Zone area photovoltaic output prediction method and system combined with power grid heterogeneous data sampling

Abstract

The invention discloses a transformer area photovoltaic output prediction method and system combined with power grid heterogeneous data sampling, and relates to the technical field of photovoltaic output prediction, and the method comprises the steps: constructing a historical output data set; constructing a user output influence curve; constructing a meteorological factor output influence curve; constructing a historical output model; acquiring a real-time power utilization parameter set sequence and a real-time power generation parameter set sequence of the photovoltaic user set in a preset acquisition window, and constructing a real-time output model; and optimizing the historical output model by using the real-time output model to obtain a photovoltaic output prediction model. According to the invention, the technical problems of relatively lagging photovoltaic output prediction and low prediction reliability in the prior art are solved, and the technical effect of improving the photovoltaic prediction accuracy is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power output prediction, and particularly to a method and system for predicting the photovoltaic power output of a distribution transformer area by combining heterogeneous data sampling of the power grid. Background Art

[0002] With the continuous development of renewable energy, photovoltaic power generation has become one of the important green energy sources globally. However, due to the influence of various factors such as meteorological conditions, environmental factors, and power grid demand on photovoltaic power generation, its power generation output has significant volatility and uncertainty. Therefore, accurate prediction of photovoltaic power output is of great significance for power grid dispatching, energy management, and system stability.

[0003] There are technical problems in the prior art such as relatively lagged prediction of photovoltaic power output and low prediction reliability. Summary of the Invention

[0004] The present application provides a method and system for predicting the photovoltaic power output of a distribution transformer area by combining heterogeneous data sampling of the power grid, which is used to solve the technical problems of relatively lagged prediction of photovoltaic power output and low prediction reliability in the prior art.

[0005] In view of the above problems, the present application provides a method and system for predicting the photovoltaic power output of a distribution transformer area by combining heterogeneous data sampling of the power grid.

[0006] In the first aspect of the present application, a method for predicting the photovoltaic power output of a distribution transformer area by combining heterogeneous data sampling of the power grid is provided. The method includes: Obtaining a historical power consumption data set and a historical power generation data set of a photovoltaic user set in a target distribution transformer area within a historical window, performing power output analysis, and constructing a historical power output data set; Obtaining a historical power consumption behavior data set of the photovoltaic user set within the historical window, combining the historical power output data set for data analysis, and constructing a user power output influence curve; Obtaining a meteorological data set within the historical window, combining the historical power output data set for data analysis, and constructing a meteorological factor power output influence curve; Constructing a historical power output model based on the user power output influence curve, the meteorological factor power output influence curve, the historical power consumption data set, the historical power generation data set, and the historical power output data set; Obtaining a sequence of real-time power consumption parameter sets and a sequence of real-time power generation parameter sets of the photovoltaic user set within a preset acquisition window, and constructing a real-time power output model; Optimizing the historical power output model by using the real-time power output model to obtain a photovoltaic power output prediction model.

[0007] Preferably, obtain the historical power consumption data set and historical power generation data set of the photovoltaic user set in the target power grid area within the historical window, perform output analysis, and construct a historical output data set, including: calculating the mapped power difference between the historical power consumption data set and the historical power generation data set to obtain an initial historical output data set; screening outliers from the initial historical output data set to obtain the initial historical output data set.

[0008] Preferably, obtain the historical power consumption behavior data set of the photovoltaic user set within the historical window, and perform data analysis in combination with the historical output data set to construct a user output influence curve, including: constructing a two-dimensional coordinate system with historical output data as the abscissa and historical power consumption behavior data as the ordinate; inputting the historical power consumption behavior data set and the historical output data set into the two-dimensional coordinate system to obtain a first influence scatter point set; fitting the first influence scatter point set to obtain the user output influence curve.

[0009] Preferably, fitting the first influence scatter point set to obtain the user output influence curve includes: preliminarily fitting the first influence scatter point set using a polynomial regression model to obtain a first fitting curve; statistically calculating the first fitting neighborhood of the first fitting curve according to a preset tolerance bandwidth, and statistically calculating the number of scatter points in the first fitting neighborhood to obtain the number of scatter points in the first fitting neighborhood; determining whether the number of scatter points above the first fitting curve is greater than the number of scatter points below the first fitting curve. If so, move the first fitting curve upward according to a preset moving step to obtain a second fitting curve; construct a second fitting neighborhood of the second fitting curve according to a preset tolerance bandwidth, and statistically calculate the number of scatter points in the second fitting neighborhood to obtain the number of scatter points in the second fitting neighborhood; when the number of scatter points in the second fitting neighborhood is greater than or equal to the number of scatter points in the first fitting neighborhood, continue to move the second fitting curve upward according to a preset moving step until a preset number of moves is satisfied to obtain the user output influence curve.

[0010] Preferably, construct a historical output model based on the user output influence curve, meteorological factor output influence curve, historical power consumption data set, historical power generation data set, and the historical output data set, including: randomly extracting data multiple times from the historical power consumption data set and the historical power generation data set, and performing mapped extraction on the historical output data set according to the extraction results, and combining the user output influence curve and the meteorological factor output influence curve to obtain a support set and a query set; using the support set to perform supervised training on a framework constructed based on a feedforward neural network, and using the query set to verify the trained framework until convergence to obtain the trained historical output model.

[0011] Preferably, a real-time output model is constructed by obtaining a sequence of real-time power consumption parameter sets and a sequence of real-time power generation parameter sets of a photovoltaic user set within a preset collection window, including: respectively performing parameter centralized screening on the sequence of real-time power consumption parameter sets and the sequence of real-time power generation parameter sets to obtain a centralized real-time power consumption parameter sequence and a centralized real-time power generation parameter sequence; calculating the mapping difference between the centralized real-time power consumption parameter sequence and the centralized real-time power generation parameter sequence to obtain a centralized real-time output data sequence; and constructing the real-time output model based on the centralized real-time power consumption parameter sequence, the centralized real-time power generation parameter sequence, and the centralized real-time output data sequence.

[0012] Preferably, respectively performing parameter centralized screening on the sequence of real-time power consumption parameter sets and the sequence of real-time power generation parameter sets to obtain a centralized real-time power consumption parameter sequence and a centralized real-time power generation parameter sequence includes: traversing the set means of the sequence of real-time power consumption parameter sets and the sequence of real-time power generation parameter sets to obtain a real-time power consumption parameter mean sequence and a real-time power generation parameter mean sequence; respectively using the real-time power consumption parameter mean sequence and the real-time power generation parameter mean sequence as the centralized screening starting points, and using a parameter centralized screening function to update and iterate the centralized screening starting points until a preset number of update iterations is satisfied, so as to obtain a centralized real-time power consumption parameter sequence and a centralized real-time power generation parameter sequence.

[0013] Preferably, the parameter centralized screening function is: ; Wherein, is the centralized screening iteration point, is the Gaussian weight kernel function, is the centralized screening starting point, is the near neighbor neighborhood composed of real-time power consumption parameters in the real-time power consumption parameter set whose distance to the centralized screening starting point is less than a preset value, is centered on the th real-time power consumption parameter in the near neighbor neighborhood, and the neighborhood density of the obtained neighborhood constructed with a preset value as the radius, is the th real-time power consumption parameter in the near neighbor neighborhood, is the Gaussian function with weight decay, is the parameter for controlling the decay speed.

[0014] Preferably, each real-time power consumption parameter in the sequence of real-time power consumption parameter sets includes current, voltage, and power, and each real-time power generation parameter in the sequence of real-time power generation parameter sets includes power generation power and power peak.

[0015] In the second aspect of the present application, a photovoltaic output prediction system for a distribution transformer area combined with grid heterogeneous data sampling is provided. The system includes: Historical output data set construction module, which is used to obtain the historical power consumption data set and historical power generation data set of the photovoltaic user set in the target substation area within the historical window, perform output analysis, and construct a historical output data set; User output influence curve construction module, which is used to obtain the historical power consumption behavior data set of the photovoltaic user set within the historical window, combine the historical output data set to perform data analysis, and construct a user output influence curve; Meteorological factor output influence curve construction module, which is used to obtain the meteorological data set within the historical window, combine the historical output data set to perform data analysis, and construct a meteorological factor output influence curve; Historical output model construction module, which is used to construct a historical output model based on the user output influence curve, meteorological factor output influence curve, historical power consumption data set, historical power generation data set, and the historical output data set; Real-time output model construction module, which is used to obtain the real-time power consumption parameter set sequence and real-time power generation parameter set sequence of the photovoltaic user set within the preset acquisition window, and construct a real-time output model; Photovoltaic output prediction model acquisition module, which is used to optimize the historical output model by using the real-time output model to obtain a photovoltaic output prediction model.

[0016] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application obtains the historical power consumption data set and historical power generation data set of the photovoltaic user set in the target substation area within the historical window, performs output analysis, constructs a historical output data set, then obtains the historical power consumption behavior data set of the photovoltaic user set within the historical window, combines the historical output data set to perform data analysis, constructs a user output influence curve, and further obtains the meteorological data set within the historical window, combines the historical output data set to perform data analysis, constructs a meteorological factor output influence curve, then constructs a historical output model based on the user output influence curve, meteorological factor output influence curve, historical power consumption data set, historical power generation data set, and historical output data set, obtains the real-time power consumption parameter set sequence and real-time power generation parameter set sequence of the photovoltaic user set within the preset acquisition window, constructs a real-time output model, and then optimizes the historical output model by using the real-time output model to obtain a photovoltaic output prediction model. It achieves the technical effect of improving the reliability of photovoltaic output prediction and enhancing the accuracy of photovoltaic output prediction in the substation area. Description of the Drawings

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0018] Figure 1 Schematic flowchart of the method for predicting the photovoltaic output of a distribution transformer area by combining the sampling of heterogeneous power grid data provided by the embodiments of the present application; Figure 2 Schematic structural diagram of the system for predicting the photovoltaic output of a distribution transformer area by combining the sampling of heterogeneous power grid data provided by the embodiments of the present application.

[0019] Explanation of reference numerals: Historical output data set construction module 11, User output influence curve construction module 12, Meteorological factor output influence curve construction module 13, Historical output model construction module 14, Real-time output model construction module 15, Photovoltaic output prediction model acquisition module 16. Detailed implementation manners

[0020] The present application provides a method and system for predicting the photovoltaic output of a distribution transformer area by combining the sampling of heterogeneous power grid data, which are used to solve the technical problems of relatively lagging photovoltaic output prediction and low prediction reliability in the prior art.

[0021] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts belong to the scope of protection of the present application.

[0022] It should be noted that the terms "include" and "have" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units does not necessarily have to be limited to those clearly listed steps or units, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products, or devices.

[0023] Embodiment 1, as Figure 1 shown, the present application provides a method for predicting the photovoltaic output of a distribution transformer area by combining the sampling of heterogeneous power grid data, and the method includes: S1: Obtain the historical electricity consumption data set and historical power generation data set of the photovoltaic user set in the target distribution transformer area within the historical window, perform output analysis, and construct a historical output data set; Further, obtain the historical power consumption data set and historical power generation data set of the photovoltaic user set in the target area within the historical window, perform output analysis, and construct a historical output data set. Step S1 of the embodiment of the present application further includes: Calculate the mapped power difference between the historical power consumption data set and the historical power generation data set to obtain an initial historical output data set; Filter out outliers from the initial historical output data set to obtain the initial historical output data set.

[0024] In a possible embodiment, the target area is any area that needs to perform photovoltaic processing prediction. The photovoltaic user set is all photovoltaic power generation users within a specific area range, and the power data of these users will be used to analyze and predict the output of photovoltaic power generation. The historical window refers to the time period used to collect historical data. For example, the past month or year, and the data during this period will be used to train the output prediction model. The historical power consumption data set reflects the power consumption data of photovoltaic users within the historical window, including power consumption, power consumption time period, etc. The historical power generation data set reflects the photovoltaic power generation data of photovoltaic users within the historical window, including power generation amount, power generation time period, etc. The historical output data set is a data set obtained through output analysis, which reflects the actual power generation capacity of the photovoltaic system at different time periods and its relationship with power consumption demand.

[0025] Preferably, it means mapping and comparing the historical power consumption data set with the historical power generation data set, and calculating the difference between them to obtain the initial historical output data set. That is, calculate the difference between the power consumption in the historical power consumption data and the power generation amount in the historical power generation data to obtain an initial historical output. In the initial historical output data set, detect and delete those data points that significantly deviate from the normal range. These data points are usually caused by sensor failures, data acquisition errors, or other factors. Among them, outliers refer to the values in the data set that significantly deviate from most data points. Removing these abnormal values from the data set can ensure the accuracy of the data and the stability of the model.

[0026] S2: Obtain the historical power consumption behavior data set of the photovoltaic user set within the historical window, and perform data analysis in combination with the historical output data set to construct a user output influence curve; Further, obtain the historical power consumption behavior data set of the photovoltaic user set within the historical window, and perform data analysis in combination with the historical output data set to construct a user output influence curve. Step S2 of the embodiment of the present application further includes: Construct a two-dimensional coordinate system with historical output data as the abscissa and historical power consumption behavior data as the ordinate; Input the historical electricity consumption behavior data set and the historical output data set into the two-dimensional coordinate system to obtain a first set of influence scatter points; Fit the first set of influence scatter points to obtain the user output influence curve.

[0027] Furthermore, when fitting the first set of influence scatter points to obtain the user output influence curve, step S2 of this application embodiment further includes: Use a polynomial regression model to preliminarily fit the first set of influence scatter points to obtain a first fitting curve; According to a preset tolerance bandwidth, count the first fitting neighborhood of the first fitting curve, and count the number of scatter points in the first fitting neighborhood to obtain the number of scatter points in the first fitting neighborhood; Judge whether the number of scatter points above the first fitting curve is greater than the number of scatter points below the first fitting curve. If so, move the first fitting curve upward according to a preset moving step to obtain a second fitting curve; Construct a second fitting neighborhood of the second fitting curve according to a preset tolerance bandwidth, and count the number of scatter points in the second fitting neighborhood to obtain the number of scatter points in the second fitting neighborhood; When the number of scatter points in the second fitting neighborhood is greater than or equal to the number of scatter points in the first fitting neighborhood, continue to move the second fitting curve upward according to a preset moving step until a preset number of moves is satisfied to obtain the user output influence curve.

[0028] In a possible embodiment, the historical electricity consumption behavior data set is the electricity consumption data of a photovoltaic user within a historical window, including the user's electricity consumption, electricity consumption patterns (such as the electricity consumption distribution during peak periods and low periods), and seasonal or temporal changes in electricity consumption. The electricity consumption behavior data helps to identify the user's electricity consumption pattern, which in turn affects power generation and output prediction. The user output influence curve refers to the curve obtained by analyzing the relationship between the user's electricity consumption behavior and the historical output data, showing the influence of different electricity consumption behaviors on the output of the photovoltaic system.

[0029] In a possible embodiment, in the two-dimensional coordinate system, the historical output data is used as the abscissa (X-axis), and the historical electricity consumption behavior data is used as the ordinate (Y-axis). This coordinate system is used to visually display these two data sets, facilitating further analysis of their relationship. The first set of influence scatter points is a scatter plot set formed by corresponding input of the historical electricity consumption behavior data set and the historical output data set into the two-dimensional coordinate system. Each scatter point represents the corresponding relationship between the electricity consumption behavior and the photovoltaic output within a specific time period. Through these scatter points, the change trend of the photovoltaic output under different electricity consumption behaviors can be observed.

[0030] By fitting these scattered points and using fitting algorithms (such as polynomial regression, etc.), the best-matching curve is found from the set of scattered points, thereby obtaining the user output influence curve. This curve can effectively reflect the specific impact of the user's electricity consumption behavior on the photovoltaic power generation, providing an important basis for subsequent photovoltaic output prediction. Through the fitted curve, the output change of the photovoltaic system under different user behavior patterns can be predicted more accurately.

[0031] Preferably, the polynomial regression model is a mathematical method for fitting data. It fits the scattered point data by using polynomials (such as quadratic, cubic polynomials, etc.). The polynomial regression model finds the best-fitting curve for the data points by minimizing the error. The first fitting curve is the curve obtained by initially fitting the first set of influence scattered points, representing the relationship between the electricity consumption behavior of photovoltaic users and historical output data. The preset tolerance bandwidth is a tolerance area preset by those skilled in the art near the fitting curve, indicating the maximum allowable distance that the scattered points can deviate from the fitting curve when they belong to the curve fitting neighborhood. It is used to determine whether the data points are within the fitting range. The first fitting neighborhood refers to an area around the first fitting curve, which is determined by the tolerance bandwidth. The number of scattered points in this neighborhood is used to judge the accuracy of the fitting curve. The number of scattered points in the fitting neighborhood is used to measure the fitting effect of the fitting curve on the data points. If this number is too small, it means the fitting effect is not good and the curve needs to be adjusted. The preset moving step is the step size for a single adjustment of the fitting curve preset by those skilled in the art, used to gradually adjust the position of the curve to ensure the best fitting effect.

[0032] Exemplarily, a quadratic polynomial regression model is selected to initially fit the first set of influence scattered points, and its form is: , where is the historical output data, is the historical electricity consumption behavior data, are the coefficients to be determined. The least squares method is used to minimize the error between the fitting curve and the data points. The coefficients of the polynomial regression model are continuously adjusted to minimize the sum of the squares of the errors (i.e., residuals) between the model prediction values and the actual data points.

[0033] Preferably, regression modeling is performed on the historical data points (i.e., the set of scattered points) to fit a regression curve. For each , the model calculates the corresponding , and then, the coefficients of the polynomial regression model are determined. By minimizing the sum of the squares of the residuals, the parameters of the best-fitting curve are found to obtain the first fitting curve. The scattered points in the first set of influence scattered points whose distances to the first fitting curve are within the preset tolerance bandwidth range are divided into the first fitting neighborhood, and the number of scattered points in the first fitting neighborhood is counted to obtain the number of scattered points in the first fitting neighborhood.

[0034] Determine whether the number of scatter points above the first fitting curve is greater than the number of scatter points below the first fitting curve. If so, it indicates that the area where the scatter points are more concentrated is above the first fitting curve. Therefore, determine the moving direction as upward, and move the first fitting curve upward according to a preset moving step to obtain a second fitting curve.

[0035] Based on the same construction principle as the first fitting neighborhood, construct the second fitting neighborhood in the first set of influential scatter points, and count the number of scatter points in the second fitting neighborhood to obtain the scatter point quantity of the second fitting neighborhood. Further, compare the scatter point quantity of the second fitting neighborhood with that of the first fitting neighborhood to determine whether to continue moving. When the scatter point quantity of the second fitting neighborhood is greater than or equal to that of the first fitting neighborhood, it indicates that the second fitting curve can better reflect the general situation of the first set of influential scatter points than the first fitting curve. At this time, continue to move the second fitting curve upward according to the preset moving step until the preset number of moving times is satisfied, and obtain the user power output influence curve. The user power output influence curve accurately reflects the influence of user electricity consumption behavior on photovoltaic power generation output. Through multiple adjustments and movements, the technical effect of providing reliable data support for subsequent construction of the historical power output model is achieved.

[0036] S3: Obtain the set of meteorological data within the historical window, perform data analysis in combination with the historical power output data set, and construct a meteorological factor power output influence curve; In a possible embodiment, the set of meteorological data includes meteorological data such as solar radiation, temperature, and wind speed. The meteorological factor power output influence curve reflects the influence of different weather conditions on the power output situation within the historical window. Use the historical power output data as the abscissa and the meteorological data as the ordinate to construct a meteorological two-dimensional coordinate system, and input the historical electricity consumption behavior data set and the set of meteorological data into the meteorological two-dimensional coordinate system to obtain a second set of influential scatter points. Furthermore, based on the same principle as obtaining the user power output influence curve, fit the second set of influential scatter points to obtain the meteorological factor power output influence curve. Among them, the meteorological factor power output influence curve reflects the influence of meteorological factors (such as light intensity, temperature, etc.) on the power generation output of the photovoltaic system, and is used to help predict the power generation capacity of the photovoltaic system under different meteorological conditions. By constructing a curve that reflects the influence of meteorological factors on the power output of the photovoltaic system, the technical effect of providing an accurate meteorological influence basis for subsequent construction of the historical power output model is achieved.

[0037] S4: Construct a historical power output model based on the user power output influence curve, the meteorological factor power output influence curve, the historical electricity consumption data set, the historical power generation data set, and the historical power output data set; Further, based on the user output influence curve, meteorological factor output influence curve, historical electricity consumption data set, historical power generation data set, and the historical output data set, a historical output model is constructed. Step S4 of the embodiment of the present application further includes: Randomly extract data from the historical electricity consumption data set and the historical power generation data set multiple times, and perform mapping extraction on the historical output data set according to the extraction results. Combine the user output influence curve and the meteorological factor output influence curve to obtain a support set and a query set; Use the support set to perform supervised training on the framework constructed based on the feedforward neural network, and use the query set to verify the trained framework until convergence to obtain the trained historical output model.

[0038] In a possible embodiment, the historical output model is a model trained based on a feedforward neural network, which can predict the photovoltaic output under different electricity consumption behaviors and meteorological conditions. The support set and the query set are two data sets extracted from historical data. The support set is used to train the model, while the query set is used to verify the accuracy of the model.

[0039] Preferably, randomly extract data from historical electricity consumption data and historical power generation data multiple times, which helps to avoid data bias and ensure the wide representativeness of the data. Then, perform mapping extraction on the historical output data set based on the extracted data, that is, extract data from the corresponding historical output data set according to the extracted historical electricity consumption data and historical power generation data to obtain a training data set. Divide the training data set into a support set and a query set according to a ratio preset by those skilled in the art. Among them, the support set is used to train the model, and the query set is used to verify the model.

[0040] Furthermore, use the support set to perform supervised training on the feedforward neural network framework. During the training process, the neural network learns the laws in the historical data and continuously adjusts the network parameters to make the output result tend to the true value. Through multiple iterations, the output of the model gradually approaches the true photovoltaic output. Finally, use the query set to verify the trained model and check its performance on unknown data. When the training process converges, it means that the model has learned an effective mapping relationship, and the historical output model is also trained. The historical output model can accurately predict the output of the photovoltaic system under specific user behaviors and meteorological conditions, providing support for power grid dispatching and the optimization management of photovoltaic power generation.

[0041] S5: Obtain a sequence of real-time electricity consumption parameter sets and a sequence of real-time power generation parameter sets of the photovoltaic user set within a preset acquisition window, and construct a real-time output model; Further, each real-time power consumption parameter in the real-time power consumption parameter set sequence includes current, voltage, and power, and each real-time power generation parameter in the real-time power generation parameter set sequence includes power generation power and power peak value.

[0042] In a possible embodiment, the preset acquisition window is a time period preset by those skilled in the art for real-time data acquisition, usually a relatively small time period, such as 30s, 50s, etc. The real-time power consumption parameter set sequence is the power consumption data of photovoltaic users collected in real-time within the preset acquisition window, including electrical parameters such as current, voltage, and power, which can directly reflect the power demand of users at different time periods. The real-time power generation parameter set sequence is the power generation data of the photovoltaic system collected within the preset acquisition window, usually including parameters such as power generation power and power peak value. The real-time output model is a prediction model constructed based on real-time power consumption and power generation data, reflecting the output situation of the photovoltaic system.

[0043] Further, to obtain the real-time power consumption parameter set sequence and the real-time power generation parameter set sequence of the photovoltaic user set within the preset acquisition window and construct a real-time output model, step S5 of the embodiment of the present application further includes: Perform parameter centralized screening on the real-time power consumption parameter set sequence and the real-time power generation parameter set sequence respectively to obtain a centralized real-time power consumption parameter sequence and a centralized real-time power generation parameter sequence; Calculate the mapping difference between the centralized real-time power consumption parameter sequence and the centralized real-time power generation parameter sequence to obtain a centralized real-time output data sequence; Construct the real-time output model based on the centralized real-time power consumption parameter sequence, the centralized real-time power generation parameter sequence, and the centralized real-time output data sequence.

[0044] Further, to perform parameter centralized screening on the real-time power consumption parameter set sequence and the real-time power generation parameter set sequence respectively to obtain a centralized real-time power consumption parameter sequence and a centralized real-time power generation parameter sequence, step S5 of the embodiment of the present application further includes: Traverse the set means of the real-time power consumption parameter set sequence and the real-time power generation parameter set sequence to obtain a real-time power consumption parameter mean sequence and a real-time power generation parameter mean sequence; Respectively use the real-time power consumption parameter mean sequence and the real-time power generation parameter mean sequence as the centralized screening starting points, and use the parameter centralized screening function to update and iterate the centralized screening starting points until the preset update and iteration times are met to obtain a centralized real-time power consumption parameter sequence and a centralized real-time power generation parameter sequence.

[0045] Further, the parameter centralized screening function is: ; Wherein, To centrally screen iterative points, is the Gaussian weight kernel function, is the starting point of central screening, is the near-neighbor neighborhood composed of real-time electricity consumption parameters in the real-time electricity consumption parameter set whose distance to the starting point of central screening is less than the preset value, is the th real-time electricity consumption parameter in the near-neighbor neighborhood as the center, and the neighborhood density of the obtained neighborhood constructed with the preset value as the radius, is the th real-time electricity consumption parameter in the near-neighbor neighborhood, is the Gaussian function with weight decay, is the parameter controlling the decay rate.

[0046] In a possible embodiment, the centralized real-time electricity consumption parameter sequence is the real-time electricity consumption data sequence after parameter central screening, reflecting the electricity consumption demand situation of photovoltaic users after screening and processing. The centralized real-time power generation parameter sequence is the real-time power generation data sequence after parameter central screening, reflecting the power generation situation of the photovoltaic system after screening and processing.

[0047] By extracting the power in the centralized real-time electricity consumption parameter sequence and calculating the mapping difference with the power generation power in the centralized real-time power generation parameter sequence, the centralized real-time output data sequence is obtained. Among them, the centralized real-time output data sequence reflects the photovoltaic output situation of the photovoltaic user set in the target substation area within the preset acquisition window. Furthermore, with the centralized real-time electricity consumption parameter sequence and the centralized real-time power generation parameter sequence as input data and the centralized real-time output data sequence as output data, the real-time output model is constructed.

[0048] Preferably, the model framework is constructed based on a convolutional neural network. With the centralized real-time electricity consumption parameter sequence and the centralized real-time power generation parameter sequence as input data, the predicted real-time output data sequence output by the framework is obtained. The cosine similarity is used to calculate the similarity between the centralized real-time output data sequence and the predicted real-time output data sequence, and it is judged whether the calculation result meets the preset requirements. If so, the trained real-time output model is obtained. If not, the network parameters of the framework are updated, and the updated framework is trained again.

[0049] In a possible embodiment, the sets in the real-time power consumption parameter set sequence and the real-time power generation parameter set sequence are respectively subjected to mean value calculation to obtain a real-time power consumption parameter mean value sequence and a real-time power generation parameter mean value sequence. Furthermore, the real-time power consumption parameter mean value sequence and the real-time power generation parameter mean value sequence are respectively used as the centralized screening starting points, where the centralized screening starting point is the initial data central point in each set. Then, the parameter centralized screening function is used to update and iterate the centralized screening starting points until the preset update iteration times (the maximum update iteration times preset by those skilled in the art) are satisfied, and a centralized real-time power consumption parameter sequence and a centralized real-time power generation parameter sequence are obtained. Among them, the parameter centralized screening function is used to iteratively screen the data with the densest distribution in the set. Among them, the parameter centralized screening function is a parameter for controlling the attenuation speed, which is used to ensure that the influence of neighborhood points farther away on the target point is smaller.

[0050] S6: Optimize the historical output model by using the real-time output model to obtain a photovoltaic output prediction model.

[0051] In a possible embodiment, the photovoltaic output prediction model is used to reliably predict the photovoltaic output situation in the target substation area. By respectively inputting the real-time power consumption parameter set sequence and the real-time power generation parameter set sequence into the real-time output model and the historical output model, a real-time prediction output result set sequence and a historical prediction output result set sequence are obtained. Calculate the difference between the real-time prediction output result set sequence and the historical prediction output result set sequence to obtain a prediction error result sequence. Among them, the prediction error result sequence reflects the data deviation between the historical output model and the real-time output model. Calculate the mean value of the prediction error result sequence, and judge whether the mean value is less than the preset error mean value (the maximum error mean value preset by those skilled in the art). If not, adjust the parameters of the historical output model until the requirements are met. That is to say, when the error between the output results of the historical output model and the real-time output model is less than the preset error mean value, the historical output model after parameter adjustment is used as the photovoltaic output prediction model.

[0052] By optimizing the historical output model based on the real-time output model, it is achieved that the historical output model not only depends on historical data, but also can timely reflect the latest photovoltaic user power consumption behavior and power generation situation, avoiding the possible lag and error in the historical model. The optimized photovoltaic output prediction model can more accurately predict the actual power generation of the photovoltaic system and provide real-time and accurate data required for power grid dispatching, load management, etc. This optimized model has stronger adaptability and flexibility, and can effectively improve the prediction accuracy of photovoltaic power generation in practical applications, providing more reliable data support for the efficient operation of the power grid.

[0053] In summary, the embodiments of the present application at least have the following technical effects: 1. By optimizing the historical output model and combining real-time data feedback and error analysis, the present application can significantly improve the accuracy of photovoltaic output prediction. The optimized model can effectively reduce systematic errors and adjust the prediction results in real time according to multi-dimensional information such as user electricity consumption behavior and meteorological factors, ensuring that the power generation prediction of the photovoltaic system is more in line with the actual situation.

[0054] 2. By constructing a photovoltaic output prediction model by combining multi-source heterogeneous data (such as historical electricity consumption data, historical power generation data, meteorological data, and real-time power data), the present application achieves the technical effects of improving the accuracy and real-time performance of photovoltaic system output prediction.

[0055] Embodiment 2, based on the same inventive concept as the method for predicting the photovoltaic output of a distribution transformer area combined with sampling of grid heterogeneous data in the foregoing embodiment, as Figure 2 shown, the present application provides a system for predicting the photovoltaic output of a distribution transformer area combined with sampling of grid heterogeneous data. The system in the embodiments of the present application and the method embodiments are based on the same inventive concept. Among them, the system includes: A historical output data set construction module 11, configured to obtain a historical electricity consumption data set and a historical power generation data set of a photovoltaic user set in a target distribution transformer area within a historical window, perform output analysis, and construct a historical output data set; A user output influence curve construction module 12, configured to obtain a historical electricity consumption behavior data set of a photovoltaic user set within a historical window, combine the historical output data set for data analysis, and construct a user output influence curve; A meteorological factor output influence curve construction module 13, configured to obtain a meteorological data set within a historical window, combine the historical output data set for data analysis, and construct a meteorological factor output influence curve; A historical output model construction module 14, configured to construct a historical output model based on the user output influence curve, the meteorological factor output influence curve, the historical electricity consumption data set, the historical power generation data set, and the historical output data set; A real-time output model construction module 15, configured to obtain a sequence of real-time electricity consumption parameter sets and a sequence of real-time power generation parameter sets of a photovoltaic user set within a preset collection window, and construct a real-time output model; A photovoltaic output prediction model obtaining module 16, configured to optimize the historical output model by using the real-time output model to obtain a photovoltaic output prediction model.

[0056] Further, the system further includes: Calculating the mapped power difference of the historical electricity consumption data set and the historical power generation data set to obtain an initial historical output data set; Perform outlier screening on the initial historical output data set to obtain the initial historical output data set.

[0057] Furthermore, the system further includes: Construct a two-dimensional coordinate system with historical output data as the abscissa and historical electricity consumption behavior data as the ordinate; Input the historical electricity consumption behavior data set and the historical output data set into the two-dimensional coordinate system to obtain a first influence scatter point set; Fit the first influence scatter point set to obtain the user output influence curve.

[0058] Furthermore, the system further includes: Use a polynomial regression model to perform preliminary fitting on the first influence scatter point set to obtain a first fitting curve; According to a preset tolerance bandwidth, count the first fitting neighborhood of the first fitting curve, and count the number of scatter points in the first fitting neighborhood to obtain the number of scatter points in the first fitting neighborhood; Judge whether the number of scatter points above the first fitting curve is greater than the number of scatter points below the first fitting curve. If so, move the first fitting curve upward according to a preset moving step to obtain a second fitting curve; Construct a second fitting neighborhood of the second fitting curve according to a preset tolerance bandwidth, and count the number of scatter points in the second fitting neighborhood to obtain the number of scatter points in the second fitting neighborhood; When the number of scatter points in the second fitting neighborhood is greater than or equal to the number of scatter points in the first fitting neighborhood, continue to move the second fitting curve upward according to a preset moving step until a preset number of moves is satisfied to obtain the user output influence curve.

[0059] Furthermore, the system further includes: Randomly extract data multiple times from the historical electricity consumption data set and the historical power generation data set, and perform mapping extraction on the historical output data set according to the extraction results. Combine the user output influence curve and the meteorological factor output influence curve to obtain a support set and a query set; Use the support set to perform supervised training on a framework constructed based on a feedforward neural network, and use the query set to verify the trained framework until convergence to obtain the trained historical output model.

[0060] Furthermore, the system further includes: Perform parameter centralized screening on the real-time electricity consumption parameter set sequence and the real-time power generation parameter set sequence respectively to obtain a centralized real-time electricity consumption parameter sequence and a centralized real-time power generation parameter sequence; Calculate the mapping difference between the centralized real-time power consumption parameter sequence and the centralized real-time power generation parameter sequence to obtain the centralized real-time output data sequence; Construct the real-time output model based on the centralized real-time power consumption parameter sequence, the centralized real-time power generation parameter sequence, and the centralized real-time output data sequence.

[0061] Furthermore, the system further includes: Traverse the set means of the real-time power consumption parameter set sequence and the real-time power generation parameter set sequence to obtain the real-time power consumption parameter mean sequence and the real-time power generation parameter mean sequence; Respectively take the real-time power consumption parameter mean sequence and the real-time power generation parameter mean sequence as the centralized screening starting points, and use the parameter centralized screening function to update and iterate the centralized screening starting points until the preset update and iteration times are met to obtain the centralized real-time power consumption parameter sequence and the centralized real-time power generation parameter sequence.

[0062] Furthermore, the parameter centralized screening function is: ; Wherein, is the centralized screening iteration point, is the Gaussian weight kernel function, is the centralized screening starting point, is the near neighbor neighborhood composed of real-time power consumption parameters in the real-time power consumption parameter set whose distance to the centralized screening starting point is less than the preset value, is the neighborhood density of the neighborhood obtained by constructing a neighborhood with the th real-time power consumption parameter in the near neighbor neighborhood as the center and the preset value as the radius, is the th real-time power consumption parameter in the near neighbor neighborhood, is the Gaussian function with weight decay, is the parameter controlling the decay speed.

[0063] Furthermore, each real-time power consumption parameter in the real-time power consumption parameter set sequence includes current, voltage, and power, and each real-time power generation parameter in each real-time power generation parameter set sequence in the real-time power generation parameter set sequence includes power generation power and power peak.

[0064] It should be noted that the above sequence of embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above describes specific embodiments of the present specification. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0065] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

[0066] This specification and the drawings are merely exemplary descriptions of the present application and are considered to have covered any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, provided that these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications therein.

Claims

1. A method for predicting the photovoltaic output of a distribution transformer area combined with heterogeneous data sampling of the power grid, characterized in that The method includes: Obtaining the historical power consumption data set and historical power generation data set of the photovoltaic user set in the target substation area within the historical window, performing output analysis, and constructing a historical output data set; Obtaining the historical electricity consumption behavior data set of the photovoltaic user set within the historical window, combining with the historical output data set for data analysis, and constructing a user output influence curve; Obtaining the meteorological data set within the historical window, combining with the historical output data set for data analysis, and constructing a meteorological factor output influence curve; Constructing a historical output model based on the user output influence curve, meteorological factor output influence curve, historical power consumption data set, historical power generation data set, and the historical output data set; Obtaining the real-time power consumption parameter set sequence and real-time power generation parameter set sequence of the photovoltaic user set within the preset acquisition window, and constructing a real-time output model; Using the real-time output model to optimize the historical output model to obtain a photovoltaic output prediction model.

2. The method for predicting the photovoltaic output of a transformer substation area combined with the sampling of heterogeneous power grid data according to claim 1, wherein, Obtaining the historical power consumption data set and historical power generation data set of the photovoltaic user set in the target substation area within the historical window, performing output analysis, and constructing a historical output data set, including: Calculating the mapped power difference of the historical power consumption data set and historical power generation data set to obtain an initial historical output data set; Performing outlier screening on the initial historical output data set to obtain the initial historical output data set.

3. The method for predicting the photovoltaic output of a transformer substation area by combining the sampling of heterogeneous power grid data according to claim 1, wherein, Obtaining the historical electricity consumption behavior data set of the photovoltaic user set within the historical window, combining with the historical output data set for data analysis, and constructing a user output influence curve, including: Constructing a two-dimensional coordinate system with historical output data as the abscissa and historical electricity consumption behavior data as the ordinate; Inputting the historical electricity consumption behavior data set and the historical output data set into the two-dimensional coordinate system to obtain a first influence scatter point set; Fitting the first influence scatter point set to obtain the user output influence curve.

4. The method for predicting the photovoltaic output of a distribution transformer area by combining heterogeneous data sampling of a power grid according to claim 3, wherein Fitting the first influence scatter point set to obtain the user output influence curve, including: Using a polynomial regression model to preliminarily fit the first influence scatter point set to obtain a first fitting curve; Counting the first fitting neighborhood of the first fitting curve according to a preset tolerance bandwidth, and counting the number of scatter points within the first fitting neighborhood to obtain the first fitting neighborhood scatter point quantity; Judging whether the number of scatter points above the first fitting curve is greater than the number of scatter points below the first fitting curve. If so, moving the first fitting curve upward by a preset moving step to obtain a second fitting curve; Constructing a second fitting neighborhood of the second fitting curve according to a preset tolerance bandwidth, and counting the number of scatter points within the second fitting neighborhood to obtain the second fitting neighborhood scatter point quantity; When the second fitting neighborhood scatter point quantity is greater than or equal to the first fitting neighborhood scatter point quantity, continue to move the second fitting curve upward by a preset moving step until a preset number of moves is satisfied to obtain the user output influence curve.

5. The method for predicting the photovoltaic output of a transformer substation area combined with heterogeneous data sampling of a power grid according to claim 1, wherein Constructing a historical output model based on the user output influence curve, meteorological factor output influence curve, historical power consumption data set, historical power generation data set, and the historical output data set, including: Perform multiple random data extractions from the historical power consumption data set and the historical power generation data set, and perform mapping extraction on the historical output data set according to the extraction results. Combine the user output influence curve and the meteorological factor output influence curve to obtain a support set and a query set; Use the support set to perform supervised training on the framework constructed based on the feedforward neural network, and use the query set to verify the trained framework until convergence to obtain the trained historical output model.

6. The method for predicting the photovoltaic output of a transformer substation area by combining the sampling of heterogeneous power grid data according to claim 1, characterized in that Obtain the real-time power consumption parameter set sequence and the real-time power generation parameter set sequence of the photovoltaic user set within a preset collection window, and construct a real-time output model, including: Perform parameter centralized screening on the real-time power consumption parameter set sequence and the real-time power generation parameter set sequence respectively to obtain a centralized real-time power consumption parameter sequence and a centralized real-time power generation parameter sequence; Calculate the mapping difference between the centralized real-time power consumption parameter sequence and the centralized real-time power generation parameter sequence to obtain a centralized real-time output data sequence; Construct the real-time output model based on the centralized real-time power consumption parameter sequence, the centralized real-time power generation parameter sequence, and the centralized real-time output data sequence.

7. The method for predicting the photovoltaic output of a distribution transformer area combined with the sampling of heterogeneous power grid data according to claim 6, wherein Perform parameter centralized screening on the real-time power consumption parameter set sequence and the real-time power generation parameter set sequence respectively to obtain a centralized real-time power consumption parameter sequence and a centralized real-time power generation parameter sequence, including: Traverse the set means of the real-time power consumption parameter set sequence and the real-time power generation parameter set sequence to obtain a real-time power consumption parameter mean sequence and a real-time power generation parameter mean sequence; Respectively use the real-time power consumption parameter mean sequence and the real-time power generation parameter mean sequence as the centralized screening starting points, and use the parameter centralized screening function to update and iterate the centralized screening starting points until the preset update iteration times are met to obtain a centralized real-time power consumption parameter sequence and a centralized real-time power generation parameter sequence.

8. The method for predicting the photovoltaic output of a substation area by combining the heterogeneous data sampling of the power grid according to claim 7, characterized in that, The parameter centralized screening function is: ； Among them, is for centralized screening of iteration points, is the Gaussian weight kernel function, is the starting point of centralized screening, is the near-neighbor neighborhood composed of real-time electricity consumption parameters in the real-time electricity consumption parameter set whose distance to the starting point of centralized screening is less than the preset value, is centered on the th real-time electricity consumption parameter in the near-neighbor neighborhood, and is the neighborhood density of the neighborhood obtained by constructing a neighborhood with the preset value as the radius, is the th real-time electricity consumption parameter in the near-neighbor neighborhood, is the Gaussian function with weight decay, is the parameter for controlling the decay speed.

9. The method for predicting the photovoltaic output of a transformer substation area by combining the sampling of heterogeneous power grid data according to claim 1, characterized in that, Each real-time power consumption parameter in the real-time power consumption parameter set sequence includes current, voltage, and power, and each real-time power generation parameter in the real-time power generation parameter set sequence includes power generation power and power peak.

10. A photovoltaic output prediction system for a distribution transformer area combined with heterogeneous data sampling of the power grid, characterized in that, The system includes: A historical output data set construction module, which is used to obtain the historical power consumption data set and the historical power generation data set of the photovoltaic user set in the target substation area within the historical window, perform output analysis, and construct a historical output data set; A user output influence curve construction module, which is used to obtain the historical power consumption behavior data set of the photovoltaic user set within the historical window, and perform data analysis in combination with the historical output data set to construct a user output influence curve; A meteorological factor output influence curve construction module, which is used to obtain the meteorological data set within the historical window, and perform data analysis in combination with the historical output data set to construct a meteorological factor output influence curve; A historical output model construction module, which is used to construct a historical output model based on the user output influence curve, the meteorological factor output influence curve, the historical power consumption data set, the historical power generation data set, and the historical output data set; A real-time output model construction module, which is used to obtain a sequence of real-time power consumption parameter sets and a sequence of real-time power generation parameter sets of a photovoltaic user set within a preset acquisition window, and construct a real-time output model; A photovoltaic output prediction model acquisition module, which is used to optimize the historical output model by using the real-time output model to obtain a photovoltaic output prediction model.

Images