Distributed photovoltaic short-term power prediction method and system

Abstract

The invention discloses a distributed photovoltaic short-term power prediction method and system, and the method comprises the steps: S101, obtaining historical generated power data of each photovoltaic power station and meteorological data corresponding to the historical generated power data of each photovoltaic power station, carrying out the data preprocessing, and constructing a sample library; s102, segmenting the sample data according to the data length L and the data segmentation step length S to realize sample library data expansion; s103, acquiring an actual power sequence of the photovoltaic power station and a corresponding actual meteorological data sequence, wherein the lengths of the actual power sequence and the actual meteorological data sequence are L-1; s104, acquiring N similar samples of the photovoltaic power station from a sample library by using the actual power sequence and the corresponding actual meteorological data sequence; and S105, calculating the power of the photovoltaic power station based on the N similar samples, and summarizing the predicted power of each photovoltaic power station to obtain the predicted power of the distributed photovoltaic power station. And through fine segmentation of historical data, the precision of the prediction model is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of photovoltaic power prediction, and particularly relates to a distributed photovoltaic short-term power prediction method and system. BACKGROUND

[0002] The photovoltaic power output has obvious randomness, volatility and intermittency. With large-scale integration of distributed photovoltaic power stations into the power grid, great challenges are brought to power grid dispatching management and power safety. Photovoltaic power prediction is one of the key technologies to solve this problem, which can promote the integration of photovoltaic power into day-ahead dispatching plan, help new energy consumption, and improve the lean management level of the power grid, and has important significance for improving the quality of photovoltaic power grid integration.

[0003] Factors affecting photovoltaic power include solar radiation value, temperature, humidity, wind speed and direction, photovoltaic components, etc., resulting in high input dimension of the prediction model, complex weather prediction results and large deviation. In order to improve the accuracy of the prediction model, Chinese patent CN117767281A divides historical data into three time periods according to the solar radiation intensity, and constructs three training sets accordingly; Chinese patent CN116484998A divides historical data into eight categories according to the two time periods of morning and afternoon and four weather types. The above two Chinese patents both divide the historical data in detail, and train a prediction model for each subdivided historical data, so as to make the prediction model more accurate. However, the illumination amplitude value and meteorological factors in each period of the same day are volatile, and further fine division of historical data is needed to obtain more accurate prediction results. SUMMARY

[0004] In order to overcome the shortcomings of the prior art, the present application provides a distributed photovoltaic short-term power prediction method and system, which further finely divides the historical power generation data and the corresponding meteorological data, finds similar samples in the historical data according to the actual power generation data and meteorological data of the prediction day, and improves the prediction accuracy of the distributed photovoltaic power based on the similar samples.

[0005] In a first aspect, the present application discloses a distributed photovoltaic short-term power prediction method, comprising the following steps:

[0006] Step S101: Obtain the historical power generation data of each photovoltaic power station, and the historical power generation data corresponding to the meteorological data of each photovoltaic power station, perform data preprocessing, and construct a sample library.

[0007] Step S102: Divide the sample data according to the data length L and the data division step S, realize sample library data expansion, wherein S is less than L, S and L are positive integers, the first L-1 bits of each sample are data characteristics, and the Lth bit is a data label.

[0008] Step S103: Obtain an actual power sequence of the photovoltaic power station and a corresponding actual meteorological data sequence, and the lengths of the actual power sequence and the actual meteorological data sequence are both L-1.

[0009] Step S104: Obtain N similar samples of the photovoltaic power station in the sample library by using the actual power sequence and the corresponding actual meteorological data sequence.

[0010] Step S105: Calculate the power of the photovoltaic power station based on the N similar samples, and aggregate the predicted powers of the photovoltaic power stations to obtain the predicted power of the distributed photovoltaic power station.

[0011] Preferably, the historical power generation data of each photovoltaic power station and the corresponding meteorological data are classified and stored according to the weather types of the meteorological data, a plurality of sample sub-libraries are established, and the weather types include sunny days, rainy and snowy days, cloudy days, and overcast days.

[0012] Preferably, each sample in each sample sub-library is traversed, and for each sample, a new sample is obtained by moving on the sample data by a set data length L and a data division step S.

[0013] Preferably, each sample in each sample sub-library is stored in a data partition according to the illumination amplitude value of the sample.

[0014] Preferably, the data is stored in a data partition according to the start and end time of the sample, or the data is stored in a data partition according to the illumination amplitude value interval of the sample.

[0015] Preferably, the N similar samples of the photovoltaic power station in the sample library are obtained by using the actual power sequence and the corresponding actual meteorological data sequence, and the specific process is as follows:

[0016] Step S1041: Determine the weather type according to the actual meteorological data sequence, obtain the probability of each weather type, sort the probabilities in descending order, and select at least one weather type.

[0017] Step S1042: Obtain the corresponding sample sub-library according to the weather type, obtain the illumination amplitude value and / or the time interval according to the actual meteorological data sequence, and determine the corresponding data partition.

[0018] Step S1043: Perform meteorological similarity calculation in the determined data partition according to the actual meteorological data sequence, and select N1 samples with the highest meteorological similarity.

[0019] Step S1044: Calculate the power similarity between the actual power sequence and the historical power generation data of the N1 samples, respectively, and select N samples with the highest power similarity.

[0020] Preferably, the power of the photovoltaic power station is calculated based on the N similar samples, and the specific process is as follows:

[0021] The power of the photovoltaic power station is calculated using the data labels of the N samples and the power similarity.

[0022] Preferably, the prediction model is trained using the N sample data, the actual meteorological data sequence is input into the prediction model, and the power of the photovoltaic power station is obtained; the prediction models of the respective photovoltaic power stations are the same or different.

[0023] Preferably, if the number of samples N is less than the sample threshold, the power of the photovoltaic power station is calculated using the data labels of the N samples and the power similarity.

[0024] If the number of samples N is not less than the sample threshold, the prediction model is trained using the N sample data, the actual meteorological data sequence is input into the prediction model, and the power of the photovoltaic power station is obtained.

[0025] In a second aspect, the present application discloses a distributed photovoltaic short-term power prediction system, comprising a sample library construction module, a sample expansion module, a data acquisition module, a sample screening module, and a power prediction module.

[0026] The sample library construction module: acquires historical power generation data of each photovoltaic power station and meteorological data corresponding to the historical power generation data of each photovoltaic power station, performs data preprocessing, and constructs a sample library.

[0027] The sample expansion module: divides the sample data according to the data length L and the data division step S to expand the sample library data, wherein S is less than L, S and L are both positive integers, the first L-1 bits of each sample are data features, and the Lth bit is a data label.

[0028] The data acquisition module: acquires an actual power sequence of a photovoltaic power station and a corresponding actual meteorological data sequence, and the lengths of the actual power sequence and the actual meteorological data sequence are both L-1.

[0029] The sample screening module: uses the actual power sequence and the corresponding actual meteorological data sequence to obtain N similar samples of the photovoltaic power station in the sample library.

[0030] The power prediction module: calculates the power of the photovoltaic power station based on the N similar samples, and aggregates the predicted powers of the respective photovoltaic power stations to obtain the predicted power of the distributed photovoltaic.

[0031] Compared with the closest prior art, the technical effect of the present application is:

[0032] 1. By further finely dividing the historical power generation data and the corresponding meteorological data, not only is the sample library expanded, but also the similarity of the obtained similar samples is higher, thereby improving the accuracy of the prediction model.

[0033] 2. Historical data is stored in separate databases using weather types, and further partitioned using solar radiation values, which improves the speed of finding similar samples. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings.

[0035] Figure 1 This is a flowchart of a distributed photovoltaic short-term power prediction method according to the present invention;

[0036] Figure 2 This is a flowchart of the present invention for screening similar samples from a sample library;

[0037] Figure 3 This is a schematic diagram of the structure of a distributed photovoltaic short-term power prediction system according to the present invention. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0039] See Figure 1 As shown, this invention discloses a method for short-term power prediction of distributed photovoltaic power, comprising the following steps:

[0040] Step S101: Obtain historical power generation data of each photovoltaic power station and meteorological data corresponding to the historical power generation data of each photovoltaic power station, perform data preprocessing, and construct a sample library.

[0041] Distributed photovoltaic (PV) power stations refer to photovoltaic power generation systems of less than one megawatt installed on rooftops or other locations that do not occupy significant land resources. Each PV power station in this invention is equipped with a small weather station to collect meteorological data corresponding to the PV power station. The small weather station collects information such as wind speed, wind direction, solar radiation, temperature, humidity, and air pressure through various sensors, and sends the collected meteorological information and the PV power station's power data to a cloud server. Preferably, the PV power generation and meteorological data sampling frequency are consistent; alternatively, if the sampling frequencies are inconsistent, the data preprocessing includes time alignment processing to ensure that the PV power generation and meteorological data are time-aligned. Preferably, data is collected every 15 minutes, i.e., 96 points of PV power generation data and meteorological data are collected daily from each PV power station.

[0042] Due to the large number of distributed photovoltaic (PV) power stations and the diverse dimensions of data collection, the sample database is enormous, impacting data query efficiency. To improve query speed, the data preprocessing includes classifying and storing historical power generation data and corresponding meteorological data for each PV power station according to weather type, establishing multiple sample sub-databases. Weather types include sunny, rainy / snowy, cloudy, and overcast. Data queries are then performed in the relevant sub-databases based on actual weather conditions, improving query speed to some extent. Especially when distributed PV power stations are geographically widespread, the weather types at each station may differ; in this case, the query volume is distributed to a certain degree of equilibrium across the different sample sub-databases.

[0043] Step S102: Segment the sample data according to the data length L and the data segmentation step size S to expand the sample database. Here, S is less than L, and both S and L are positive integers. The first L-1 bits of each sample are data features, and the Lth bit is the data label.

[0044] Solar radiation and meteorological factors are volatile, and multiple weather types can even exist within a single day, such as sunny in the morning and rainy or snowy in the afternoon. This places higher demands on data samples and prediction models. This invention iterates through each sample in each sample sub-database. For each sample, it moves the data based on a set data length L and data segmentation step size S, obtaining a new sample every S positions. For example, for a 96-point sample, a data length L of 13 and a data segmentation step size S of 1 can be used. This generates 84 new samples from the original 96-point sample, thus expanding the sample database and significantly increasing the probability that each new sample represents a single weather type.

[0045] For each sample, the first L-1 bits are data features, and the Lth bit is the data label. That is, when training the prediction model, the first L-1 data features are used to fit the Lth data label.

[0046] To further improve data query efficiency, data is partitioned and stored according to the illumination amplitude value of each sample in each sample sub-database. Preferably, data is partitioned and stored according to the start and end times of the samples, or according to the range of illumination amplitude values ​​of the samples. Taking a data length L of 13 as an example, samples from 3:00 to 6:00 form one data partition, and samples from 3:15 to 6:15 form another data partition. For example, 200 to 350 W / m²... 2 The sample of illumination amplitude values ​​is a data partition, ranging from 351 to 600 W / m. 2 The sample of illumination amplitude values ​​constitutes another data partition. Data partitioning allows for the selection of specific partitions for querying based on the query sample, reducing the scope of the query and further improving data retrieval speed.

[0047] Step S103: Obtain the actual power sequence of the photovoltaic power station and the corresponding actual meteorological data sequence. The length of both the actual power sequence and the actual meteorological data sequence is L-1.

[0048] When predicting the power of a photovoltaic power station at time t+1, the actual power sequence (w) of the photovoltaic power station is obtained. t-L+2 ,…,w t-1 , w t Where is the actual power of the photovoltaic power station at time t. Simultaneously, data sequences of various meteorological factors are obtained, such as wind speed, wind direction, solar radiation, temperature, humidity, and air pressure.

[0049] Step S104: Obtain N similar samples of photovoltaic power plants from the sample library using the actual power sequence and the corresponding actual meteorological data sequence. Specifically, the process involves obtaining N similar samples of photovoltaic power plants from the sample library using the actual power sequence and the corresponding actual meteorological data sequence, as follows: Figure 2 As shown:

[0050] Step S1041: Determine the weather type based on the actual meteorological data sequence, obtain the probability of each weather type, sort the probabilities in descending order, and select at least one weather type.

[0051] Weather types are determined by analyzing meteorological factors such as wind speed, wind direction, solar radiation, temperature, humidity, and air pressure from actual meteorological data sequences. Specifically, a classification model can be used for this determination; this invention does not impose any specific limitations. Taking the selection of two weather types as an example, assuming the determination results are: a probability of 0.7 for sunny days, a probability of 0.1 for rain / snow, a probability of 0.4 for cloudy days, and a probability of 0.45 for partly cloudy days, then the two weather types selected are sunny and partly cloudy.

[0052] Step S1042: Obtain the corresponding sample sub-library according to the weather type, obtain the illumination amplitude value and / or time interval according to the actual meteorological data sequence, and determine the corresponding data partition.

[0053] When the determined weather type is sunny or cloudy, a corresponding sample sub-library is determined. The data partitions of the sample sub-library are then determined based on the start and end times of the samples and the illumination amplitude values. Similar samples are then retrieved from these data partitions. Of course, if no more than a set threshold of similar samples is obtained from a determined data partition, similar samples are searched in neighboring data partitions.

[0054] Step S1043: Calculate meteorological similarity in the determined data partition based on the actual meteorological data sequence, and select the N1 samples with the highest meteorological similarity.

[0055] Similarity calculation is a key technique used in various fields to assess the similarity between two data objects. Similarity measures such as cosine similarity and Euclidean distance can be used, and this invention does not impose specific limitations. Both N1 and N are positive integers, and N1 is greater than N.

[0056] Step S1044: Calculate the power similarity between the actual power sequence and the historical power generation data of the N1 samples, and select the N samples with the highest power similarity. To further improve the accuracy of similar samples, the actual power sequence is used to further optimize the N1 similar samples.

[0057] Preferably, the power of the photovoltaic power station is calculated based on N similar samples, specifically as follows:

[0058] The power of a photovoltaic power station is calculated by weighting the data labels and power similarity of N samples. Specifically, the power similarity of each sample can be normalized first, and then the sum of the products of the normalized power similarity and the sample data labels can be obtained as the predicted power value of the photovoltaic power station at time t.

[0059] Alternatively, a prediction model can be trained using N sample data points. The actual meteorological data sequence is then input into the prediction model to obtain the power output of the photovoltaic power station. The prediction models for each photovoltaic power station may be the same or different. The prediction model can employ convolutional neural network models, temporal neural network models, etc., and this invention does not impose any specific limitations.

[0060] Preferably, if the number of samples N is less than the sample threshold, the power of the photovoltaic power station is calculated by weighting the data labels and power similarity of the N samples. If the number of samples N is not less than the sample threshold, the prediction model is trained using the N sample data, and the actual meteorological data sequence is input into the prediction model to obtain the power of the photovoltaic power station.

[0061] Step S105: Calculate the power of the photovoltaic power station based on N similar samples, and summarize the predicted power of each photovoltaic power station to obtain the predicted power of distributed photovoltaic power. By summarizing the predicted power of each photovoltaic power station, the predicted power value of distributed photovoltaic power is obtained.

[0062] Secondly, this invention discloses a distributed photovoltaic short-term power prediction system, such as... Figure 3 As shown, it includes a sample library construction module, a sample expansion module, a data acquisition module, a sample screening module, and a power prediction module;

[0063] Sample library construction module: Obtain historical power generation data of each photovoltaic power station, as well as meteorological data corresponding to the historical power generation data of each photovoltaic power station, perform data preprocessing, and construct a sample library.

[0064] Sample expansion module: The sample data is segmented according to the data length L and the data segmentation step size S to expand the sample library data. S is less than L, and both S and L are positive integers. The first L-1 bits of each sample are data features, and the Lth bit is the data label.

[0065] Data acquisition module: acquires the actual power sequence and the corresponding actual meteorological data sequence of the photovoltaic power station. The length of both the actual power sequence and the actual meteorological data sequence is L-1.

[0066] Sample screening module: Use the actual power sequence and the corresponding actual meteorological data sequence to obtain N similar samples of photovoltaic power plants in the sample library.

[0067] Power prediction module: Calculates the power of photovoltaic power plants based on N similar samples, summarizes the predicted power of each photovoltaic power plant, and obtains the predicted power of distributed photovoltaic.

[0068] It should be noted that, depending on the implementation needs, the various components / steps described in the embodiments of this application can be broken down into more components / steps, or two or more components / steps or parts of the operation of components / steps can be combined into new components / steps to achieve the purpose of the embodiments of this application.

[0069] The methods described in the embodiments of this application can be implemented in hardware, firmware, or as software or computer code that can be stored in a recording medium (such as a CD-ROM, RAM, floppy disk, hard disk, or magneto-optical disk), or as computer code originally stored on a remote recording medium or a non-transitory machine-readable medium and subsequently stored on a local recording medium, downloaded over a network. Thus, the apparatus described herein can be stored on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as an ASIC or FPGA) for such software processing. It is understood that the computer, processor, microprocessor controller, or programmable hardware includes storage components (e.g., RAM, ROM, flash memory, etc.) capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the apparatus described herein. Furthermore, when a general-purpose computer accesses the code used to implement the apparatus shown herein, the execution of the code transforms the general-purpose computer into a dedicated computer for executing the apparatus shown herein.

[0070] Those skilled in the art will recognize that the unit and device steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different devices to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the embodiments of this application.

[0071] The above embodiments are only used to illustrate the embodiments of this application, and are not intended to limit the embodiments of this application. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the embodiments of this application. Therefore, all equivalent technical solutions also fall within the scope of the embodiments of this application, and the patent protection scope of the embodiments of this application should be defined by the claims.

Claims

1. A method for short-term power prediction of distributed photovoltaic power, characterized in that, Includes the following steps: Step S101: Obtain historical power generation data of each photovoltaic power station and meteorological data corresponding to the historical power generation data of each photovoltaic power station, perform data preprocessing, and construct a sample library; Step S102: Segment the sample data according to the data length L and the data segmentation step size S to expand the sample database. Here, S is less than L, and both S and L are positive integers. The first L-1 bits of each sample are data features, and the Lth bit is the data label. Step S103: Obtain the actual power sequence of the photovoltaic power station and the corresponding actual meteorological data sequence. The length of both the actual power sequence and the actual meteorological data sequence is L-1. Step S104: Use the actual power sequence and the corresponding actual meteorological data sequence to obtain N similar samples of photovoltaic power plants in the sample library; Step S105: Calculate the power of the photovoltaic power station based on N similar samples, summarize the predicted power of each photovoltaic power station, and obtain the predicted power of distributed photovoltaic.

2. The distributed photovoltaic short-term power prediction method as described in claim 1, characterized in that, Historical power generation data of each photovoltaic power station and corresponding meteorological data were obtained. Data preprocessing was performed to construct a sample library, specifically: Based on the weather type of the meteorological data, the historical power generation data and corresponding meteorological data of each photovoltaic power station are classified and stored, and multiple sample sub-databases are established; the weather types include sunny, rainy / snowy, cloudy, and partly cloudy.

3. The distributed photovoltaic short-term power prediction method as described in claim 2, characterized in that, The sample data is segmented according to the data length L and the data segmentation step size S to expand the sample database, specifically as follows: Iterate through each sample in each sample sub-database. For each sample, move it across the sample data according to the set data length L and data splitting step size S. A new sample is obtained every S positions moved.

4. The distributed photovoltaic short-term power prediction method as described in claim 3, characterized in that, Data is partitioned and stored based on the illumination amplitude value of each sample in each sample sub-library.

5. The distributed photovoltaic short-term power prediction method as described in claim 4, characterized in that, Data can be partitioned and stored based on the start and end times of the samples; or, data can be partitioned and stored based on the range of illumination amplitude values ​​of the samples.

6. A method for short-term power prediction of distributed photovoltaic power as described in claim 4 or 5, characterized in that, N similar samples of photovoltaic power plants are obtained from the sample library using actual power sequences and corresponding actual meteorological data sequences, specifically: Step S1041: Determine the weather type based on the actual meteorological data sequence, obtain the probability of each weather type, sort the probabilities in descending order, and select at least one weather type; Step S1042: Obtain the corresponding sample sub-library according to the weather type, obtain the illumination amplitude value and / or time interval according to the actual meteorological data sequence, and determine the corresponding data partition; Step S1043: Calculate meteorological similarity in the determined data partition based on the actual meteorological data sequence, and select the N1 samples with the highest meteorological similarity. Step S1044: Calculate the power similarity between the actual power sequence and the historical power generation data of N1 samples respectively, and select the N samples with the highest power similarity.

7. The distributed photovoltaic short-term power prediction method according to claim 6, characterized in that, The power of a photovoltaic power station is calculated based on N similar samples, specifically as follows: The power of a photovoltaic power station is calculated by weighting the data labels and power similarity of N samples.

8. The distributed photovoltaic short-term power prediction method as described in claim 6, characterized in that, The prediction model is trained using N sample data. The actual meteorological data sequence is input into the prediction model to obtain the power of the photovoltaic power station. The prediction models of each photovoltaic power station may be the same or different.

9. A method for short-term power prediction of distributed photovoltaic power as described in claim 6, characterized in that, If the number of samples N is less than the sample threshold, the power of the photovoltaic power station is calculated by weighting the data labels and power similarity of the N samples. If the number of samples N is not less than the sample threshold, the prediction model is trained using N sample data. The actual meteorological data sequence is then input into the prediction model to obtain the power of the photovoltaic power station.

10. A distributed photovoltaic short-term power prediction system, characterized in that, It includes a sample library construction module, a sample expansion module, a data acquisition module, a sample screening module, and a power prediction module; Sample library construction module: Obtain historical power generation data of each photovoltaic power station, as well as meteorological data corresponding to the historical power generation data of each photovoltaic power station, perform data preprocessing, and construct a sample library; Sample expansion module: The sample data is segmented according to the data length L and the data segmentation step size S to expand the sample library data. S is less than L, and both S and L are positive integers. The first L-1 positions of each sample are data features, and the Lth position is the data label. Data acquisition module: acquires the actual power sequence of the photovoltaic power station and the corresponding actual meteorological data sequence. The length of both the actual power sequence and the actual meteorological data sequence is L-1. Sample screening module: Uses the actual power sequence and the corresponding actual meteorological data sequence to obtain N similar samples of photovoltaic power plants from the sample library; Power prediction module: Calculates the power of photovoltaic power plants based on N similar samples, summarizes the predicted power of each photovoltaic power plant, and obtains the predicted power of distributed photovoltaic.

Images