Distributed photovoltaic power station power prediction method

Abstract

The invention relates to the technical field of intelligent power grids. The invention provides a distributed photovoltaic power station power prediction method. The method comprises the following steps: acquiring data of all power generation units in a distributed photovoltaic power station area; according to the data of the power generation units, the power generation units are divided step by step, and the power generation units are divided into a plurality of power generation clusters; preprocessing the historical power data to obtain a cluster historical power data set; based on the geographical location interval distribution condition of the power generation cluster, performing grid division on the historical meteorological data to obtain historical grid meteorological data of the power generation cluster; according to the correlation between the historical power data and the historical grid meteorological data, obtaining grid prediction meteorological data of the power generation cluster; and constructing a power prediction model based on historical data, and inputting the grid prediction meteorological data of the power generation cluster into the power prediction model to obtain predicted power. The problems that in existing distributed photovoltaic power station power prediction, data acquisition is difficult, the power generation performance difference is large, and a prediction model is difficult to establish are solved.

Description

Technical Field

[0001] The present invention relates to the field of smart grid technology, and in particular to a distributed photovoltaic power station power prediction method. Background Art

[0002] Distributed photovoltaic projects have the characteristics of dispersed distribution of power generation units, small size, high risk and management difficulty. There are three voltage levels for distributed photovoltaic access to the grid: 10kV, 380V (including 400V), and 220V. The distributed photovoltaic access power generation units above 10kV are relatively concentrated, and the measured data is relatively easy to obtain; the distributed photovoltaic power generation units of 380V (including 400V) / 220V are relatively dispersed, and the measured data are difficult to obtain. The capacity of each power generation unit is small, and the power generation performance varies greatly.

[0003] At present, in actual usage scenarios, due to the small installed capacity of distributed photovoltaics and the difficulty in obtaining measured data, it is difficult to achieve modeling and prediction based on current data-driven prediction algorithms, and distributed photovoltaic power prediction is difficult. Summary of the invention

[0004] The purpose of the present invention is to provide a distributed photovoltaic power station power prediction method, aiming to solve the problems of difficult data acquisition, large differences in power generation performance and difficulty in establishing a prediction model in existing distributed photovoltaic power station power prediction.

[0005] The present invention is achieved through the following technical solutions:

[0006] A distributed photovoltaic power station power prediction method comprises the following steps:

[0007] Obtain the ledger information data, historical meteorological data and historical actual power data of all power generation units in the distributed photovoltaic power station area;

[0008] According to the ledger information data of the power generation units, the power generation units are divided into levels, and according to the geographical location intervals corresponding to the power generation units, the power generation units are divided into several power generation clusters;

[0009] Preprocess the historical power data corresponding to each power generation unit in each power generation cluster to obtain a cluster historical power data set;

[0010] Based on the geographical distribution of power generation clusters, the historical meteorological data is divided into grids to obtain the historical grid meteorological data corresponding to each power generation cluster;

[0011] According to the correlation between the historical power data of each power generation cluster and the corresponding historical grid meteorological data, the future weather of each power generation cluster is predicted to obtain the grid predicted meteorological data of each power generation cluster;

[0012] Based on the historical grid meteorological data and cluster historical power data set corresponding to each power generation cluster, a power prediction model is constructed. The grid predicted meteorological data of each power generation cluster is input into the power prediction model to obtain the predicted power of all power generation units in the distributed photovoltaic power station area.

[0013] Optionally, the specific process of obtaining the ledger information data, historical meteorological data and historical actual power data of all power generation units in the distributed photovoltaic power station area is:

[0014] Collect the ledger information data of all power generation units in the distributed photovoltaic power station area through the power station management system or database interface; obtain historical meteorological data in the distributed photovoltaic power station area from the meteorological data provider or meteorological monitoring system; collect the historical actual power data of all power generation units in the distributed photovoltaic power station area through the power station's data acquisition system or monitoring system.

[0015] Optionally, the specific process of classifying the power generation units step by step according to the ledger information data of the power generation units is:

[0016] According to the supply area information in the ledger information data of the power generation unit, the power generation unit is divided into several supply area clusters;

[0017] In each supply area cluster, the power generation units are divided into several substation clusters according to the grid-connected substation information;

[0018] In each substation cluster, the power generation units are divided into several voltage level clusters according to the voltage level information;

[0019] In each voltage level cluster, the power generation units are divided into dispatch clusters and non-dispatch clusters according to the dispatch attribute information.

[0020] Optionally, the specific process of dividing the power generation units into a plurality of power generation clusters according to the geographical location intervals corresponding to the power generation units is:

[0021] On the basis of the power generation units divided step by step, the power generation units in the same administrative area are clustered according to the administrative area information to which each power generation unit belongs, so as to obtain several power generation clusters. Each power generation cluster is used as a virtual electric field. Each virtual electric field is numbered, recorded and stored.

[0022] Optionally, the specific process of preprocessing the historical power data corresponding to each power generation unit in each power generation cluster to obtain the cluster historical power data set is:

[0023] Review historical power data for each generating unit within each generating cluster, identifying and marking missing or anomalous data points;

[0024] If the proportion of power generation units with missing data at the same time is greater than the preset threshold, the historical power data of all power generation units at this time will be deleted;

[0025] If the proportion of power generation units with missing data is less than the preset threshold, the equal capacity strategy is used to fill the gap;

[0026] The supplemented data are smoothed, sorted and summarized to form a cluster historical power data set for each power generation cluster.

[0027] Optionally, the equal capacity strategy is expressed as shown in the following formula (1):

[0028]

[0029] Among them, M v represents missing values, P l Indicates the historical power value of the power generation unit, C l represents the capacity value of the power generation unit, n represents the number of power generation units with no missing data, C m The capacity of the power generation unit with missing data is proportionally allocated to the power generation units with missing data based on the average power of the power generation units with data in the same cluster.

[0030] Optionally, the specific process of dividing the historical meteorological data into grids based on the geographical location interval distribution of the power generation clusters to obtain the historical grid meteorological data corresponding to each power generation cluster is:

[0031] Determine the geographical area covered by each power generation cluster according to the geographical location information of the power generation cluster;

[0032] According to the resolution and accuracy of meteorological data, the historical meteorological data are divided into several grids, each grid corresponding to a geographical area;

[0033] According to the geographical area covered by each power generation cluster, the grid meteorological data corresponding to each power generation cluster is selected, and the historical grid meteorological data corresponding to each power generation cluster is sorted and summarized.

[0034] Optionally, the specific process of predicting the future weather of each power generation cluster based on the correlation between the cluster historical power data of each power generation cluster and the corresponding historical grid meteorological data to obtain the grid predicted meteorological data of each power generation cluster is:

[0035] Analyze the correlation between the cluster historical power data set and the corresponding historical grid meteorological data of each power generation cluster to determine the key meteorological factors affecting power generation;

[0036] Use the weather forecast model or the services of the weather data provider to forecast the weather data within a specified period of time in the future and obtain the future grid forecast weather data;

[0037] According to the geographical location information of the power generation cluster and the spatial distribution characteristics of the meteorological data, the corresponding grid forecast meteorological data is selected for each power generation cluster;

[0038] The selected grid forecast meteorological data are verified and adjusted to obtain the grid forecast meteorological data of each power generation cluster for subsequent power forecast model input.

[0039] Optionally, the specific process of constructing the power prediction model based on the historical grid meteorological data and cluster historical power data set corresponding to each power generation cluster is:

[0040] Combine the historical grid meteorological data of each power generation cluster and the corresponding cluster historical power data set in chronological order to obtain a time series integrated data set;

[0041] Build an initial power prediction model through machine learning or deep learning algorithms;

[0042] The time series integration data set is divided into a training set and a test set, the initial power prediction model is trained using the training set data, the hyperparameters of the initial power prediction model are adjusted, and the initial power prediction model is optimized;

[0043] After the initial power prediction model training is completed, the model is verified using the test set data to evaluate the prediction accuracy and stability of the initial power prediction model;

[0044] The trained and verified initial power prediction model is used as the power prediction model.

[0045] Optionally, the specific process of inputting the grid forecasted meteorological data of each power generation cluster into the power forecast model to obtain the forecasted power of all power generation units in the distributed photovoltaic power station area is:

[0046] The grid forecast meteorological data corresponding to each power generation cluster is used as input to the constructed and trained power forecast model;

[0047] The power prediction model performs prediction calculations based on the input grid prediction meteorological data and the mapping relationship between the historical grid meteorological data and the cluster historical power data learned within the model, and outputs the predicted power data for each power generation cluster;

[0048] According to the corresponding relationship between power generation clusters and power generation units, the predicted power data of each power generation cluster is decomposed into the corresponding power generation units to obtain the predicted power of all power generation units in the distributed photovoltaic power station area;

[0049] The predicted power data of all power generation units are sorted and summarized to form the final prediction results for power prediction and dispatch management of distributed photovoltaic power stations.

[0050] The technical solution of the present invention has at least the following advantages and beneficial effects:

[0051] Improve prediction accuracy: By obtaining detailed inventory information data, historical meteorological data, and historical actual power data of all power generation units in the distributed photovoltaic power station area, the uniqueness and differences of each power generation unit can be fully considered; clustering is carried out based on the geographical location of the power generation unit, and the model is built based on the historical data of the cluster, which can more effectively capture the power change characteristics in space and time, thereby improving the accuracy of power prediction.

[0052] Optimize resource allocation: By dividing the power generation units into levels and managing them in clusters, the resource allocation of photovoltaic power stations can be more flexibly allocated and optimized, and operation and maintenance plans, energy storage scheduling and power trading can be arranged more effectively, thereby improving overall operational efficiency and economic benefits.

[0053] Enhanced meteorological adaptability: Historical meteorological data is gridded and combined with the historical power data of the power generation cluster. The impact of future meteorological conditions on power is predicted through correlation analysis. This can more accurately reflect the dynamic impact of meteorological factors on photovoltaic power generation and enhance the adaptability of power stations to meteorological changes.

[0054] Improve the level of intelligence: By building a power prediction model and inputting grid prediction meteorological data into the model for prediction, the intelligent prediction of photovoltaic power generation is realized, which not only improves the degree of automation of the prediction, but also provides strong support for the intelligent management and decision-making of the power station.

[0055] Promoting sustainable development: Accurate power forecasting helps photovoltaic power plants better participate in grid dispatching and operation, improves the utilization rate of renewable energy and the stability of the grid; at the same time, by optimizing resource allocation and improving operational efficiency, it helps reduce the cost of photovoltaic power generation and promotes the sustainable development of renewable energy. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 This is a schematic diagram of a process of a distributed photovoltaic power station power prediction method according to Embodiment 1 of the present invention;

[0057] Figure 2 A schematic diagram of a power prediction model construction process of a distributed photovoltaic power station power prediction method according to Embodiment 1 of the present invention;

[0058] Figure 3 A schematic diagram of grid forecast meteorological data of a power generation cluster of a distributed photovoltaic power station power forecasting method according to Embodiment 1 of the present invention;

[0059] Figure 4 This is a schematic diagram of a flow chart of a distributed photovoltaic power station power prediction method according to Embodiment 2 of the present invention;

[0060] Figure 5 This is a schematic diagram of the historical power accumulation of all power generation units in a certain substation from January to February 2022;

[0061] Figure 6 The distributed photovoltaic power station power prediction method according to the second embodiment of the present invention is Figure 5 Schematic diagram after missing data is filled in. DETAILED DESCRIPTION

[0062] The following is a specific implementation method in conjunction with the drawings.

[0063] Example 1

[0064] Reference Figure 1 , Figure 2 , Figure 3 , a distributed photovoltaic power station power prediction method, comprising the steps of:

[0065] Step 1: Obtain the ledger information data, historical meteorological data, and historical actual power data of all power generation units in the distributed photovoltaic power station area. A power generation unit is a power generation unit that is equipped with photovoltaic panels and can perform photoelectric conversion, such as a building or factory with photovoltaic panels installed. A distributed photovoltaic power station contains several power generation units.

[0066] In this embodiment, the specific process of obtaining the ledger information data, historical meteorological data and historical actual power data of all power generation units in the distributed photovoltaic power station area is as follows:

[0067] Collect the ledger information data of all power generation units in the distributed photovoltaic power station area through the power station management system or database interface; obtain historical meteorological data in the distributed photovoltaic power station area from the meteorological data provider or meteorological monitoring system; collect the historical actual power data of all power generation units in the distributed photovoltaic power station area through the power station's data acquisition system or monitoring system.

[0068] Step 2: According to the ledger information data of the power generation units, the power generation units are divided into levels, and according to the geographical location intervals corresponding to the power generation units, the power generation units are divided into several power generation clusters.

[0069] In this embodiment, the specific process is:

[0070] According to the supply area information in the ledger information data of the power generation unit, the power generation unit is divided into several supply area clusters; within each supply area cluster, according to the grid-connected substation information, the power generation unit is divided into several substation clusters; within each substation cluster, according to the voltage level information, the power generation unit is divided into several voltage level clusters; within each voltage level cluster, according to the scheduling attribute information, the power generation unit is divided into scheduling clusters and non-scheduling clusters.

[0071] On the basis of the power generation units divided step by step, the power generation units in the same administrative area are clustered according to the administrative area information to which each power generation unit belongs, so as to obtain several power generation clusters. Each power generation cluster is used as a virtual electric field. Each virtual electric field is numbered, recorded and stored.

[0072] The clustering process of power generation units in the distributed photovoltaic power station area can be realized by setting the clustering function. Assume that the input of the clustering function is DataFrame, which contains the ledger information of the power generation unit to be predicted. The specific parameters are shown in Table 1 below:

[0073] Table 1: Cluster function input parameter table

[0074]

[0075]

[0076] The return value of the cluster function is shown in Table 2 below:

[0077] Table 2: Cluster function return value table

[0078] Parameter name Parameter Description Parameter Type wfid Virtual id after clustering string user_id User Number int powercap Capacity of user_id float

[0079] According to the ledger data in the project, clusters are divided according to different supply areas (220kV), different substations (110kV), different voltage levels (10kV, 220 / 380V) and different dispatch types (unified dispatch, non-unified dispatch), and then clusters are divided according to different administrative regions to obtain virtual electric fields. Each virtual electric field contains multiple user numbers (hereinafter referred to as household numbers), and the virtual electric field information remains unchanged (unless there is new attribute information such as administrative regions). The association between the virtual electric field and the household number is updated according to the ledger data update, and the divided power generation clusters are stored in the basic information library of the system database. Convert the power data into a unified unit, such as kilowatts (kW) or megawatts (MW), to ensure the comparability of data between different power generation units; if there are different time resolutions in the data (such as every hour, every minute, etc.), the time resolution needs to be unified for subsequent analysis. For each power generation cluster, the historical power data of all power generation units in the cluster are aggregated to obtain the total power data of the cluster; the data can be summarized according to the time window (such as one day, one week, one month, etc.) to analyze the power changes in different time periods. Perform quality checks on the preprocessed data to ensure data accuracy, completeness, and consistency; use statistical methods (such as mean, standard deviation, maximum, minimum, etc.) to describe the distribution characteristics of the data. Store the preprocessed cluster historical power data set in the system database or data warehouse for subsequent analysis and modeling; ensure data accessibility and security to prevent data leakage or damage. Example of cluster mode and virtual electric field generation:

[0080] Example 1: A virtual electric field is generated for the town and street under the 10kV electric field under the 110kV substation under the 220kV supply area, corresponding to multiple household numbers;

[0081] Example 2: A virtual electric field is generated for the town and street under the non-centrally-controlled 10kV electric field under the 110kV substation under the 220kV supply area, corresponding to multiple household numbers;

[0082] Example 3: A virtual electric field is generated in the towns and streets under the 220 / 380V electric field under the 110kV substation under the 220kV supply area, corresponding to multiple household numbers.

[0083] Step 3: pre-process the historical power data corresponding to each power generation unit in each power generation cluster to obtain a cluster historical power data set.

[0084] In this embodiment, the preprocessing method is: check the abnormal values ​​in the data, such as a sudden increase or decrease in power value, unreasonable negative values, etc., and correct or eliminate them; process missing values, which can be estimated through interpolation methods (such as linear interpolation, Lagrange interpolation, etc.) or based on the data trends of adjacent time points; remove duplicate data to ensure that the data corresponding to each timestamp is unique.

[0085] Step 4: Based on the geographical distribution of the power generation clusters, the historical meteorological data is divided into grids to obtain the historical grid meteorological data corresponding to each power generation cluster.

[0086] In this embodiment, the specific process is:

[0087] According to the geographical location information of the power generation cluster, the geographical area covered by each power generation cluster is determined. The specific boundaries of these geographical areas can be determined with the assistance of the Geographic Information System (GIS); according to the resolution and accuracy of the meteorological data, the historical meteorological data is divided into several grids, each grid corresponds to a geographical area, and the size and shape of the grid can be adjusted according to the actual situation, but it should be ensured that each power generation cluster can be completely contained in one or more grids. The division of the grid should take into account the availability and accuracy of the meteorological data, as well as the limitation of computing resources; according to the geographical area covered by each power generation cluster, the grid meteorological data corresponding to each power generation cluster is selected, and the historical meteorological data (such as temperature, humidity, wind speed, irradiance, etc.) is mapped to each grid. If the resolution of the historical meteorological data does not match the grid division, interpolation or other data processing techniques need to be used to fill or adjust the meteorological data; the historical grid meteorological data corresponding to each power generation cluster is sorted and summarized, and the quality of the sorted historical grid meteorological data is checked and verified to check the integrity, accuracy and consistency of the data to ensure that there are no missing values, outliers or inconsistent data. If problems are found, data cleaning, correction or re-collection is required.

[0088] Step 5: According to the correlation between the historical cluster power data of each power generation cluster and the corresponding historical grid meteorological data, the future weather of each power generation cluster is predicted to obtain the grid predicted meteorological data of each power generation cluster.

[0089] In this embodiment, the specific process is:

[0090] Analyze the correlation between the cluster historical power data set of each power generation cluster and the corresponding historical grid meteorological data. The correlation between the cluster historical power data set of each power generation cluster and the corresponding historical grid meteorological data can be obtained through statistical analysis and data mining techniques, such as correlation coefficient calculation, scatter plot drawing, linear regression analysis, etc., to determine the key meteorological factors affecting power generation, such as irradiance, temperature, humidity, wind speed, etc.; use meteorological prediction models or the services of meteorological data providers to predict meteorological data within a specified period of time in the future to obtain future grid predicted meteorological data; according to the determined key meteorological factors, select appropriate meteorological prediction models or build new meteorological prediction models. These models can be physics-based models (such as numerical weather forecast models) or data-driven models (such as machine learning models); select corresponding grid predicted meteorological data for each power generation cluster according to the geographical location information of the power generation cluster and the spatial distribution characteristics of meteorological data; verify and adjust the selected grid predicted meteorological data to obtain grid predicted meteorological data for each power generation cluster, such as Figure 3 As shown, it is used as the input of the subsequent power prediction model. The verification may include comparison with known historical meteorological data to ensure the rationality and accuracy of the prediction results, smoothing the prediction data, removing outliers, or making corrections based on professional knowledge.

[0091] Step 6: Based on the historical grid meteorological data and cluster historical power data set corresponding to each power generation cluster, a power prediction model is constructed, and the grid prediction meteorological data of each power generation cluster is input into the power prediction model to obtain the predicted power of all power generation units in the distributed photovoltaic power station area.

[0092] In this embodiment, the specific process is:

[0093] Combine the historical grid meteorological data of each power generation cluster and the corresponding cluster historical power data set in chronological order to obtain a time series integrated data set; construct an initial power prediction model through machine learning or deep learning algorithms; divide the time series integrated data set into a training set and a test set, use the training set data to train the initial power prediction model, adjust the hyperparameters of the initial power prediction model, and optimize the initial power prediction model; after the initial power prediction model is trained, use the test set data to verify the model and evaluate the prediction accuracy and stability of the initial power prediction model; use the trained and verified initial power prediction model as the power prediction model.

[0094] In this embodiment, the benchmark model in the distributed photovoltaic power station can use the SVM, MLP, KerasG and other models designed for the centralized power station. In addition, since the grid prediction meteorological data of the distributed photovoltaic power station is in the data format of the grid image, for the distributed photovoltaic power station, Figure 2As shown, in this embodiment, the initial power prediction model is constructed based on the ConvLSTM network structure combined with a deep learning algorithm. The difference from the LSTM algorithm is that the input of the ConvLSTM model is image data. The ConvLSTM model replaces the fully connected operation in the LSTM network with a convolutional form, which can better extract the features of the area represented by the image, and these features also have time series features. Figure 2 Here G represents grid irradiance, P represents power, t represents time t, and tn represents time n before t.

[0095] In this embodiment, the specific process of inputting the grid forecasted meteorological data of each power generation cluster into the power forecast model to obtain the forecasted power of all power generation units in the distributed photovoltaic power station area is as follows:

[0096] The grid forecast meteorological data corresponding to each power generation cluster is taken as input and input into the constructed and trained power prediction model; the power prediction model performs forecast calculation based on the input grid forecast meteorological data and the mapping relationship between the historical grid meteorological data and the cluster historical power data learned inside the model, and outputs the forecast power data of each power generation cluster; according to the corresponding relationship between the power generation cluster and the power generation unit, the forecast power data of each power generation cluster is decomposed into the corresponding power generation units to obtain the forecast power of all power generation units in the distributed photovoltaic power station area; the forecast power data of all power generation units are sorted and summarized to form the final forecast result, which is used for power forecasting and scheduling management of distributed photovoltaic power stations.

[0097] Example 2

[0098] Based on Example 1, refer to Figure 3 , Figure 4 , Figure 5 In this embodiment, after the distributed photovoltaic power station cluster, there may be data quality problems such as missing power generation unit data and data anomalies in the same power generation cluster. In order to ensure the correctness and integrity of the data, the data quality detection is processed. Taking a substation as an example, the historical power data of distributed photovoltaic power generation obtained in the early stage is seriously missing. The historical power accumulation of all power generation units from January to February 2022 is as follows Figure 4 As shown in the figure, the historical power data of all power generation units on some dates are missing, and the overall power generation situation is unstable, which is mainly reflected in the sudden increase of power generation on some dates. In view of the missing historical power data of some power generation units, the historical power data corresponding to each power generation unit in each power generation cluster is preprocessed to obtain the cluster historical power data set. The specific process is:

[0099] Check the historical power data of each power generation unit in each power generation cluster, identify and mark missing or abnormal data points; if the proportion of power generation units with missing data at the same time is greater than the preset threshold, delete the historical power data of all power generation units at this time; if the proportion of power generation units with missing data is less than the preset threshold, adopt the equal capacity strategy to fill the data; smooth the filled data, organize and summarize the smoothed data to form a cluster historical power data set for each power generation cluster.

[0100] In this embodiment, the expression of the equal capacity strategy is shown in the following formula (1):

[0101]

[0102] Among them, M v represents missing values, P l Indicates the historical power value of the power generation unit, C l represents the capacity value of the power generation unit, n represents the number of power generation units with no missing data, C m The capacity of the power generation unit with missing data is proportionally allocated to the power generation units with missing data according to the average power of the power generation units with data in the same cluster. Figure 5 As shown in the red box, the data has improved after the supplementation.

Claims

1. A distributed photovoltaic power station power prediction method, characterized in that: Includes steps: Obtain the ledger information data, historical meteorological data and historical actual power data of all power generation units in the distributed photovoltaic power station area; According to the ledger information data of the power generation units, the power generation units are divided into levels, and according to the geographical location intervals corresponding to the power generation units, the power generation units are divided into several power generation clusters; Preprocess the historical power data corresponding to each power generation unit in each power generation cluster to obtain a cluster historical power data set; Based on the geographical distribution of power generation clusters, the historical meteorological data is divided into grids to obtain the historical grid meteorological data corresponding to each power generation cluster; According to the correlation between the historical power data of each power generation cluster and the corresponding historical grid meteorological data, the future weather of each power generation cluster is predicted to obtain the grid predicted meteorological data of each power generation cluster; Based on the historical grid meteorological data and cluster historical power data set corresponding to each power generation cluster, a power prediction model is constructed. The grid predicted meteorological data of each power generation cluster is input into the power prediction model to obtain the predicted power of all power generation units in the distributed photovoltaic power station area.

2. The distributed photovoltaic power station power prediction method according to claim 1, characterized in that: The specific process of obtaining the ledger information data, historical meteorological data and historical actual power data of all power generation units in the distributed photovoltaic power station area is as follows: Collect the ledger information data of all power generation units in the distributed photovoltaic power station area through the power station management system or database interface; obtain historical meteorological data in the distributed photovoltaic power station area from the meteorological data provider or meteorological monitoring system; collect the historical actual power data of all power generation units in the distributed photovoltaic power station area through the power station's data acquisition system or monitoring system.

3. The distributed photovoltaic power station power prediction method according to claim 1, characterized in that: The specific process of classifying the power generation units step by step according to the ledger information data of the power generation units is as follows: According to the supply area information in the ledger information data of the power generation unit, the power generation unit is divided into several supply area clusters; In each supply area cluster, the power generation units are divided into several substation clusters according to the grid-connected substation information; In each substation cluster, the power generation units are divided into several voltage level clusters according to the voltage level information; In each voltage level cluster, the power generation units are divided into dispatch clusters and non-dispatch clusters according to the dispatch attribute information.

4. The distributed photovoltaic power station power prediction method according to claim 1, characterized in that: The specific process of dividing the power generation units into a plurality of power generation clusters according to the geographical location intervals corresponding to the power generation units is as follows: On the basis of the power generation units divided step by step, the power generation units in the same administrative area are clustered according to the administrative area information to which each power generation unit belongs, so as to obtain several power generation clusters. Each power generation cluster is used as a virtual electric field. Each virtual electric field is numbered, recorded and stored.

5. The distributed photovoltaic power station power prediction method according to claim 1, characterized in that: The specific process of preprocessing the historical power data corresponding to each power generation unit in each power generation cluster to obtain the cluster historical power data set is: Review historical power data for each generating unit within each generating cluster, identifying and marking missing or anomalous data points; If the proportion of power generation units with missing data at the same time is greater than the preset threshold, the historical power data of all power generation units at this time will be deleted; If the proportion of power generation units with missing data is less than the preset threshold, the equal capacity strategy is used to fill the gap; The supplemented data are smoothed, sorted and summarized to form a cluster historical power data set for each power generation cluster.

6. The distributed photovoltaic power station power prediction method according to claim 5, characterized in that: The expression of the equal capacity strategy is shown in the following formula (1): Among them, M v represents missing values, P l Indicates the historical power value of the power generation unit, C l represents the capacity value of the power generation unit, n represents the number of power generation units with no missing data, C m The capacity of the power generation unit with missing data is proportionally allocated to the power generation units with missing data based on the average power of the power generation units with data in the same cluster.

7. The distributed photovoltaic power station power prediction method according to claim 1, characterized in that: The specific process of dividing the historical meteorological data into grids based on the geographical location interval distribution of the power generation cluster to obtain the historical grid meteorological data corresponding to each power generation cluster is as follows: Determine the geographical area covered by each power generation cluster according to the geographical location information of the power generation cluster; According to the resolution and accuracy of meteorological data, the historical meteorological data are divided into several grids, each grid corresponding to a geographical area; According to the geographical area covered by each power generation cluster, the grid meteorological data corresponding to each power generation cluster is selected, and the historical grid meteorological data corresponding to each power generation cluster is sorted and summarized.

8. The distributed photovoltaic power station power prediction method according to claim 1, characterized in that: The specific process of predicting the future weather of each power generation cluster based on the correlation between the cluster historical power data of each power generation cluster and the corresponding historical grid meteorological data to obtain the grid predicted meteorological data of each power generation cluster is as follows: Analyze the correlation between the cluster historical power data set and the corresponding historical grid meteorological data of each power generation cluster to determine the key meteorological factors affecting power generation; Use the weather forecast model or the services of the weather data provider to forecast the weather data within a specified period of time in the future and obtain the future grid forecast weather data; According to the geographical location information of the power generation cluster and the spatial distribution characteristics of the meteorological data, the corresponding grid forecast meteorological data is selected for each power generation cluster; The selected grid forecast meteorological data are verified and adjusted to obtain the grid forecast meteorological data of each power generation cluster for subsequent power forecast model input.

9. The distributed photovoltaic power station power prediction method according to claim 1, characterized in that: The specific process of constructing the power prediction model based on the historical grid meteorological data and cluster historical power data set corresponding to each power generation cluster is as follows: Combine the historical grid meteorological data of each power generation cluster and the corresponding cluster historical power data set in chronological order to obtain a time series integrated data set; Build an initial power prediction model through machine learning or deep learning algorithms; The time series integration data set is divided into a training set and a test set, the initial power prediction model is trained using the training set data, the hyperparameters of the initial power prediction model are adjusted, and the initial power prediction model is optimized; After the initial power prediction model training is completed, the model is verified using the test set data to evaluate the prediction accuracy and stability of the initial power prediction model; The trained and verified initial power prediction model is used as the power prediction model.

10. The distributed photovoltaic power station power prediction method according to claim 9, characterized in that: The specific process of inputting the grid forecasted meteorological data of each power generation cluster into the power forecast model to obtain the forecasted power of all power generation units in the distributed photovoltaic power station area is as follows: The grid forecast meteorological data corresponding to each power generation cluster is used as input to the constructed and trained power forecast model; The power prediction model performs prediction calculations based on the input grid prediction meteorological data and the mapping relationship between the historical grid meteorological data and the cluster historical power data learned within the model, and outputs the predicted power data for each power generation cluster; According to the corresponding relationship between power generation clusters and power generation units, the predicted power data of each power generation cluster is decomposed into the corresponding power generation units to obtain the predicted power of all power generation units in the distributed photovoltaic power station area; The predicted power data of all power generation units are sorted and summarized to form the final prediction results for power prediction and dispatch management of distributed photovoltaic power stations.

Images