CN117216526B - Photovoltaic output prediction method and system based on artificial intelligence - Google Patents
Photovoltaic output prediction method and system based on artificial intelligence Download PDFInfo
- Publication number
- CN117216526B CN117216526B CN202311481673.9A CN202311481673A CN117216526B CN 117216526 B CN117216526 B CN 117216526B CN 202311481673 A CN202311481673 A CN 202311481673A CN 117216526 B CN117216526 B CN 117216526B
- Authority
- CN
- China
- Prior art keywords
- photovoltaic
- cloud layer
- photovoltaic output
- moment
- layer block
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 17
- 230000005540 biological transmission Effects 0.000 claims abstract description 82
- 239000011159 matrix material Substances 0.000 claims abstract description 76
- 238000010248 power generation Methods 0.000 claims abstract description 42
- 238000012937 correction Methods 0.000 claims abstract description 26
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 5
- 238000005286 illumination Methods 0.000 claims description 17
- 238000004590 computer program Methods 0.000 claims description 6
- 230000011218 segmentation Effects 0.000 claims description 5
- 238000012549 training Methods 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 4
- 230000005855 radiation Effects 0.000 description 8
- 230000008569 process Effects 0.000 description 7
- 230000002159 abnormal effect Effects 0.000 description 4
- 230000009467 reduction Effects 0.000 description 4
- 238000002834 transmittance Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 239000013598 vector Substances 0.000 description 3
- 238000005070 sampling Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 238000010521 absorption reaction Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000029087 digestion Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Landscapes
- Photovoltaic Devices (AREA)
Abstract
The invention relates to the technical field of photovoltaic output prediction, in particular to a photovoltaic output prediction method and a photovoltaic output prediction system based on artificial intelligence, wherein the method comprises the following steps: collecting relevant data of a photovoltaic power generation area; acquiring a photovoltaic panel space distribution array of a distributed power generation area; building the photovoltaic output coefficients of all positions of the array, performing singular value decomposition on the photovoltaic panel space distribution array to obtain a photovoltaic panel space feature matrix, and further building a photovoltaic output space stability matrix at all times; calculating the transmission weight of each position of the cloud layer block; acquiring a weighted transmission index and a corrected weighted transmission index of the cloud layer block; calculating cloud layer correction weighted transmission indexes at all times; obtaining a corrected photovoltaic output space stability matrix at each moment; acquiring a photovoltaic output time-space correlation matrix at the current moment; and predicting the photovoltaic output by combining an LSTM time sequence prediction model. Thus realizing high-precision prediction of photovoltaic output.
Description
Technical Field
The invention relates to the technical field of photovoltaic output prediction, in particular to a photovoltaic output prediction method and system based on artificial intelligence.
Background
Due to the characteristics of the distributed photovoltaic, such as inabservable, uncontrollable and the like, the problems of overrun, blind tuning and the like are caused, so that the peak shaving difficulty of the power grid is further increased. In order to support the reasonable consumption of distributed photovoltaics and the safe power distribution requirements, reasonable prediction of photovoltaic output conditions is required.
The traditional photovoltaic output prediction generally adopts historical power generation data as a basis to construct a time sequence prediction model to predict the photovoltaic output condition of a period of time in the future, and the traditional time sequence prediction model only considers time sequence information and cannot reflect the difference of the distributed photovoltaic output equipment in spatial distribution. Meanwhile, the photovoltaic power generation is characterized in that the photovoltaic power generation is influenced by factors such as solar radiation intensity, weather conditions and seasons, and the photovoltaic output has larger random fluctuation, so that challenges are brought to the accuracy of prediction work.
Disclosure of Invention
In order to solve the technical problems, the invention aims to provide an artificial intelligence-based photovoltaic output prediction method and system, and the adopted technical scheme is as follows:
in a first aspect, an embodiment of the present invention provides an artificial intelligence-based photovoltaic output prediction method, including the steps of:
collecting historical time sequence photovoltaic power generation capacity of a photovoltaic power generation area to form a photovoltaic power generation sequence, and collecting environment illumination intensity, environment temperature and remote sensing images of the photovoltaic power generation area;
acquiring a photovoltaic panel space distribution array of a distributed power generation area; obtaining the photovoltaic output coefficients of all the positions according to the number of the photovoltaic panels at all the positions of the photovoltaic panel spatial distribution array, the ambient temperature and the illumination intensity, and taking the photovoltaic output coefficients as the element values of all the positions of the photovoltaic panel spatial distribution array; singular value decomposition is carried out on the photovoltaic panel space distribution array to obtain a photovoltaic panel space feature matrix; constructing each element of a photovoltaic output space stability matrix at each moment according to the difference between each element and the neighborhood element in the photovoltaic panel space characteristic matrix at each moment;
for the remote sensing image at each moment, semantic segmentation is carried out to obtain cloud layer blocks in the remote sensing image, and the transmission weights of all positions of the cloud layer blocks are obtained according to the distances between all positions of the cloud layer blocks and the nearest boundary points of the cloud layer blocks; acquiring a weighted transmission index of the cloud layer block according to the transmission weights of all positions in the cloud layer block and the distances; obtaining a corrected weighted transmission index of the cloud layer block according to the weighted transmission index of the cloud layer block and the gray value of each pixel point of the cloud layer block in the remote sensing image; obtaining cloud layer correction weighted transmission indexes at all times according to the areas of all cloud layer blocks in the remote sensing images at all times and the correction weighted transmission indexes; obtaining a corrected photovoltaic output space stability matrix at each moment according to the photovoltaic output space stability matrix at each moment and the cloud layer correction weighted transmission index; taking the absolute value of the difference value of the corresponding position element of the corrected photovoltaic output space stability matrix at the current moment and the next moment as each element of the photovoltaic output space-time correlation matrix at the current moment;
and taking the photovoltaic output sequence and the photovoltaic output space-time correlation matrix as training samples of an LSTM time sequence prediction model, and outputting a photovoltaic output predicted value by the trained LSTM time sequence prediction model.
Further, the obtaining a photovoltaic panel spatial distribution array of the distributed power generation area includes:
and taking the distributed power generation area as a rectangular area, wherein the rectangular area covers all the photovoltaic panels, and dividing the rectangular area into square small areas serving as positions of the photovoltaic panel space distribution array.
Further, the obtaining the photovoltaic output coefficients of each position according to the number of the photovoltaic panels, the ambient temperature and the illumination intensity at each position of the photovoltaic panel spatial distribution array comprises:
counting the number of photovoltaic panels, the ambient temperature and the illumination intensity at each position of a photovoltaic space distribution array, and taking the product of the number of the photovoltaic panels, the inverse of the ambient temperature and the illumination intensity as the photovoltaic output coefficient at each position;
the photovoltaic output coefficient of the position with zero photovoltaic panel arrangement quantity in the photovoltaic panel space distribution array is recorded as 0.
Further, the construction of each element of the photovoltaic output spatial stability matrix at each moment according to the difference between each element and the neighbor element in the photovoltaic panel spatial characteristic matrix at each moment includes:
for the photovoltaic panel space feature matrix at each moment, eight neighborhood elements of each element in the photovoltaic panel space feature matrix are obtained, the average value of the eight neighborhood elements of each element is obtained, the absolute value of the difference between each element in the photovoltaic panel space feature matrix and the average value is calculated, and the reciprocal of the absolute value of the difference is used as each element of the photovoltaic output space stability matrix at each moment.
Further, the obtaining the transmission weight of each position of the cloud layer block according to the distance between each position of the cloud layer block and the nearest boundary point comprises:
for each position of the cloud layer block, obtaining a boundary point nearest to each position, and taking the distance between each position and the corresponding boundary point as the boundary distance of each position; large reference neighborhood of M1 x M1 and M2 x M2 built centered on each positionA small reference neighborhood, wherein M, M is a preset large and small reference neighborhood size, M1>M2, setting a sliding window with the same size as the small reference neighborhood in the large reference neighborhood, sliding in the large reference neighborhood, and counting sliding times of the sliding window as follows；
Calculating the average value of the boundary distances of all the positions in the small reference neighborhood of each position, marking the average value as a first average value, and calculating the average value of the boundary distances of all the positions of the sliding window in the large reference neighborhood of each position, marking the average value as a second average value; and calculating a normalized value of the absolute value of the difference between the first average value and the second average value, and taking the average value of all the normalized values as the transmission weight of each position.
Further, the obtaining the weighted transmission index of the cloud layer block includes:
and obtaining the ratio of the transmission weight of each position of the cloud layer block to the boundary distance, and taking the average value of the ratio of all positions in the cloud layer block as the weighted transmission index of the cloud layer block.
Further, the obtaining the corrected weighted transmission index of the cloud layer block according to the weighted transmission index of the cloud layer block and the gray value of each pixel point of the cloud layer block in the remote sensing image includes:
and calculating the gray average value of all pixel points of the cloud layer block in the remote sensing image, and taking the product of the average value and the weighted transmission index of the cloud layer block as the corrected weighted transmission index of the cloud layer block.
Further, the obtaining the cloud layer correction weighted transmission index at each moment according to the area of each cloud layer block in the remote sensing image at each moment and the correction weighted transmission index specifically includes:
and counting the area of each cloud layer block in the remote sensing image for the remote sensing image at each moment, obtaining the product of the area of each cloud layer block and the correction weighted transmission index, and taking the average value of the products of all the cloud layer blocks in the remote sensing image as the cloud layer correction weighted transmission index at the corresponding moment.
Further, the obtaining the corrected photovoltaic output spatial stability matrix at each moment according to the photovoltaic output spatial stability matrix at each moment and the cloud layer correction weighted transmission index comprises:
and determining a matrix obtained by multiplying the cloud layer correction weighted transmission index at each moment by the photovoltaic output space stability matrix as the corrected photovoltaic output space stability matrix at each moment.
In a second aspect, an embodiment of the present invention further provides an artificial intelligence-based photovoltaic output prediction system, including a memory, a processor, and a computer program stored in the memory and running on the processor, where the processor implements the steps of any one of the methods described above when the processor executes the computer program.
The invention has at least the following beneficial effects:
the invention provides a photovoltaic output prediction method and a photovoltaic output prediction system based on artificial intelligence, which are used for predicting distributed photovoltaic output equipment by analyzing historical output information of a distributed photovoltaic output system. The method comprises the steps of dividing distributed photovoltaic output equipment in space, obtaining a space association relation, and obtaining a photovoltaic output space-time association matrix sequence by combining a transmission index of cloud layers to solar radiationBy combining the original photovoltaic output sequences +.>The characteristic of the change of the photovoltaic output equipment in time is considered, the spatial distribution characteristic of the photovoltaic output equipment can be obtained from the difference of the distribution positions of the photovoltaic panels, the more accurate prediction of the photovoltaic output equipment is realized, and the photovoltaic output prediction accuracy is improved.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions and advantages of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of steps of an artificial intelligence-based photovoltaic output prediction method according to an embodiment of the present invention.
Detailed Description
In order to further describe the technical means and effects adopted by the invention to achieve the preset aim, the following detailed description is given below of the artificial intelligence-based photovoltaic output prediction method and system according to the invention, which are specific implementation, structure, characteristics and effects thereof, with reference to the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" means that the embodiments are not necessarily the same. Furthermore, the particular features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The following specifically describes a specific scheme of the photovoltaic output prediction method and system based on artificial intelligence provided by the invention with reference to the accompanying drawings.
Referring to fig. 1, a flowchart of steps of an artificial intelligence-based photovoltaic output prediction method according to an embodiment of the present invention is shown, where the method includes the following steps:
and S001, acquiring photovoltaic output history information and related data of a photovoltaic power generation area.
Under the background of the development of the photovoltaic whole county in a pushing scale, the distributed photovoltaic power generation plates are required to be arranged according to the actual conditions of different areas during installation, so that the photovoltaic power generation plates have the characteristics of discreteness, flexibility, randomness and the like. In order to effectively manage and predict the output condition of photovoltaic power generation, the present embodiment predicts in the same manner in other areas by analyzing the photovoltaic output of one distributed power generation area and making relevant power predictions.
Specifically, in the process of predicting the photovoltaic output, reference is first required to the historical photovoltaic power generation output condition of the distributed photovoltaic power generation area.The photovoltaic power generation capacity in each hour is obtained through a real-time monitoring system of a local photovoltaic power generation system operator, and a photovoltaic output sequence is formed by the power generation capacity recorded in each hour and is recorded as. The efficiency of photovoltaic power generation is greatly influenced by illumination intensity, and because distributed photovoltaic power generation facilities need to be irregular in facility distribution due to local conditions, illumination intensity information of a current area is obtained in each small area through arrangement of photoresistors, and local environment temperature is obtained through a temperature sensor. And simultaneously acquiring remote sensing satellite images above the distributed photovoltaic power generation area for analyzing cloud layer absorption characteristics, and adopting the sampling frequency which is the same as that of the photovoltaic output sequence, namely sampling once per hour, so as to obtain the remote sensing image sequence of the photovoltaic output area.
Thus, the method according to the embodiment can obtain the basic data for photovoltaic output analysis.
And step S002, constructing the spatial distribution characteristics of the distributed photovoltaic power generation sequence according to the historical photovoltaic output condition information so as to improve the model prediction capability.
According to the embodiment, for the prediction of the distributed photovoltaic output, the historical photovoltaic output data and the spatial information of photovoltaic facility distribution at the position where the historical photovoltaic output data are located are used as a training set, and the effect of accurately predicting the distributed photovoltaic output is achieved through a training time sequence prediction model. The method comprises the following steps:
1) Photovoltaic output sequenceAnd (5) preprocessing.
The photovoltaic output sequence is taken as important data of photovoltaic output prediction, and the accuracy and the integrity of the data play a very important role in the whole prediction system. In order to eliminate errors such as abnormal values, missing values and the like in the data and provide reliable data for the follow-up photovoltaic output prediction, the sequence is needed to beProceeding withProcessing of deleted values and outliers to obtain a processed sequence +.>The method is characterized by comprising the following steps:
for the missing value possibly caused by equipment faults and other reasons, the average value of 3 data in the front and rear neighborhoods of the missing value is adopted as the filling of the missing value; the abnormal value is processed by adopting a front-back difference method, namely, a difference sequence is constructed by taking difference between two points adjacent in front and back in the sequence, the abnormal value is detected on the new sequence, and the previous observed value is adopted for replacing the abnormal value.
2) Photovoltaic panel spatially distributed array for building photovoltaic power generation equipmentAnd performing dimension reduction treatment.
The distributed photovoltaic power generation facility has the characteristics of flexible construction, complex and diversified scenes, close and convenient digestion and the like. The application places of the photovoltaic power generation system can be a roof, an open parking lot, an industrial area and the like, the scenes are various and non-uniform, and the photovoltaic power generation system can be flexibly expanded and contracted according to actual needs, but is complex in photovoltaic power output prediction. We therefore constructed a spatially distributed array of photovoltaic panelsTo provide its spatial distribution characteristics for photovoltaic output prediction.
By spatially dividing the distributed generation area into one rectangular area, the rectangular area will contain the locations of all photovoltaic panels in the distributed generation area. Then, the rectangular area is uniformly divided into different small areas by small square areas having a side length of 10 meters. Next, a spatially distributed array B of photovoltaic panels is formed by the square small areas, wherein each square small area is an element of the spatially distributed array of photovoltaic panels, and each element value represents a photovoltaic output coefficient of the square small area, specifically:
wherein,representing a spatially distributed array of photovoltaic panels constructed from distributed power generation regions>Middle position->Photovoltaic output coefficient of>Representing a spatially distributed array in a photovoltaic panel>Middle position->The number of photovoltaic panels in the area; />Representing a spatially distributed array in a photovoltaic panel>In (1) position->The ambient temperature of the area in which the device is located; />Representing a spatially distributed array in a photovoltaic panel>In (1) position->The illumination intensity of the region where the light source is located, < >>To avoid the parameter with zero denominator, the practitioner sets itself, in this embodiment, the value is 0.01.
The number and illumination intensity of the photovoltaic panels in each small area are in direct proportion to the photovoltaic output coefficients of the area, namely the larger the number of the photovoltaic panels is, the stronger the illumination intensity is, the higher the corresponding photovoltaic output coefficients are, otherwise, the smaller the number of the photovoltaic panels is, the weaker the illumination intensity is, and the lower the corresponding photovoltaic output coefficients are; the photovoltaic panel can generate corresponding heat in the working process, and if the ambient temperature is too high, the heat radiation capacity of the photovoltaic panel can be affected, and the output power of the photovoltaic panel is further reduced, so that the ambient temperature and the output coefficient of the photovoltaic panel are in inverse proportion, namely, the higher the ambient temperature is, the lower the output coefficient of the photovoltaic panel is, and otherwise, the lower the ambient temperature is, the higher the output coefficient of the photovoltaic panel is.
In view of spatially distributed arrays in photovoltaic panelsIn the method, since a plurality of areas are not provided with the photovoltaic panel, namely the element of the position of the photovoltaic panel is 0, the array is a sparse matrix, so that singular value decomposition SVD is adopted to perform dimension reduction treatment on the matrix to obtain a photovoltaic panel space characteristic matrix->The method specifically comprises the following steps:
first, for a spatially distributed array of raw photovoltaic panelsSingular value decomposition is performed. According to the formula->Three matrices are obtained, wherein ∈ ->Is left singular vector, ">Is the right singular vector, ">Is the singular value of a spatially distributed array of photovoltaic panels.
Then, the process is carried out,the first k principal components are taken. Matrix is formed、/>And->The first k column vectors of (a) constitute a matrix +.>、/>And->Wherein k is generally smaller than the singular value matrix +.>Where k takes a value of 50.
And finally, reconstructing the matrix after dimension reduction. Will approximate a matrix、/>And->Reduction as a new matrix to a photovoltaic panel spatial feature matrix +.>=/>。
3) And calculating a corrected photovoltaic output spatial stability matrix sequence D (t).
In the process of predicting the photovoltaic output, not only the spatial distribution characteristic is considered, but also the dimension of time is added, wherein the time interval adopts the photovoltaic output sequence obtained in the step oneAnd the consistency is one hour. Namely, in the construction of the above step 2) of the spatially distributed array of photovoltaic panels>During the process, the acquired ambient temperature and illumination intensity are acquired once every other hour, so that a photovoltaic panel space distribution array sequence +.>A photovoltaic panel spatial feature matrix sequence +.>。
Since the more uniform the photovoltaic output distribution in the current distributed power generation region, the less the data volatility when photovoltaic output prediction is performed. Thus, next, the stability of the output of the photovoltaic panel in the spatial distribution is considered, i.e. by the firstPhotovoltaic panel spatial feature matrix sequence at individual moments +.>Calculate its photovoltaic output space stability matrix at this moment +.>Thereby obtaining a photovoltaic output space stability matrix sequence +.>The method is characterized by comprising the following steps:
wherein,indicate->Photovoltaic output spatial stability matrix at each moment +.>Middle position->Representing the stability of the photovoltaic output at that location in a spatial distribution; />Indicate->Photovoltaic panel spatial feature matrix sequence at individual moments +.>In (1) position->Element values of (2); m represents->Photovoltaic panel spatial feature matrix sequence at individual moments +.>In (1) position->The number of neighborhood elements of the element value of (2), wherein the neighborhood elements are the elements of the neighborhood of each position 8; />Indicate->Photovoltaic panel spatial feature matrix sequence at individual moments +.>In (1) position->Neighborhood of element values of (2)The u-th element of (a)>To avoid the parameter with zero denominator, the practitioner sets itself, in this embodiment, the value is 0.01.
When the first isPhotovoltaic panel spatial feature matrix sequence at individual moments +.>In (1) position->When the difference between the element of (a) and other element values in the neighborhood is larger, the smaller the photovoltaic output spatial stability of the position is indicated, and otherwise, the larger the photovoltaic output spatial stability of the position is indicated.
Further, in the photovoltaic power generation process, the size and thickness of the cloud layer can directly influence the transmittance of solar radiation, so that the output condition of the photovoltaic panel is influenced. Here, a cloud correction weighted transmission index sequence is constructed by analyzing the remote sensing image at each moment. The method comprises the following steps:
a) And dividing the cloud layer area.
For the remote sensing image, the U-Net network model is adopted for semantic segmentation. The semantic segmentation and the specific neural network model structure are the prior art and are not described in detail herein. The semantic segmentation can accurately extract the precise outline and shape information of the cloud layer, and each cloud layer connected region is marked as a cloud layer block to be processed respectively.
B) Calculating weighted transmission index of pixel block of cloud layer area。
And C) carrying out distance transformation on each cloud layer block divided in the step A) of the remote sensing image at the ith moment, wherein the value of each element of the cloud layer block represents the distance from the nearest boundary point and is recorded as the boundary distance of the element. The further a pixel value is from the boundary, the closer the element value is to the middle of the cloud block.
Then, traversing each element value in the cloud layer block, constructing a large reference neighborhood of M1 x M1 and a small reference neighborhood of M2 x M2 with the current element as the center for each element, wherein the M1 value is 21, the M2 value is 5, then, taking a sliding window which is the same as the small reference neighborhood of M2 x M2 in the large reference neighborhood of M1 x M1, setting the step length to be 5, sliding and traversing in the large reference neighborhood, and finally, calculating the boundary distance similarity of the sliding window and the small reference neighborhood as the transmission weight of each cloud layer blockThe specific calculation is as follows:
wherein,representing the transmission weight of the v element of the j cloud layer block in the remote sensing image at the i moment; />The number of sliding times of the sliding window in the large reference neighborhood of the v element of the j cloud layer block in the remote sensing image at the i moment is represented; />Normalizing the data; />Representing the boundary distance average value of the small reference neighborhood of the v element of the j-th cloud layer block in the remote sensing image at the i-th moment, and marking the boundary distance average value as a first average value for convenience in expression; />The v of the j cloud layer block in the remote sensing image representing the i momentThe average value of the boundary distance of the sliding window in the large reference neighborhood of the element is recorded as a second average value. The v element herein, that is, the v position of the j cloud layer block in the remote sensing image, and this embodiment of the expression method is not specifically limited, so that it is convenient to understand as a key.
When the calculated boundary distance average value of the small reference neighborhood of the pixel is larger than the boundary distance average value of the sliding window around the pixel, the position of the pixel is closer to the center position of the cloud layer block, namely the radiation of sunlight is less likely to penetrate through the cloud layer, namely the transmission weight of the pixel is lower, otherwise, the radiation of the sunlight is more likely to penetrate through the cloud layer, namely the transmission weight of the pixel is higher.
Further, a weighted transmission index of each cloud layer block is calculatedThe method is characterized by comprising the following steps:
wherein,a weighted transmission index of a jth cloud layer block in the remote sensing image at the ith moment is represented; />The number of elements of the jth cloud layer block in the remote sensing image at the ith moment is represented; />Representing the transmission weight of the v element of the j cloud layer block in the remote sensing image at the i moment; />Boundary distance of the v-th element of the j-th cloud layer block in the remote sensing image representing the i-th moment,/v>To avoid parameters with zero denominator, practitionersThe value of the self-setting is 0.01 in the embodiment.
The larger the transmission weight of the current element is, the stronger the solar radiation can pass through the cloud layer, the larger the corresponding weighted transmission index is, and conversely, the smaller the corresponding weighted transmission index is; and when the boundary distance of the current element is larger, the closer the boundary distance is to the middle position of the cloud layer, the smaller the corresponding weighted transmission index should be, and conversely, the larger the corresponding weighted transmission index should be.
C) And correcting the weighted transmission index of each cloud layer block according to the cloud layer distribution thickness.
Since the weighted transmission index of the cloud block calculated in the above step B) can only represent the solar radiation transmittance of the cloud block, and the solar radiation transmittance is also related to the thickness of the cloud distribution, when the cloud distribution is thin, the transmittance of the cloud may be increased, so that the correction of the thickness of the cloud distribution to the weighted transmission index of the element is considered through the gray value of the cloud. The method comprises the following steps:
wherein,the correction weighted transmission index of the jth cloud layer block in the remote sensing image at the ith moment is represented; />A weighted transmission index of a jth cloud layer block in the remote sensing image at the ith moment is represented; />The number of elements of the jth cloud layer block in the remote sensing image at the ith moment is represented; />The gray value of the v element of the j cloud layer block in the remote sensing image at the i moment is represented,to avoid the parameter with zero denominator, the practitioner sets itself, in this embodiment, the value is 0.01.
When the average gray value of the cloud block is larger, the corresponding weighted transmission index should be lower if the thickness of the cloud block is larger, and conversely, the corresponding weighted transmission index should be higher if the thickness of the cloud block is smaller.
In summary, we have thus far obtained the corrected weighted transmission index of each cloud layer block of the remote sensing image at each moment, and then calculated the corrected weighted transmission index of the remote sensing image at each moment, thereby obtaining the cloud layer corrected weighted transmission index sequenceThe method is characterized by comprising the following steps:
wherein,a cloud correcting weighted transmission index at the ith moment; />The number of cloud layer blocks at the ith moment is represented; />The area of the jth cloud layer block at the ith moment is represented; />And the correction weighted transmission index of the jth cloud layer block at the ith moment is represented.
When the cloud layer weighted transmission index of each moment is calculated, the area of each cloud layer block is used as the weight of the correction weighted transmission index, and the final result can be considered to be the overall cloud layer coverage rate of the image.
The spatial stability of the corrected photovoltaic output can be obtained by calculating the cloud layer correction weighted transmission index at each momentMatrix sequenceThe method specifically comprises the following steps:
wherein,indicate->Correcting the photovoltaic output space stability matrix at moment; />Indicate->A moment photovoltaic output space stability matrix; />Indicate->The cloud correction weighted transmission index at the moment.
4) By correcting the sequence of the photovoltaic output space stability matrixCalculating the photovoltaic output time-space correlation matrix sequence +.>。
Further, according to the corrected photovoltaic output space stability matrix sequenceCalculating a photovoltaic output space-time correlation matrix sequence +.>. Mainly embodied in that according to the embodied lightCorrection of the spatial distribution information of the photovoltaic panels for a matrix sequence of spatial stability of the photovoltaic output>At the change characteristics of different moments, constructing a photovoltaic output space-time correlation matrix sequence with space-time correlation characteristics>The method specifically comprises the following steps:
wherein,indicate->Time photovoltaic output time-space correlation matrix sequence at position +.>The value of the element representing +.>The difference value between the position element at the moment and the photovoltaic output stability at the previous moment; />Indicate->Correcting the space stability matrix of the photovoltaic output at the moment at the position +.>Values of the element; />Indicate->Correcting the space stability matrix of the photovoltaic output at the moment at the position +.>The value of the element.
Thus, the photovoltaic output time-space correlation matrix sequence capable of representing the output time-space characteristics of the photovoltaic panel is obtained。
And step S003, according to the photovoltaic output history information and the spatial distribution characteristics of the constructed distributed photovoltaic power generation sequence, a long-short-time memory network LSTM is adopted as a prediction model to predict the photovoltaic output.
Obtaining a pretreated photovoltaic output sequence from the stepsAnd their corresponding large photovoltaic output space-time correlation matrix sequences->。
And then, the two time sequences are used as training samples, a long and short time memory network LSTM is used as a time sequence prediction model, and the photovoltaic output sequence is predicted.
Wherein the input time step of the LSTM time series prediction model is set to 24 and the prediction time step is set to 6. The optimizer uses Adam and the loss function uses Mean Square Error (MSE).
The foregoing is intended to provide a better understanding of the embodiments of the invention to those skilled in the art and is not intended to limit the invention.
Based on the same inventive concept as the above method, the embodiment of the invention further provides an artificial intelligence-based photovoltaic output prediction system, which comprises a memory, a processor and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to realize the steps of any one of the above artificial intelligence-based photovoltaic output prediction methods.
In summary, embodiments of the present invention provide for the generation of light by combining distributed lightThe photovoltaic output equipment is spatially divided, a spatial association relation is obtained, and a photovoltaic output time-space association matrix sequence is obtained by combining the transmission indexes of the cloud layer on solar radiationBy combining the original photovoltaic output sequences +.>The characteristic of the change of the photovoltaic output equipment in time is considered, the spatial distribution characteristic of the photovoltaic output equipment can be obtained from the difference of the distribution positions of the photovoltaic panels, the more accurate prediction of the photovoltaic output equipment is realized, and the photovoltaic output prediction accuracy is improved.
It should be noted that: the sequence of the embodiments of the present invention is only for description, and does not represent the advantages and disadvantages of the embodiments. And the foregoing description has been directed to specific embodiments of this specification. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments.
The foregoing description of the preferred embodiments of the present invention is not intended to be limiting, but rather, any modifications, equivalents, improvements, etc. that fall within the principles of the present invention are intended to be included within the scope of the present invention.
Claims (3)
1. The photovoltaic output prediction method based on artificial intelligence is characterized by comprising the following steps of:
collecting historical time sequence photovoltaic power generation capacity of a photovoltaic power generation area to form a photovoltaic power generation sequence, and collecting environment illumination intensity, environment temperature and remote sensing images of the photovoltaic power generation area;
acquiring a photovoltaic panel space distribution array of a distributed power generation area; obtaining the photovoltaic output coefficients of all the positions according to the number of the photovoltaic panels at all the positions of the photovoltaic panel spatial distribution array, the ambient temperature and the illumination intensity, and taking the photovoltaic output coefficients as the element values of all the positions of the photovoltaic panel spatial distribution array; singular value decomposition is carried out on the photovoltaic panel space distribution array to obtain a photovoltaic panel space feature matrix; constructing each element of a photovoltaic output space stability matrix at each moment according to the difference between each element and the neighborhood element in the photovoltaic panel space characteristic matrix at each moment;
for the remote sensing image at each moment, semantic segmentation is carried out to obtain cloud layer blocks in the remote sensing image, and the transmission weights of all positions of the cloud layer blocks are obtained according to the distances between all positions of the cloud layer blocks and the nearest boundary points of the cloud layer blocks; acquiring a weighted transmission index of the cloud layer block according to the transmission weights of all positions in the cloud layer block and the distances; obtaining a corrected weighted transmission index of the cloud layer block according to the weighted transmission index of the cloud layer block and the gray value of each pixel point of the cloud layer block in the remote sensing image; obtaining cloud layer correction weighted transmission indexes at all times according to the areas of all cloud layer blocks in the remote sensing images at all times and the correction weighted transmission indexes; obtaining a corrected photovoltaic output space stability matrix at each moment according to the photovoltaic output space stability matrix at each moment and the cloud layer correction weighted transmission index; taking the absolute value of the difference value of the corresponding position element of the corrected photovoltaic output space stability matrix at the current moment and the next moment as each element of the photovoltaic output space-time correlation matrix at the current moment;
taking the photovoltaic output sequence and the photovoltaic output space-time correlation matrix as training samples of an LSTM time sequence prediction model, and outputting a photovoltaic output predicted value by the trained LSTM time sequence prediction model;
the photovoltaic output coefficients of all the positions are obtained according to the number of the photovoltaic panels at all the positions of the photovoltaic panel space distribution array, the ambient temperature and the illumination intensity, and the method comprises the following steps: counting the number of photovoltaic panels, the ambient temperature and the illumination intensity at each position of a photovoltaic space distribution array, and taking the product of the number of the photovoltaic panels, the inverse of the ambient temperature and the illumination intensity as the photovoltaic output coefficient at each position; the photovoltaic output coefficient of the position with zero photovoltaic panel arrangement quantity in the photovoltaic panel space distribution array is recorded as 0;
the construction of each element of the photovoltaic output space stability matrix at each moment according to the difference between each element and the neighborhood element in the photovoltaic panel space characteristic matrix at each moment comprises the following steps: for a photovoltaic panel space feature matrix at each moment, acquiring eight neighborhood elements of each element in the photovoltaic panel space feature matrix, acquiring the average value of the eight neighborhood elements of each element, calculating the absolute value of the difference between each element in the photovoltaic panel space feature matrix and the average value, and taking the reciprocal of the absolute value of the difference as each element of a photovoltaic output space stability matrix at each moment;
the obtaining the transmission weight of each position of the cloud layer block according to the distance between each position of the cloud layer block and the nearest boundary point comprises the following steps: for each position of the cloud layer block, obtaining a boundary point nearest to each position, and taking the distance between each position and the corresponding boundary point as the boundary distance of each position; a large reference neighborhood of M1 and a small reference neighborhood of M2 which are constructed by taking each position as a center, wherein M1 and M2 are respectively the preset large reference neighborhood size and the preset small reference neighborhood size, M1 is larger than M2, a sliding window with the same size as the small reference neighborhood size is set in the large reference neighborhood, the sliding window slides in the large reference neighborhood, and the sliding number of the sliding window is recorded as N1; calculating the average value of the boundary distances of all the positions in the small reference neighborhood of each position, marking the average value as a first average value, and calculating the average value of the boundary distances of all the positions of the sliding window in the large reference neighborhood of each position, marking the average value as a second average value; calculating a normalized value of an absolute value of a difference value between the first average value and the second average value, and taking the average value of all the normalized values as the transmission weight of each position;
the obtaining the weighted transmission index of the cloud layer block comprises the following steps: acquiring the ratio of the transmission weight of each position of the cloud layer block to the boundary distance, and taking the average value of the ratio of all positions in the cloud layer block as the weighted transmission index of the cloud layer block;
the method for obtaining the corrected weighted transmission index of the cloud layer block according to the weighted transmission index of the cloud layer block and the gray value of each pixel point of the cloud layer block in the remote sensing image comprises the following steps: calculating the gray average value of all pixel points of the cloud layer block in the remote sensing image, and taking the product of the average value and the weighted transmission index of the cloud layer block as the corrected weighted transmission index of the cloud layer block;
the cloud layer correction weighted transmission index at each moment is obtained according to the area of each cloud layer block in the remote sensing image at each moment and the correction weighted transmission index, and specifically comprises the following steps: for the remote sensing image at each moment, counting the area of each cloud layer block in the remote sensing image, obtaining the product of the area of each cloud layer block and the correction weighted transmission index, and taking the average value of the products of all the cloud layer blocks in the remote sensing image as the cloud layer correction weighted transmission index at the corresponding moment;
the method for obtaining the corrected photovoltaic output space stability matrix at each moment according to the photovoltaic output space stability matrix at each moment and the cloud layer correction weighted transmission index comprises the following steps: and determining a matrix obtained by multiplying the cloud layer correction weighted transmission index at each moment by the photovoltaic output space stability matrix as the corrected photovoltaic output space stability matrix at each moment.
2. The artificial intelligence based photovoltaic power generation prediction method according to claim 1, wherein the obtaining a spatially distributed array of photovoltaic panels of a distributed power generation area comprises:
and taking the distributed power generation area as a rectangular area, wherein the rectangular area covers all the photovoltaic panels, and dividing the rectangular area into square small areas serving as positions of the photovoltaic panel space distribution array.
3. An artificial intelligence based photovoltaic output prediction system comprising a memory, a processor and a computer program stored in the memory and running on the processor, characterized in that the processor implements the steps of the method according to any one of claims 1-2 when executing the computer program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311481673.9A CN117216526B (en) | 2023-11-09 | 2023-11-09 | Photovoltaic output prediction method and system based on artificial intelligence |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311481673.9A CN117216526B (en) | 2023-11-09 | 2023-11-09 | Photovoltaic output prediction method and system based on artificial intelligence |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117216526A CN117216526A (en) | 2023-12-12 |
| CN117216526B true CN117216526B (en) | 2024-01-26 |
Family
ID=89051516
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311481673.9A Active CN117216526B (en) | 2023-11-09 | 2023-11-09 | Photovoltaic output prediction method and system based on artificial intelligence |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117216526B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117833346B (en) * | 2024-03-04 | 2024-05-14 | 湖南璟泰信息系统有限公司 | Photovoltaic and power grid complementary direct-current energy-saving air conditioner power supply method and system |
| CN117879047B (en) * | 2024-03-13 | 2024-05-24 | 国网山西省电力公司经济技术研究院 | Optimization method and system for distributed photovoltaic access distribution network |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103077300A (en) * | 2012-12-20 | 2013-05-01 | 东南大学 | Forecasting method for generating capacity of distributed photovoltaic power supply on basis of type-2 fuzzy logic |
| CN109510198A (en) * | 2018-12-14 | 2019-03-22 | 国网山东省电力公司经济技术研究院 | A kind of photovoltaic power generation receiving capability assessment method based on Radiation Characteristics meteorology scene |
| KR20220064760A (en) * | 2020-11-12 | 2022-05-19 | 김승섭 | Solar power system capable of predicting the amount of power generation |
| CN115239002A (en) * | 2022-07-29 | 2022-10-25 | 深圳市慧科达科技有限公司 | New energy distributed photovoltaic power station monitoring and early warning system and method |
| CN115811044A (en) * | 2022-11-18 | 2023-03-17 | 国网河北省电力有限公司营销服务中心 | Photovoltaic power abrupt change prediction method |
| CN116435984A (en) * | 2023-02-28 | 2023-07-14 | 云南电网有限责任公司昆明供电局 | Distributed photovoltaic output prediction method based on graph convolution long-short-term memory network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9825580B2 (en) * | 2014-09-26 | 2017-11-21 | State Grid Corporation Of China | Method for constructing real-time solar irradiation metering network of gigawatts level photovoltaic power generation base |
-
2023
- 2023-11-09 CN CN202311481673.9A patent/CN117216526B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103077300A (en) * | 2012-12-20 | 2013-05-01 | 东南大学 | Forecasting method for generating capacity of distributed photovoltaic power supply on basis of type-2 fuzzy logic |
| CN109510198A (en) * | 2018-12-14 | 2019-03-22 | 国网山东省电力公司经济技术研究院 | A kind of photovoltaic power generation receiving capability assessment method based on Radiation Characteristics meteorology scene |
| KR20220064760A (en) * | 2020-11-12 | 2022-05-19 | 김승섭 | Solar power system capable of predicting the amount of power generation |
| CN115239002A (en) * | 2022-07-29 | 2022-10-25 | 深圳市慧科达科技有限公司 | New energy distributed photovoltaic power station monitoring and early warning system and method |
| CN115811044A (en) * | 2022-11-18 | 2023-03-17 | 国网河北省电力有限公司营销服务中心 | Photovoltaic power abrupt change prediction method |
| CN116435984A (en) * | 2023-02-28 | 2023-07-14 | 云南电网有限责任公司昆明供电局 | Distributed photovoltaic output prediction method based on graph convolution long-short-term memory network |
Non-Patent Citations (4)
| Title |
|---|
| Interval prediction of ophotovoltaic power generation based on cloud theory;Haibo Zhao等;《IOP Conference Series:Earth and Environmental Science》;第446卷;第1-7页 * |
| Prediction of Short-Term Photovoltaic Power Via Self-Attention-Based Deep Learning Approach;Jie Li等;《Journal of Energy Reaources Technology》;第144卷(第10期);第1-13页 * |
| 微电网光伏发电的Adaboost天气聚类超短期预测方法;谭津等;《电力系统自动化》;41(21);第33-39页 * |
| 考虑光伏出力时空关联特征的电力系统跨区优化调度研究;李欣彤;《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》(第09期);第C042-277页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117216526A (en) | 2023-12-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN117216526B (en) | Photovoltaic output prediction method and system based on artificial intelligence | |
| JP6968378B2 (en) | Power prediction method and system of photovoltaic power plant based on operation data of grid interconnection inverter | |
| CN110909919A (en) | Photovoltaic power prediction method of depth neural network model with attention mechanism fused | |
| CN103489038A (en) | Photovoltaic ultra-short-term power prediction method based on LM-BP neural network | |
| CN111178616A (en) | Wind speed prediction method based on negative correlation learning and regularization extreme learning machine integration | |
| CN112561058A (en) | Short-term photovoltaic power prediction method based on Stacking-ensemble learning | |
| CN105160423A (en) | Photovoltaic power generation prediction method based on Markov residual error correction | |
| Golam et al. | A long short-term memory-based solar irradiance prediction scheme using meteorological data | |
| CN112100911A (en) | Solar radiation prediction method based on deep BISLTM | |
| CN114462718A (en) | CNN-GRU wind power prediction method based on time sliding window | |
| CN116128170B (en) | Photovoltaic power station power ultra-short-term prediction method and device and related equipment | |
| CN116128141B (en) | Storm surge prediction method and device, storage medium and electronic equipment | |
| CN114297907A (en) | Greenhouse environment spatial distribution prediction method and device | |
| CN112862274A (en) | WRF-LES and Deepar combined wind power prediction method | |
| CN116014722A (en) | Sub-solar photovoltaic power generation prediction method and system based on seasonal decomposition and convolution network | |
| CN117293809A (en) | Multi-time space scale new energy generation power prediction method based on large model | |
| Asanza et al. | Fpga based meteorological monitoring station | |
| Lin et al. | GAOformer: An adaptive spatiotemporal feature fusion transformer utilizing GAT and optimizable graph matrixes for offshore wind speed prediction | |
| CN116108761B (en) | Regional climate simulation method and system for coupling deep learning and HASM | |
| CN109447843B (en) | Photovoltaic output prediction method and device | |
| CN116525135A (en) | Method for predicting epidemic situation development situation by space-time model based on meteorological factors | |
| CN116611023A (en) | Lightning automatic prediction method based on predicted high-abstraction feature network | |
| CN115204467A (en) | Power load prediction method, device and storage medium | |
| CN114841461A (en) | Air quality integration prediction method based on time sequence missing perception and multi-source factor fusion | |
| CN112288694B (en) | Method for identifying defects of power transformation equipment based on mask region convolution neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |