Distributed photovoltaic generating capacity prediction method and device

Abstract

The invention relates to the technical field of photovoltaic power generation detection, and particularly discloses a distributed photovoltaic power generation capacity prediction method and equipment, and the method comprises the steps: obtaining the installation coordinates of a photovoltaic power generation unit, and building a power generation system diagram according to the installation coordinates; counting the photovoltaic power generation capacity of each day based on the power generation system diagram, and taking the photovoltaic power generation capacity as photovoltaic characteristics; arranging the photovoltaic characteristics according to a time sequence, performing self-recognition on the arranged photovoltaic characteristics, and determining a change period; recording weather information of each day, taking the weather information as input, taking the photovoltaic characteristics as output, and training a neural network model; and predicting photovoltaic characteristics in future time based on the change period and the trained neural network model, and determining the photovoltaic generating capacity. According to the method, the power generation result is jointly predicted in two aspects of weather information and periodicity, the prediction result is accurate, and correspondingly, the detection process based on the prediction process is relatively accurate.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power generation detection, and in particular to a distributed photovoltaic power generation prediction method and device. Background Art

[0002] Distributed photovoltaic power generation refers to photovoltaic power generation systems installed at or near the user's location. Its main features are power generation, grid connection, conversion and use nearby. It is usually used on the roofs of buildings, industrial parks or other dispersed places. Unlike traditional large-scale centralized photovoltaic power generation systems, distributed photovoltaic systems can effectively utilize idle rooftop resources to achieve energy self-sufficiency.

[0003] In the process of distributed photovoltaic power generation, a very important task is the power generation detection process. The existing power generation detection process is mostly based on the detection process of the prediction process. First, a value is predicted. If the value is predicted accurately, it means that the photovoltaic power generation process is highly predictable and naturally in a stable state. Therefore, the quality of the prediction process also affects the accuracy of the detection results. Summary of the invention

[0004] The purpose of the present invention is to provide a distributed photovoltaic power generation prediction method and device to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A distributed photovoltaic power generation prediction method, the method comprising:

[0007] Acquire the installation coordinates of the photovoltaic power generation unit, and create a power generation system diagram according to the installation coordinates; the power generation system diagram is a map composed of photovoltaic power generation units, including distribution information of the photovoltaic power generation units;

[0008] Based on the power generation system diagram, the photovoltaic power generation of each day is counted as the photovoltaic feature;

[0009] Arrange the photovoltaic characteristics according to the time sequence, self-identify the arranged photovoltaic characteristics, and determine the change cycle;

[0010] Record weather information for each day, use weather information as input and photovoltaic characteristics as output to train the neural network model;

[0011] Based on the change cycle and the trained neural network model, the photovoltaic characteristics in the future are predicted to determine the photovoltaic power generation.

[0012] As a further solution of the present invention: the step of obtaining the installation coordinates of the photovoltaic power generation unit and creating a power generation system diagram according to the installation coordinates includes:

[0013] Obtain the center position and boundary coordinates of the photovoltaic power generation unit during installation;

[0014] Obtain the preset system diagram size, and calculate the maximum distance between photovoltaic power generation units according to the center position of the photovoltaic power generation units;

[0015] Determine a scale according to the maximum distance and the preset size, convert the center position and the boundary coordinates into a graphic outline based on the scale, insert it into the system diagram, and obtain a power generation system diagram;

[0016] Among them, the method of obtaining the center position and boundary coordinates of the photovoltaic power generation unit during installation includes reading the registered data and identifying the remote sensing image.

[0017] As a further solution of the present invention: the step of counting the photovoltaic power generation of each day based on the power generation system diagram as the photovoltaic feature includes:

[0018] Receive the instantaneous power generation at each time of the day acquired by the detection module in the photovoltaic power generation unit, and query the corresponding outline in the power generation system diagram;

[0019] Determine a value according to the instantaneous power generation, and insert the value into the point in the corresponding contour; wherein the point selection rule when inserting the value is that when the total value of the value at a certain point reaches a preset threshold, select the next point in a preset order, and the initial point is the pre-selected point;

[0020] When the total value of all points in any contour reaches the threshold, the power generation system diagram is output, and the blank power generation system diagram is read, and the insertion process is executed repeatedly until all instantaneous power generation in a day is filled;

[0021] All power generation system diagrams with output within one day are counted as photovoltaic characteristics.

[0022] As a further solution of the present invention: the step of arranging the photovoltaic characteristics according to the time sequence, self-identifying the arranged photovoltaic characteristics, and determining the change period includes:

[0023] Arrange the PV characteristics of different dates in chronological order;

[0024] Extract the characteristic values ​​of photovoltaic characteristics of each day to obtain a characteristic value sequence;

[0025] Perform periodic identification on the characteristic value sequence to determine the change period and p-value;

[0026] The process of extracting the characteristic value of the photovoltaic characteristics of each day at least includes:

[0027] Converting photovoltaic characteristics into a histogram, and inputting the histogram into a preset conversion model to obtain a characteristic value;

[0028] The periodicity identification process is completed by the scipy.stats submodule in the Python scientific computing library SciPy. The p-value is used to indicate whether the periodicity is strict. The smaller the p-value, the greater the possibility that the data has strict periodicity.

[0029] As a further solution of the present invention: the step of recording weather information of each day, taking weather information as input and photovoltaic characteristics as output, and training the neural network model comprises:

[0030] Record weather information for each day; the weather information includes at least solar radiation, sunshine hours, temperature and cloud cover;

[0031] Take weather information as the independent variable, read the photovoltaic characteristics of the corresponding date as the dependent variable, and construct the training set, test set and validation set;

[0032] The neural network model is trained based on the training set, and the error rate of the neural network model is determined based on the test set and the validation set. When the error rate is less than the preset error threshold, the trained neural network model is output.

[0033] As a further solution of the present invention: the step of predicting photovoltaic characteristics in the future based on the variation cycle and the trained neural network model and determining photovoltaic power generation includes:

[0034] Get weather information for future time based on weather service;

[0035] Sending the weather information to a trained neural network model to output photovoltaic characteristics at a future time;

[0036] Selecting photovoltaic features corresponding to the future time from the historical photovoltaic features based on the period as similar photovoltaic features;

[0037] The photovoltaic characteristics output by the neural network model are modified according to similar photovoltaic characteristics;

[0038] determining photovoltaic power generation according to the photovoltaic characteristics;

[0039] The correction process is:

[0040] Where Z(i,j) is the value at the midpoint (i,j) of the corrected photovoltaic feature, X(i,j) is the value at the midpoint (i,j) of the similar photovoltaic feature, and Y(i,j) is the value at the midpoint (i,j) of the photovoltaic feature output by the neural network model; p is the p-value obtained in the periodicity identification process; p max The preset maximum value.

[0041] The technical solution of the present invention also provides a distributed photovoltaic power generation prediction device, the device comprising:

[0042] A system diagram creation module, used to obtain the installation coordinates of the photovoltaic power generation unit, and create a power generation system diagram according to the installation coordinates; the power generation system diagram is a map composed of photovoltaic power generation units, including distribution information of the photovoltaic power generation units;

[0043] The power generation statistics module is used to count the photovoltaic power generation of each day based on the power generation system diagram as a photovoltaic feature;

[0044] A cycle identification module is used to arrange the photovoltaic characteristics according to the time sequence, self-identify the arranged photovoltaic characteristics, and determine the change cycle;

[0045] The model training module is used to record the weather information of each day, take the weather information as input, and take the photovoltaic characteristics as output to train the neural network model;

[0046] The power generation prediction module is used to predict the photovoltaic characteristics in the future based on the change cycle and the trained neural network model to determine the photovoltaic power generation.

[0047] As a further solution of the present invention: the system diagram creation module includes:

[0048] A coordinate acquisition unit, used to acquire the center position and boundary coordinates of the photovoltaic power generation unit during installation;

[0049] A distance calculation unit, used to obtain a preset system diagram size and calculate a maximum distance between photovoltaic power generation units according to the center position of the photovoltaic power generation units;

[0050] A conversion unit, configured to determine a scale according to the maximum distance and a preset size, convert the center position and the boundary coordinates into a graphic outline based on the scale, insert the contour into the system diagram, and obtain a power generation system diagram;

[0051] Among them, the method of obtaining the center position and boundary coordinates of the photovoltaic power generation unit during installation includes reading the registered data and identifying the remote sensing image.

[0052] As a further solution of the present invention: the power generation statistics module includes:

[0053] A profile query unit, used to receive the instantaneous power generation at each time of the day acquired by the detection module in the photovoltaic power generation unit, and query the corresponding profile in the power generation system diagram;

[0054] A numerical value insertion unit is used to determine a numerical value according to the instantaneous power generation, and insert the numerical value into a point in the corresponding contour; wherein the point selection rule when inserting the numerical value is that when the total value of the numerical value at a certain point reaches a preset threshold, the next point is selected according to a preset order, and the initial point is a pre-selected point;

[0055] The insertion cutoff unit is used to output the power generation system diagram when the total value of all points in any contour reaches the threshold value, read the blank power generation system diagram, and execute the insertion process cyclically until all instantaneous power generation in a day is filled;

[0056] The system diagram statistics unit is used to count all power generation system diagrams output within a day as photovoltaic characteristics.

[0057] As a further solution of the present invention: the cycle identification module includes:

[0058] an arrangement unit for arranging the photovoltaic characteristics of different dates according to the time sequence;

[0059] An extraction unit, used for extracting characteristic values ​​of photovoltaic characteristics of each day to obtain a characteristic value sequence;

[0060] An identification unit, used for periodically identifying the characteristic value sequence, and determining the change period and p-value;

[0061] The process of extracting the characteristic value of the photovoltaic characteristics of each day at least includes:

[0062] Converting photovoltaic characteristics into a histogram, and inputting the histogram into a preset conversion model to obtain a characteristic value;

[0063] The periodicity identification process is completed by the scipy.stats submodule in the Python scientific computing library SciPy. The p-value is used to indicate whether the periodicity is strict. The smaller the p-value, the greater the possibility that the data has strict periodicity.

[0064] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention counts the output electricity of distributed photovoltaic power generation units in the form of a map as a photovoltaic feature. On the one hand, the periodicity is determined according to the changes in the photovoltaic feature itself, and then similar photovoltaic features are selected. On the other hand, the relationship between weather information and photovoltaic features is obtained, and the photovoltaic features are predicted by weather information. Combining weather information and periodicity, an extremely accurate prediction scheme is provided, and accordingly, the detection process based on the prediction process is also relatively accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, 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 some embodiments of the present invention.

[0066] Figure 1 This is a flowchart of the distributed photovoltaic power generation prediction method.

[0067] Figure 2 This is a structural block diagram of the distributed photovoltaic power generation prediction equipment. DETAILED DESCRIPTION

[0068] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0069] Figure 1 The flowchart of the distributed photovoltaic power generation prediction method is as follows. In an embodiment of the present invention, a distributed photovoltaic power generation prediction method is provided, and the method includes:

[0070] Step S100: obtaining the installation coordinates of the photovoltaic power generation unit, and creating a power generation system diagram according to the installation coordinates; the power generation system diagram is a map composed of photovoltaic power generation units, including distribution information of the photovoltaic power generation units;

[0071] Step S200: Counting the photovoltaic power generation of each day based on the power generation system diagram as a photovoltaic feature;

[0072] Step S300: Arrange the photovoltaic characteristics according to the time sequence, perform self-identification on the arranged photovoltaic characteristics, and determine the change cycle;

[0073] Step S400: recording weather information of each day, taking the weather information as input and the photovoltaic characteristics as output, and training the neural network model;

[0074] Step S500: predicting photovoltaic characteristics in the future based on the variation cycle and the trained neural network model, and determining photovoltaic power generation.

[0075] Distributed photovoltaic power generation units are usually used on the roofs of buildings, industrial parks or other decentralized locations. Unlike traditional large-scale centralized photovoltaic power generation systems, distributed photovoltaic systems can effectively utilize idle rooftop resources to achieve energy self-sufficiency, obtain the installation coordinates of photovoltaic power generation units, and use a map to represent the distribution of photovoltaic power generation units to obtain a power generation system diagram.

[0076] When the power generation system diagram is created, the power generation of each photovoltaic power generation unit is obtained in real time, the power generation is converted into a numerical value, and filled into the power generation system diagram. A diagram in which each pixel contains a numerical value can be obtained, which is called a photovoltaic feature; the photovoltaic feature is a power generation system diagram containing a numerical value.

[0077] By arranging the photovoltaic characteristics in date order and extracting the features of the photovoltaic characteristics, at least one characteristic value can be obtained. The characteristic values ​​are counted in time order to obtain a characteristic sequence. The characteristic sequence is identified to determine the cycle. The role of the cycle is to determine which dates have similar photovoltaic characteristics. When prediction is needed, similar photovoltaic characteristics can be queried according to the cycle, and prediction can be made based on the similar photovoltaic characteristics. It should be noted that the more obvious the periodic characteristics are, the simpler the prediction process is, and even the historical photovoltaic characteristics can be directly used as the prediction result.

[0078] On the basis of the above content, since photovoltaic power generation is strongly correlated with weather information, after obtaining the photovoltaic characteristics of each day, the weather information of the corresponding date is obtained. The neural network model is trained with weather information as input and photovoltaic characteristics as output. When prediction is needed, the weather information in the future time is obtained with the help of weather services, and the weather information is input into the trained neural network model to obtain the prediction results.

[0079] In the present invention, the above two prediction schemes are combined. The prediction results are first obtained by the trained neural network model, and then the prediction results are corrected by periodically querying the most similar photovoltaic characteristics. Since the prediction results themselves are photovoltaic characteristics, the correction process can adopt the superposition scheme.

[0080] Finally, the inverse process of step S200 is adopted to convert the corrected photovoltaic characteristics into photovoltaic power generation as the final prediction result.

[0081] Regarding step S100, the step of obtaining the installation coordinates of the photovoltaic power generation unit and creating a power generation system diagram according to the installation coordinates includes:

[0082] Obtain the center position and boundary coordinates of the photovoltaic power generation unit during installation;

[0083] Obtain the preset system diagram size, and calculate the maximum distance between photovoltaic power generation units according to the center position of the photovoltaic power generation units;

[0084] The scale is determined according to the maximum distance and the preset size, and the center position and the boundary coordinates are converted into a graphic outline based on the scale, and inserted into the system diagram to obtain the power generation system diagram.

[0085] Photovoltaic power generation units will be recorded when they are installed. The process of obtaining the center position and boundary coordinates of the photovoltaic power generation units during installation is a simple data reading process. While obtaining the center position, the distance between any two photovoltaic power generation units is calculated to determine the maximum distance. Based on the maximum distance and the preset size, the scale can be determined. The preset size is the size of the map, such as the size of the display screen. Based on the scale, the center position and boundary coordinates are converted into graphic outlines, which are then inserted into the system diagram to obtain the power generation system diagram.

[0086] Among them, the method of obtaining the center position and boundary coordinates of the photovoltaic power generation unit during installation includes reading the registered data and identifying the remote sensing image; in an example of the technical solution of the present invention, when obtaining the center position and boundary coordinates, in addition to reading the registered data, the remote sensing image in the public map service can also be directly used to identify the remote sensing image, so as to obtain the center position and boundary coordinates of the photovoltaic power generation unit.

[0087] Regarding step S200, the step of counting the photovoltaic power generation of each day based on the power generation system diagram as the photovoltaic feature includes:

[0088] Receive the instantaneous power generation at each time of the day acquired by the detection module in the photovoltaic power generation unit, and query the corresponding outline in the power generation system diagram;

[0089] Determine a value according to the instantaneous power generation, and insert the value into the point in the corresponding contour; wherein the point selection rule when inserting the value is that when the total value of the value at a certain point reaches a preset threshold, select the next point in a preset order, and the initial point is the pre-selected point;

[0090] The above content describes the generation process of photovoltaic characteristics. The key point is to convert the instantaneous power generation into a numerical value and then insert it into the corresponding outline of the photovoltaic power generation unit. Each pixel in the power generation system diagram contains a threshold. The process of filling in the numerical value is a cumulative process. The instantaneous power generation is obtained every few seconds or minutes in a day, and the instantaneous power generation is converted into a numerical value (the conversion model can use a direct proportional function), and then the numerical value is filled in from left to right and from top to bottom to each point.

[0091] Under normal circumstances, for example, instantaneous power generation is obtained every few minutes. The larger the photovoltaic power generation unit, the larger the corresponding contour, and the filling process will hardly exceed the range. If the range is exceeded, that is, when the total value of all points in any contour reaches the threshold, the power generation system diagram is output, and the blank power generation system diagram is read, and the insertion process is executed repeatedly until all instantaneous power generation within a day is filled. At the same time, a warning message needs to be generated to indicate that the corresponding photovoltaic power generation unit has produced a very large amount of power.

[0092] Finally, all power generation system diagrams output within a day are counted as photovoltaic characteristics. In most cases, there is only one power generation system diagram output within a day.

[0093] Regarding step S300, the step of arranging the photovoltaic characteristics according to the time sequence, self-identifying the arranged photovoltaic characteristics, and determining the change period includes:

[0094] Arrange the PV characteristics of different dates in chronological order;

[0095] Extract the characteristic values ​​of photovoltaic characteristics of each day to obtain a characteristic value sequence;

[0096] Perform periodic identification on the characteristic value sequence to determine the change period and p-value;

[0097] The process of extracting the characteristic value of the photovoltaic characteristics of each day at least includes:

[0098] Converting photovoltaic characteristics into a histogram, and inputting the histogram into a preset conversion model to obtain a characteristic value;

[0099] The periodicity identification process is completed by the scipy.stats submodule in the Python scientific computing library SciPy. The p-value is used to indicate whether the periodicity is strict. The smaller the p-value, the greater the possibility that the data has strict periodicity.

[0100] The photovoltaic characteristics themselves are graphs containing numerical values, and the process of comparing them and determining the period is relatively difficult. Therefore, in an example of the technical solution of the present invention, the characteristic values ​​of the photovoltaic characteristics of each day are first extracted, the characteristic values ​​are sorted to obtain a characteristic value sequence, and the characteristic value sequence is periodically identified to determine whether there is obvious periodicity and what its period is.

[0101] It should be noted that scipy.stats is a submodule of the Python scientific computing library SciPy, which provides a large number of statistical test functions. For periodicity detection, scipy.stats itself does not have a function specifically designed for periodicity detection, but periodicity can be indirectly evaluated through some common statistical test methods. These common statistical test methods belong to conventional prior art and will not be described in detail in this application.

[0102] Regarding step S400, the steps of recording weather information of each day, taking weather information as input and photovoltaic characteristics as output, and training the neural network model include:

[0103] Record weather information for each day; the weather information includes at least solar radiation, sunshine hours, temperature and cloud cover;

[0104] Take weather information as the independent variable, read the photovoltaic characteristics of the corresponding date as the dependent variable, and construct the training set, test set and validation set;

[0105] The neural network model is trained based on the training set, and the error rate of the neural network model is determined based on the test set and the validation set. When the error rate is less than the preset error threshold, the trained neural network model is output.

[0106] With the help of existing weather measurement stations, weather information is obtained and recorded, weather information is used as the independent variable, photovoltaic characteristics are used as the dependent variable to construct data pairs, and all data pairs are counted to obtain training sets, test sets and validation sets. The ratio of training sets, test sets and validation sets can be 9:0.5:0.5.

[0107] The neural network model is trained based on the training set, and the error rate of the neural network model is determined based on the test set and the validation set. When the error rate is small enough, the trained neural network model is output.

[0108] Regarding step S500, the step of predicting photovoltaic characteristics in the future based on the variation cycle and the trained neural network model and determining photovoltaic power generation includes:

[0109] Get weather information for future time based on weather service;

[0110] Sending the weather information to a trained neural network model to output photovoltaic characteristics at a future time;

[0111] Selecting photovoltaic features corresponding to the future time from the historical photovoltaic features based on the period as similar photovoltaic features;

[0112] The photovoltaic characteristics output by the neural network model are modified according to similar photovoltaic characteristics;

[0113] determining photovoltaic power generation according to the photovoltaic characteristics;

[0114] The correction process is:

[0115] Where Z(i,j) is the value at the midpoint (i,j) of the corrected photovoltaic feature, X(i,j) is the value at the midpoint (i,j) of the similar photovoltaic feature, and Y(i,j) is the value at the midpoint (i,j) of the photovoltaic feature output by the neural network model; p is the p-value obtained in the periodicity identification process; p max The preset maximum value.

[0116] When it is necessary to predict the photovoltaic power generation at a certain time in the future, the weather information of the future time is obtained based on the weather service, and the weather information is sent to the trained neural network model. The output obtained is the photovoltaic characteristics; the determined cycle is queried, and the historical data corresponding to the future time is obtained based on the cycle, that is, the historical photovoltaic characteristics, as similar photovoltaic characteristics; the photovoltaic characteristics obtained by the neural network model are corrected according to the similar photovoltaic characteristics, and the photovoltaic power generation can be determined.

[0117] The purpose of the above correction process is to adjust the weights of X(i, j) and Y(i, j) in the superposition process. It should be noted that the smaller p is, the more obvious the periodicity is. Correspondingly, the weight of X(i, j) is larger. Therefore, there is actually an inversion between the weights of the two. This is a simplified solution. Under normal circumstances, it can be Composite a decreasing function on it as the weight of X(i, j).

[0118] Figure 2 : is a structural block diagram of a distributed photovoltaic power generation prediction device. In an embodiment of the present invention, a distributed photovoltaic power generation prediction device, the device 10 includes:

[0119] The system diagram creation module 11 is used to obtain the installation coordinates of the photovoltaic power generation unit and create a power generation system diagram according to the installation coordinates; the power generation system diagram is a map composed of photovoltaic power generation units, including distribution information of the photovoltaic power generation units;

[0120] A power generation statistics module 12 is used to count the photovoltaic power generation of each day based on the power generation system diagram as a photovoltaic feature;

[0121] A cycle identification module 13 is used to arrange the photovoltaic characteristics according to the time sequence, self-identify the arranged photovoltaic characteristics, and determine the change cycle;

[0122] The model training module 14 is used to record the weather information of each day, take the weather information as input, take the photovoltaic characteristics as output, and train the neural network model;

[0123] The power generation prediction module 15 is used to predict the photovoltaic characteristics in the future based on the change cycle and the trained neural network model to determine the photovoltaic power generation.

[0124] Furthermore, the system diagram creation module 11 includes:

[0125] A coordinate acquisition unit, used to acquire the center position and boundary coordinates of the photovoltaic power generation unit during installation;

[0126] A distance calculation unit, used to obtain a preset system diagram size and calculate a maximum distance between photovoltaic power generation units according to the center position of the photovoltaic power generation units;

[0127] A conversion unit, configured to determine a scale according to the maximum distance and a preset size, convert the center position and the boundary coordinates into a graphic outline based on the scale, insert the contour into the system diagram, and obtain a power generation system diagram;

[0128] Among them, the method of obtaining the center position and boundary coordinates of the photovoltaic power generation unit during installation includes reading the registered data and identifying the remote sensing image.

[0129] Specifically, the power generation statistics module 12 includes:

[0130] A profile query unit, used to receive the instantaneous power generation at each time of the day acquired by the detection module in the photovoltaic power generation unit, and query the corresponding profile in the power generation system diagram;

[0131] A numerical value insertion unit is used to determine a numerical value according to the instantaneous power generation, and insert the numerical value into a point in the corresponding contour; wherein the point selection rule when inserting the numerical value is that when the total value of the numerical value at a certain point reaches a preset threshold, the next point is selected according to a preset order, and the initial point is a pre-selected point;

[0132] The insertion cutoff unit is used to output the power generation system diagram when the total value of all points in any contour reaches the threshold value, read the blank power generation system diagram, and execute the insertion process cyclically until all instantaneous power generation in a day is filled;

[0133] The system diagram statistics unit is used to count all power generation system diagrams output within a day as photovoltaic characteristics.

[0134] In addition, the cycle identification module 13 includes:

[0135] an arrangement unit for arranging the photovoltaic characteristics of different dates according to the time sequence;

[0136] An extraction unit, used for extracting characteristic values ​​of photovoltaic characteristics of each day to obtain a characteristic value sequence;

[0137] An identification unit, used for periodically identifying the characteristic value sequence, and determining the change period and p-value;

[0138] The process of extracting the characteristic value of the photovoltaic characteristics of each day at least includes:

[0139] Converting photovoltaic characteristics into a histogram, and inputting the histogram into a preset conversion model to obtain a characteristic value;

[0140] The periodicity identification process is completed by the scipy.stats submodule in the Python scientific computing library SciPy. The p-value is used to indicate whether the periodicity is strict. The smaller the p-value, the greater the possibility that the data has strict periodicity.

[0141] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A distributed photovoltaic power generation prediction method, characterized in that: The method comprises: Acquire the installation coordinates of the photovoltaic power generation unit, and create a power generation system diagram according to the installation coordinates; the power generation system diagram is a map composed of photovoltaic power generation units, including distribution information of the photovoltaic power generation units; Based on the power generation system diagram, the photovoltaic power generation of each day is counted as the photovoltaic feature; Arrange the photovoltaic characteristics according to the time sequence, self-identify the arranged photovoltaic characteristics, and determine the change cycle; Record weather information for each day, use weather information as input and photovoltaic characteristics as output to train the neural network model; Based on the change cycle and the trained neural network model, the photovoltaic characteristics in the future are predicted to determine the photovoltaic power generation.

2. The distributed photovoltaic power generation prediction method according to claim 1, characterized in that: The step of obtaining the installation coordinates of the photovoltaic power generation unit and creating a power generation system diagram according to the installation coordinates includes: Obtain the center position and boundary coordinates of the photovoltaic power generation unit during installation; Obtain the preset system diagram size, and calculate the maximum distance between photovoltaic power generation units according to the center position of the photovoltaic power generation units; Determine a scale according to the maximum distance and the preset size, convert the center position and the boundary coordinates into a graphic outline based on the scale, insert it into the system diagram, and obtain a power generation system diagram; Among them, the method of obtaining the center position and boundary coordinates of the photovoltaic power generation unit during installation includes reading the registered data and identifying the remote sensing image.

3. The distributed photovoltaic power generation prediction method according to claim 1, characterized in that: The step of counting the photovoltaic power generation of each day based on the power generation system diagram as the photovoltaic feature includes: Receive the instantaneous power generation at each time of the day acquired by the detection module in the photovoltaic power generation unit, and query the corresponding outline in the power generation system diagram; Determine a value according to the instantaneous power generation, and insert the value into the point in the corresponding contour; wherein the point selection rule when inserting the value is that when the total value of the value at a certain point reaches a preset threshold, select the next point in a preset order, and the initial point is the pre-selected point; When the total value of all points in any contour reaches the threshold, the power generation system diagram is output, and the blank power generation system diagram is read, and the insertion process is executed repeatedly until all instantaneous power generation in a day is filled; All power generation system diagrams with output within one day are counted as photovoltaic characteristics.

4. The distributed photovoltaic power generation prediction method according to claim 1, characterized in that: The steps of arranging the photovoltaic characteristics according to the time sequence, self-identifying the arranged photovoltaic characteristics, and determining the change period include: Arrange the PV characteristics of different dates in chronological order; Extract the characteristic values ​​of photovoltaic characteristics of each day to obtain a characteristic value sequence; Perform periodic identification on the characteristic value sequence to determine the change period and p-value; The process of extracting the characteristic value of the photovoltaic characteristics of each day at least includes: Converting photovoltaic characteristics into a histogram, and inputting the histogram into a preset conversion model to obtain a characteristic value; The periodicity identification process is completed by the scipy.stats submodule in the Python scientific computing library SciPy. The p-value is used to indicate whether the periodicity is strict. The smaller the p-value, the greater the possibility that the data has strict periodicity.

5. The distributed photovoltaic power generation prediction method according to claim 1, characterized in that: The steps of recording weather information of each day, taking weather information as input, taking photovoltaic characteristics as output, and training the neural network model include: Record weather information for each day; the weather information includes at least solar radiation, sunshine hours, temperature and cloud cover; Take weather information as the independent variable, read the photovoltaic characteristics of the corresponding date as the dependent variable, and construct the training set, test set and validation set; The neural network model is trained based on the training set, and the error rate of the neural network model is determined based on the test set and the validation set. When the error rate is less than the preset error threshold, the trained neural network model is output.

6. The distributed photovoltaic power generation prediction method according to claim 4, characterized in that: The step of predicting photovoltaic characteristics in the future based on the variation cycle and the trained neural network model and determining photovoltaic power generation includes: Get weather information for future time based on weather service; Sending the weather information to a trained neural network model to output photovoltaic characteristics at a future time; Selecting photovoltaic features corresponding to the future time from the historical photovoltaic features based on the period as similar photovoltaic features; The photovoltaic characteristics output by the neural network model are modified according to similar photovoltaic characteristics; determining photovoltaic power generation according to the photovoltaic characteristics; The correction process is: Where Z(i,j) is the value at the midpoint (i,j) of the corrected photovoltaic feature, X(i,j) is the value at the midpoint (i,j) of the similar photovoltaic feature, and Y(i,j) is the value at the midpoint (i,j) of the photovoltaic feature output by the neural network model; p is the p-value obtained in the periodicity identification process; p max The preset maximum value.

7. A distributed photovoltaic power generation prediction device, characterized in that: The device comprises: A system diagram creation module, used to obtain the installation coordinates of the photovoltaic power generation unit, and create a power generation system diagram according to the installation coordinates; the power generation system diagram is a map composed of photovoltaic power generation units, including distribution information of the photovoltaic power generation units; The power generation statistics module is used to count the photovoltaic power generation of each day based on the power generation system diagram as a photovoltaic feature; A cycle identification module is used to arrange the photovoltaic characteristics according to the time sequence, self-identify the arranged photovoltaic characteristics, and determine the change cycle; The model training module is used to record the weather information of each day, take the weather information as input, and take the photovoltaic characteristics as output to train the neural network model; The power generation prediction module is used to predict the photovoltaic characteristics in the future based on the change cycle and the trained neural network model to determine the photovoltaic power generation.

8. The distributed photovoltaic power generation prediction device according to claim 7, characterized in that: The system diagram creation module includes: A coordinate acquisition unit, used to acquire the center position and boundary coordinates of the photovoltaic power generation unit during installation; A distance calculation unit, used to obtain a preset system diagram size and calculate a maximum distance between photovoltaic power generation units according to the center position of the photovoltaic power generation units; A conversion unit, configured to determine a scale according to the maximum distance and a preset size, convert the center position and the boundary coordinates into a graphic outline based on the scale, insert the contour into the system diagram, and obtain a power generation system diagram; Among them, the method of obtaining the center position and boundary coordinates of the photovoltaic power generation unit during installation includes reading the registered data and identifying the remote sensing image.

9. The distributed photovoltaic power generation prediction device according to claim 7, characterized in that: The power generation statistics module includes: A profile query unit, used to receive the instantaneous power generation at each time of the day acquired by the detection module in the photovoltaic power generation unit, and query the corresponding profile in the power generation system diagram; A numerical value insertion unit is used to determine a numerical value according to the instantaneous power generation, and insert the numerical value into a point in the corresponding contour; wherein the point selection rule when inserting the numerical value is that when the total value of the numerical value at a certain point reaches a preset threshold, the next point is selected according to a preset order, and the initial point is a pre-selected point; The insertion cutoff unit is used to output the power generation system diagram when the total value of all points in any contour reaches the threshold value, read the blank power generation system diagram, and execute the insertion process cyclically until all instantaneous power generation in a day is filled; The system diagram statistics unit is used to count all power generation system diagrams output within a day as photovoltaic characteristics.

10. The distributed photovoltaic power generation prediction device according to claim 7, characterized in that: The cycle identification module comprises: an arrangement unit for arranging the photovoltaic characteristics of different dates according to the time sequence; An extraction unit, used for extracting characteristic values ​​of photovoltaic characteristics of each day to obtain a characteristic value sequence; An identification unit, used for periodically identifying the characteristic value sequence, and determining the change period and p-value; The process of extracting the characteristic value of the photovoltaic characteristics of each day at least includes: Converting photovoltaic characteristics into a histogram, and inputting the histogram into a preset conversion model to obtain a characteristic value; The periodicity identification process is completed by the scipy.stats submodule in the Python scientific computing library SciPy. The p-value is used to indicate whether the periodicity is strict. The smaller the p-value, the greater the possibility that the data has strict periodicity.

Images