-th iteration. The actual data at any sample time t is the environmental data belonging to the environmental data type collected at the sample time t; Denote the mean square error calculated based on the actual data at all sample times and the predicted data at the sample time t under the parameter combination corresponding to the i-th iteration as the optimal index value of the parameter combination corresponding to the i-th iteration under the environmental data type.

10. A smart control method for a distributed new energy power station according to claim 9, characterized in that, The method for obtaining the predicted data at each sample time under the parameter combination corresponding to the i-th iteration includes: Denote the integer value of the result obtained by multiplying the target quantity characteristic value of the environmental data type at the sample time t by the experimental slope and then adding the experimental intercept as T; In front of the sample time t, obtain a continuous data segment with a collection time adjacent to the sample time t and a length of T, and denote it as the experimental data window of the sample time t under the parameter combination corresponding to the i-th iteration. Input the experimental data window of the sample time t under the parameter combination corresponding to the i-th iteration into the prediction model of the environmental data type, and denote the output result of the prediction model of the environmental data type as the predicted data of the sample time t under the parameter combination corresponding to the i-th iteration.

11. A smart control method for a distributed new energy power station according to claim 1, characterized in that, The method for obtaining the predicted power generation of a distributed new energy power station according to the predicted environmental data of different environmental data types and the power generation prediction model includes: Construct a power prediction model, input the predicted environmental data of all environmental data types at future monitoring times into the power generation prediction model, and denote the result output by the power generation prediction model as the predicted power generation of the distributed new energy power station at future monitoring times.

12. The intelligent control method based on a distributed new energy power station according to claim 1, characterized in that, The power prediction model is an LSTM model.

Images