Day-ahead power prediction method for photovoltaic power plants

Abstract

The present invention discloses a method for predicting the day-ahead power of a photovoltaic power station, including: obtaining current environmental parameters that affect the power of a photovoltaic power station, and constructing a sequence distribution curve for the current environmental parameters according to a time series; the current environmental parameters are not environmental parameters at the current moment, but environmental parameters within the current time period, that is, the environmental parameters within the set time before the prediction are set as the current environmental parameters. When it is determined according to the sequence distribution curve that the rate of change of any parameter in the environmental parameters is greater than or equal to a set first threshold, a fluctuation factor is created according to the rate of change of all parameters in the environmental parameters; a corresponding prediction model is found from a fluctuation factor database according to the fluctuation factor; the current environmental parameters are input into the found prediction model, and a prediction result is output. The prediction result is the predicted value of the day-ahead power under the current environmental parameters. The prediction result can be obtained more accurately through this solution.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power stations, and in particular to a method for predicting the day-ahead power of a photovoltaic power station. Background Art

[0002] As the installed capacity of photovoltaic power generation systems expands, their impact on the power system becomes more and more obvious. The power generation forecast of photovoltaic arrays plays an important role in reducing the impact of the uncertainty of photovoltaic power generation on the operation of the power grid. The accuracy of the forecast results is directly related to the economy and safety of the power system operation. From the perspective of stable operation of the power system, accurate power forecasting allows the power grid department to accurately dispatch based on the predicted power, reduce the impact of large-scale photovoltaic grid connection on the power grid, improve the utilization efficiency of new energy power generation, and cooperate with the power grid department to achieve maximum intermittent energy consumption. From the perspective of power grid dispatching and operation, accurate power forecasting is conducive to the regional power grid to formulate power generation plans, real-time power regulation and other decisions to provide scientific data reference, and ensure the overall operation efficiency of the power station. Therefore, improving the accuracy of photovoltaic power generation prediction has always been a technical problem to be solved in this field at this stage. Summary of the invention

[0003] The present invention provides a method for predicting the day-ahead power of a photovoltaic power station to solve the above-mentioned problems existing in the prior art.

[0004] The present invention provides a method for predicting the day-ahead power of a photovoltaic power station, the method comprising:

[0005] S100, obtaining current environmental parameters that affect the power of the photovoltaic power station, and constructing a sequence distribution curve for the current environmental parameters according to a time series;

[0006] S200, when it is determined according to the sequence distribution curve that the change rate of any parameter in the environmental parameters is greater than or equal to a set first threshold, a fluctuation factor is created according to the change rates of all parameters in the environmental parameters;

[0007] S300, searching a corresponding prediction model from a fluctuation factor database according to the fluctuation factor; the prediction model predicts the power according to the relationship between the current environmental parameters and the output power;

[0008] S400, inputting the current environmental parameters into the found prediction model and outputting the prediction result.

[0009] Preferably, the S100 includes:

[0010] S101, current environmental parameters include: solar radiation parameters, current ambient temperature, current ambient humidity and current wind speed;

[0011] S102, setting the horizontal axis of the sequence distribution curve to the monitoring time points, the monitoring time points are evenly spaced, and the vertical axis of the sequence distribution curve is the monitoring value of each current environmental parameter at the corresponding monitoring time point;

[0012] S103, forming a sequence distribution curve of the solar radiation parameter, the current ambient temperature, the current ambient humidity and the current wind speed according to the time series.

[0013] Preferably, the S200 includes:

[0014] S201, calculating the slope of each monitoring time point on each sequence distribution curve, obtaining all slopes on each sequence distribution curve, and constructing a current environment parameter-slope set; the slope is the rate of change of the parameter; the current environment parameter-slope set is the set of slopes of all points on the sequence distribution curve corresponding to each current environment parameter;

[0015] S202, determining whether the slope in each current environment parameter-slope set is greater than or equal to a set first threshold; each current environment parameter-slope set is set with a different first threshold;

[0016] S203, screening out all slopes that are greater than or equal to a set first threshold, and current environmental parameters corresponding to the screened slopes;

[0017] S204: Determine a fluctuation factor according to the screened slope and corresponding parameters.

[0018] Preferably, the S204 includes:

[0019] S2041, setting a first weight value of all parameters in the current environment parameters, wherein the relationship between the first weight value, the parameter and the output power is determined;

[0020] S2042, performing a weighted sum calculation on the screened slope and the corresponding first weight value to obtain a weighted sum value of all the screened parameters of the current environmental parameters;

[0021] S2043: Parameters whose weighted sum values ​​are greater than or equal to the sum threshold are used as fluctuation factors.

[0022] Preferably, the S300 includes:

[0023] S301, constructing a fluctuation factor database, recording a prediction model corresponding to each fluctuation factor in the fluctuation factor database, the prediction model including: a solar radiation parameter-prediction model, a current ambient temperature-prediction model, a current ambient humidity-prediction model, and a current wind speed-prediction model;

[0024] S302, determining whether the number of fluctuation factors is greater than or equal to 1, if greater than or equal to 1, finding greater than or equal to 1 prediction models to form a prediction model set.

[0025] Preferably, the S400 includes:

[0026] S401, setting a second weight value for each prediction model in the prediction model set;

[0027] S402, inputting the current environmental parameters into all prediction models in the prediction model set respectively, and obtaining multiple prediction results through different prediction models;

[0028] S403, corresponding to the second weight values ​​of different prediction models and the corresponding prediction results, calculate and determine the final prediction result.

[0029] Preferably, the S400 further includes: constructing a prediction model; and the method for constructing a prediction model includes:

[0030] S404, obtaining historical big data, the historical big data including: solar radiation parameters, current ambient temperature, current ambient humidity and current wind speed, and the relationship between the solar radiation parameters, current ambient temperature, current ambient humidity and current wind speed and the output power respectively;

[0031] S405, according to the solar radiation parameter, the current ambient temperature, the current ambient humidity and the current wind speed having large fluctuation data as a classification principle, different data sets with the solar radiation parameter, the current ambient temperature, the current ambient humidity and the current wind speed as large fluctuation factors are classified respectively;

[0032] S406, for different data sets, training is performed using the relationship between solar irradiation parameters, current ambient temperature, current ambient humidity and current wind speed and output power, to obtain different prediction models using solar irradiation parameters, current ambient temperature, current ambient humidity and current wind speed as fluctuation factors.

[0033] Preferably, the S406 includes:

[0034] S4061, with the monitoring time point as the horizontal coordinate, and the monitoring value of each solar radiation parameter, current ambient temperature, current ambient humidity and current wind speed at the corresponding monitoring time point as the vertical coordinate, four sequence distribution curves with solar radiation parameter, current ambient temperature, current ambient humidity and current wind speed as the vertical coordinate are formed respectively; the horizontal coordinates are equally spaced;

[0035] S4062, for a data set with different fluctuation factors, using the monitoring values ​​of the parameters corresponding to the fluctuation factors as the basis for change, distributing the monitoring values ​​in equal intervals, converting the sequence distribution curve into a conversion distribution curve with a constant speed change of the monitoring values, and the abscissa of the conversion distribution curve is distributed in unequal intervals;

[0036] S4063, the conversion distribution curve and the corresponding output power are used as output parameters and input into the recurrent neural network RNN ​​for training to obtain a prediction result.

[0037] Preferably, the step S100 further includes:

[0038] After obtaining the current environmental parameters, the abnormal data of the current environmental parameters are screened to filter out the abnormal data;

[0039] Replace the abnormal data and add the replaced data to the original data position to form complete data, and construct a sequence distribution curve with the complete data.

[0040] Preferably, the screening out of abnormal data includes: if the data in a certain time period is abnormal compared with the change pattern of the day, but the data before and after the current time are periodic, then the data is non-abnormal data;

[0041] The replacing operation on the abnormal data to add the replaced data to the original data position includes:

[0042] The data before the abnormal data is set as the front data, and the data after the abnormal data is set as the back data;

[0043] Based on the previous data and the next data, the data is input into the data filling model, and the replacement data of the abnormal data is output;

[0044] Supplement the replacement data corresponding to the monitoring time point to the original data position;

[0045] The data filling model is obtained by training a deep neural network.

[0046] Compared with the prior art, the present invention has the following advantages:

[0047] The present invention provides a method for predicting the day-ahead power of a photovoltaic power station, including: obtaining current environmental parameters that affect the power of a photovoltaic power station, and constructing a sequence distribution curve for the current environmental parameters according to a time series; the current environmental parameters are not environmental parameters at the current moment, but environmental parameters within the current time period, that is, the environmental parameters within the set time before the prediction are set as the current environmental parameters. When it is determined according to the sequence distribution curve that the rate of change of any parameter in the environmental parameters is greater than or equal to a set first threshold, a fluctuation factor is created according to the rate of change of all parameters in the environmental parameters; a corresponding prediction model is searched from a fluctuation factor database according to the fluctuation factor; the prediction model predicts the power through the relationship between the current environmental parameters and the output power; the current environmental parameters are input into the searched prediction model, and a prediction result is output. The prediction result is the predicted value of the day-ahead power under the current environmental parameters. The prediction result can be obtained more accurately through this solution.

[0048] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings.

[0049] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0051] Figure 1 Flow chart of a method for predicting day-ahead power of a photovoltaic power station according to an embodiment of the present invention;

[0052] Figure 2 A flow chart of a method for creating a fluctuation factor according to the change rates of all parameters in the environmental parameters in an embodiment of the present invention;

[0053] Figure 3 The following is a flow chart of a method for constructing a prediction model in an embodiment of the present invention. DETAILED DESCRIPTION

[0054] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0055] The present invention provides a method for predicting the day-ahead power of a photovoltaic power station. Figure 1 , the photovoltaic power station day-ahead power prediction method includes:

[0056] S100, obtaining current environmental parameters that affect the power of the photovoltaic power station, and constructing a sequence distribution curve for the current environmental parameters according to a time series;

[0057] S200, when it is determined according to the sequence distribution curve that the change rate of any parameter in the environmental parameters is greater than or equal to a set first threshold, a fluctuation factor is created according to the change rates of all parameters in the environmental parameters;

[0058] S300, searching a corresponding prediction model from a fluctuation factor database according to the fluctuation factor; the prediction model predicts the power according to the relationship between the current environmental parameters and the output power;

[0059] S400, inputting the current environmental parameters into the found prediction model and outputting the prediction result.

[0060] The working principle and beneficial effects of the above technical solution are as follows: the solution adopted in this embodiment is to obtain the current environmental parameters that affect the power of the photovoltaic power station, and to construct a sequence distribution curve for the current environmental parameters according to the time series; the current environmental parameters are not the environmental parameters at the current moment, but the environmental parameters within the current time period, that is, the environmental parameters within the set time before the prediction are set as the current environmental parameters. When it is judged according to the sequence distribution curve that the change rate of any parameter in the environmental parameters is greater than or equal to the set first threshold, a fluctuation factor is created according to the change rate of all parameters in the environmental parameters; the corresponding prediction model is found from the fluctuation factor database according to the fluctuation factor; the prediction model predicts the power through the relationship between the current environmental parameters and the output power; the current environmental parameters are input into the found prediction model, and the prediction result is output. The prediction result is the predicted value of the day-ahead power under the current environmental parameters.

[0061] In the prior art, fluctuations in different factors will cause deviations in different prediction results. For example, large fluctuations in solar irradiation parameters and wind speed may result in different prediction results for the final output power. Using the same prediction model to make unified predictions for fluctuations in multiple parameters will result in inaccurate prediction results.

[0062] Therefore, the solution adopted in this embodiment is to first screen out the volatility factors with large fluctuations, and select the corresponding prediction model according to the parameters with large volatility factors. The selected prediction model is trained mainly for the volatility factor during training. Therefore, in the scenario where other parameters change normally and only the parameter corresponding to the volatility factor has a large range of fluctuations, the solution provided by this embodiment can obtain more accurate prediction results. If the fluctuations of all parameters are not large and do not exceed the set threshold, conventional prediction methods can be used.

[0063] In summary, the solution of this embodiment is used to select a targeted and unique prediction model for the above scenario, ensuring that there is a strong correlation between the prediction model and the actual parameter changes, thereby ensuring the accuracy of the prediction results of the prediction model. This can guide the operation of photovoltaic power stations and formulate multi-power coordination scheduling plans; help the scheduling department predict photovoltaic fluctuations and reasonably formulate decisions that comply with the power market, etc., meeting the requirements for prediction accuracy under the above conditions.

[0064] In another embodiment, the S100 includes:

[0065] S101, current environmental parameters include: solar radiation parameters, current ambient temperature, current ambient humidity and current wind speed;

[0066] S102, setting the horizontal axis of the sequence distribution curve to the monitoring time points, the monitoring time points are evenly spaced, and the vertical axis of the sequence distribution curve is the monitoring value of each current environmental parameter at the corresponding monitoring time point;

[0067] S103, forming a sequence distribution curve of the solar radiation parameter, the current ambient temperature, the current ambient humidity and the current wind speed according to the time series.

[0068] The working principle and beneficial effects of the above technical solution are as follows: the solution adopted in this embodiment is to obtain the current environmental parameters that affect the power of the photovoltaic power station, and to construct a specific implementation method of a sequence distribution curve for the current environmental parameters according to a time series. The scheme adopted includes: the current environmental parameters include: solar radiation parameters, current ambient temperature, current ambient humidity and current wind speed; the horizontal axis of the sequence distribution curve is set to the monitoring time point, the monitoring time points are evenly spaced, and the vertical axis of the sequence distribution curve is the monitoring value of each current environmental parameter at the corresponding monitoring time point; the solar radiation parameters, the current ambient temperature, the current ambient humidity and the current wind speed are respectively formed into a sequence distribution curve according to the time series.

[0069] The sequence distribution curve is constructed according to the monitored time nodes, which is convenient for further obtaining the distribution curve of each parameter, forming the solar radiation parameter-sequence distribution curve, the current ambient temperature-sequence distribution curve, the current ambient humidity-sequence distribution curve and the current wind speed-sequence distribution curve respectively, so as to facilitate the screening of the fluctuation factors with larger fluctuations according to different distribution curves, and then screen the corresponding prediction model.

[0070] In another embodiment, please refer to Figure 2 , the S200 includes:

[0071] S201, calculating the slope of each monitoring time point on each sequence distribution curve, obtaining all slopes on each sequence distribution curve, and constructing a current environment parameter-slope set; the slope is the rate of change of the parameter; the current environment parameter-slope set is the set of slopes of all points on the sequence distribution curve corresponding to each current environment parameter;

[0072] S202, determining whether the slope in each current environment parameter-slope set is greater than or equal to a set first threshold; each current environment parameter-slope set is set with a different first threshold;

[0073] S203, screening out all slopes that are greater than or equal to a set first threshold, and current environmental parameters corresponding to the screened slopes;

[0074] S204: Determine a fluctuation factor according to the screened slope and corresponding parameters.

[0075] The working principle and beneficial effects of the above technical solution are as follows: the solution adopted in this embodiment is the process of building a fluctuation factor, and the specific method adopted is as follows: first, calculate the slope of each monitoring time point on each sequence distribution curve, obtain all slopes on each sequence distribution curve, and construct the current environmental parameter-slope set; the slope is the rate of change of the parameter; the current environmental parameter-slope set is the set of slopes of all points on the sequence distribution curve corresponding to each current environmental parameter; secondly, determine whether the slope in each current environmental parameter-slope set is greater than or equal to the set first threshold; each current environmental parameter-slope set is set with a different first threshold; then, filter out all slopes that are greater than or equal to the set first threshold, and the current environmental parameter corresponding to the filtered slope; finally, determine the fluctuation factor based on the filtered slope and the corresponding parameter. According to the solution provided in this embodiment, the rate of change can be calculated by the slope of the curve, and then the parameters with large fluctuations can be filtered out.

[0076] In another embodiment, the S204 includes:

[0077] S2041, setting a first weight value of all parameters in the current environment parameters, wherein the relationship between the first weight value, the parameter and the output power is determined;

[0078] S2042, performing a weighted sum calculation on the screened slope and the corresponding first weight value to obtain a weighted sum value of all the screened parameters of the current environmental parameters;

[0079] S2043: Parameters whose weighted sum values ​​are greater than or equal to the sum threshold are used as fluctuation factors.

[0080] The working principle and beneficial effects of the above technical solution are as follows: the solution adopted in this embodiment is a specific implementation method for determining the fluctuation factor based on the screened slope and the corresponding parameters, including: setting the first weight value of all parameters in the current environmental parameters, the first weight value is determined by the relationship between the parameters and the output power; performing weighted sum calculation on the screened slope and the corresponding first weight value to obtain the weighted sum value of all parameters of the screened current environmental parameters; taking the parameter whose weighted sum value is greater than or equal to the sum threshold as the fluctuation factor. The solution provided in this embodiment ensures that different parameters correspond to different first weight values ​​by setting the first weight value, distinguishes the way in which all parameters are set according to the same rule, and ensures that the relationship between different parameters and the output power is different by setting the weight value, further improving the accuracy of the prediction result.

[0081] In another embodiment, the S300 includes:

[0082] S301, constructing a fluctuation factor database, recording a prediction model corresponding to each fluctuation factor in the fluctuation factor database, the prediction model including: a solar radiation parameter-prediction model, a current ambient temperature-prediction model, a current ambient humidity-prediction model, and a current wind speed-prediction model;

[0083] S302, determining whether the number of fluctuation factors is greater than or equal to 1, if greater than or equal to 1, finding greater than or equal to 1 prediction models to form a prediction model set.

[0084] The working principle and beneficial effects of the above technical solution are as follows: the solution adopted in this embodiment is the process of constructing a fluctuation factor database, and the specific construction method is as follows: construct a fluctuation factor database, record the prediction model corresponding to each fluctuation factor in the fluctuation factor database, and the prediction model includes: solar irradiation parameter-prediction model, current ambient temperature-prediction model, current ambient humidity-prediction model and current wind speed-prediction model; judge whether the number of fluctuation factors is greater than or equal to 1, if greater than or equal to 1, find out greater than or equal to 1 prediction model to form a prediction model set. The solution provided in this embodiment has recorded the prediction models corresponding to the fluctuation factors in the fluctuation factor database, which are solar irradiation parameter-prediction model, current ambient temperature-prediction model, current ambient humidity-prediction model and current wind speed-prediction model, so for the situation where the solar irradiation parameter fluctuates greatly, it can be predicted by the solar irradiation parameter-prediction model. Therefore, this solution can adjust the prediction model in real time according to the current parameters, ensure the matching degree between the prediction model and the current parameters, and then ensure the accuracy of the prediction results. In addition, according to the actual situation, if multiple fluctuation factors are selected and multiple prediction models are corresponding, the current environmental parameters can be predicted by multiple prediction models, and the final prediction results can be obtained by comprehensive prediction results.

[0085] In another embodiment, the S400 includes:

[0086] S401, setting a second weight value for each prediction model in the prediction model set;

[0087] S402, inputting the current environmental parameters into all prediction models in the prediction model set respectively, and obtaining multiple prediction results through different prediction models;

[0088] S403, corresponding to the second weight values ​​of different prediction models and the corresponding prediction results, calculate and determine the final prediction result.

[0089] The working principle and beneficial effects of the above technical solution are as follows: the solution adopted in this embodiment is based on the previous embodiment, and the method of obtaining the prediction result includes: setting a second weight value for each prediction model in the prediction model set; inputting the current environmental parameters into all the prediction models in the prediction model set respectively, and obtaining multiple prediction results through different prediction models; corresponding to the second weight values ​​of different prediction models and the corresponding prediction results, the final prediction result is calculated and determined. Different second weight values ​​are determined according to the relationship between different parameters and output power, and the final prediction result is obtained according to the comprehensive calculation of the weight value and the prediction result. The comprehensive prediction method of multiple prediction models is used to ensure that each factor affecting the output power can play a full role, and the accuracy of the prediction result is further ensured through the comprehensive prediction method.

[0090] In another embodiment, please refer to Figure 3 , the S400 further includes: constructing a prediction model; the method for constructing a prediction model includes:

[0091] S404, obtaining historical big data, the historical big data including: solar radiation parameters, current ambient temperature, current ambient humidity and current wind speed, and the relationship between the solar radiation parameters, current ambient temperature, current ambient humidity and current wind speed and the output power respectively;

[0092] S405, according to the solar radiation parameter, the current ambient temperature, the current ambient humidity and the current wind speed having large fluctuation data as a classification principle, different data sets with the solar radiation parameter, the current ambient temperature, the current ambient humidity and the current wind speed as large fluctuation factors are classified respectively;

[0093] S406, for different data sets, training is performed using the relationship between solar irradiation parameters, current ambient temperature, current ambient humidity and current wind speed and output power, to obtain different prediction models using solar irradiation parameters, current ambient temperature, current ambient humidity and current wind speed as fluctuation factors.

[0094] The working principle and beneficial effects of the above technical scheme are as follows: the scheme adopted in this embodiment is to construct a prediction model; the method of constructing the prediction model includes: obtaining historical big data, the historical big data including: solar irradiation parameters, current ambient temperature, current ambient humidity and current wind speed, and the relationship between solar irradiation parameters, current ambient temperature, current ambient humidity and current wind speed and output power respectively; according to the classification principle that solar irradiation parameters, current ambient temperature, current ambient humidity and current wind speed have large fluctuation data, different data sets with solar irradiation parameters, current ambient temperature, current ambient humidity and current wind speed as large fluctuation factors are classified respectively; for different data sets, the relationship between solar irradiation parameters, current ambient temperature, current ambient humidity and current wind speed and output power is used for training, and different prediction models with solar irradiation parameters, current ambient temperature, current ambient humidity and current wind speed as fluctuation factors are obtained respectively. Different prediction models are trained mainly for the corresponding fluctuation factors. Therefore, in the scenario where other parameters change normally and only the parameters corresponding to the fluctuation factors change widely, a more accurate prediction model can be constructed through the solution provided by this embodiment. The training data sets of the prediction model are all trained for the parameters with large fluctuations, so the prediction results can be obtained more accurately. If the fluctuations of all parameters are not large and do not exceed the set threshold, conventional prediction methods can be used.

[0095] In another embodiment, the S406 includes:

[0096] S4061, with the monitoring time point as the horizontal coordinate, and the monitoring value of each solar radiation parameter, current ambient temperature, current ambient humidity and current wind speed at the corresponding monitoring time point as the vertical coordinate, four sequence distribution curves with solar radiation parameter, current ambient temperature, current ambient humidity and current wind speed as the vertical coordinate are formed respectively; the horizontal coordinates are equally spaced;

[0097] S4062, for a data set with different fluctuation factors, using the monitoring values ​​of the parameters corresponding to the fluctuation factors as the basis for change, distributing the monitoring values ​​in equal intervals, converting the sequence distribution curve into a conversion distribution curve with a constant speed change of the monitoring values, and the abscissa of the conversion distribution curve is distributed in unequal intervals;

[0098] S4063, the conversion distribution curve and the corresponding output power are used as output parameters and input into the recurrent neural network RNN ​​for training to obtain a prediction result.

[0099] The working principle and beneficial effects of the above technical solution are as follows: the solution adopted in this embodiment is to use the monitoring time point as the horizontal coordinate, and the monitoring value of each solar irradiation parameter, current ambient temperature, current ambient humidity and current wind speed at the corresponding monitoring time point as the vertical coordinate, respectively forming four sequence distribution curves with solar irradiation parameter, current ambient temperature, current ambient humidity and current wind speed as the vertical coordinate; the horizontal coordinate is equally spaced distribution; for data sets with different fluctuation factors, the monitoring values ​​of the parameters corresponding to the fluctuation factors are used as the basis for change, the monitoring values ​​are equally spaced, and the sequence distribution curve is converted into a conversion distribution curve with a constant speed change of the monitoring values, and the horizontal coordinate of the conversion distribution curve is unequally spaced distribution; the conversion distribution curve and the corresponding output power are used as output parameters, input into the recurrent neural network RNN ​​for training, and obtain the prediction result. This embodiment provides a change in the distribution curve, distributes the monitoring values ​​equally spaced, so that the change law of the fluctuation is more accurate, and uses the converted distribution curve as the input of the training model to obtain a prediction model that is more in line with the corresponding scenario.

[0100] In another embodiment, the step S100 further includes:

[0101] After obtaining the current environmental parameters, the abnormal data of the current environmental parameters are screened to filter out the abnormal data;

[0102] Replace the abnormal data and add the replaced data to the original data position to form complete data, and construct a sequence distribution curve with the complete data.

[0103] The working principle and beneficial effects of the above technical solution are as follows: the solution adopted in this embodiment is to filter abnormal data of the current environmental parameters after obtaining the current environmental parameters, filter out abnormal data; replace the abnormal data, add the replaced data to the original data position, form complete data, and construct a sequence distribution curve with the complete data. By filtering and replacing the abnormal data, the integrity of the data can be guaranteed compared to directly deleting the abnormal data. Therefore, using complete data prediction can further improve the accuracy of the prediction results.

[0104] In another embodiment, the filtering out of abnormal data includes: if the data in a certain time period is abnormal compared with the change pattern of the day, but the data before and after the current time are periodic, then the data is non-abnormal data;

[0105] The replacing operation on the abnormal data to add the replaced data to the original data position includes:

[0106] The data before the abnormal data is set as the front data, and the data after the abnormal data is set as the back data;

[0107] Based on the previous data and the next data, the data is input into the data filling model, and the replacement data of the abnormal data is output;

[0108] Supplement the replacement data corresponding to the monitoring time point to the original data position;

[0109] The data filling model is obtained by training a deep neural network.

[0110] The working principle and beneficial effects of the above technical solution are as follows: the scheme adopted in this embodiment is the screening of abnormal data, including: if the data in a certain time period is abnormal compared with the change law of the day, but the data before and after the current time are periodic, then the data is non-abnormal data; the replacement operation of the abnormal data is performed, and the replacement data is added to the original data position, including: the data before the abnormal data is set as the front data, and the data after the abnormal data is set as the back data; based on the front data and the back data, the data filling model is input, and the replacement data of the abnormal data is output; the replacement data is added to the original data position corresponding to the monitoring time point; the data filling model is obtained by training a deep neural network. By screening and replacing the abnormal data, the integrity of the data can be guaranteed compared to directly deleting the abnormal data. Therefore, the accuracy of the prediction result can be further improved by using the complete data prediction. And by training the data filling model, the data at the abnormal data position can be predicted. Compared with the data obtained by averaging the two points before and after the abnormal data, the predicted data obtained by the scheme provided in this embodiment is more in line with the law of data change, and the predicted data is used as the replacement data to form complete data, and then the complete data is predicted by the prediction model to obtain the prediction result.

[0111] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. Day-ahead power prediction method for photovoltaic power plants, It is characterized in that include: S100, obtaining current environmental parameters that affect the power of the photovoltaic power station, and constructing a sequence distribution curve for the current environmental parameters according to a time series; S200, when it is determined according to the sequence distribution curve that the change rate of any parameter in the environmental parameters is greater than or equal to a set first threshold, a fluctuation factor is created according to the change rates of all parameters in the environmental parameters; S300, searching a corresponding prediction model from a fluctuation factor database according to the fluctuation factor; the prediction model predicts the power according to the relationship between the current environmental parameters and the output power; S400, inputting the current environmental parameters into the found prediction model and outputting the prediction result; The S100 includes: S101, current environmental parameters include: solar radiation parameters, current ambient temperature, current ambient humidity and current wind speed; S102, setting the horizontal axis of the sequence distribution curve to the monitoring time points, the monitoring time points are evenly spaced, and the vertical axis of the sequence distribution curve to the monitoring value of each current environmental parameter at the corresponding monitoring time point; S103, forming a sequence distribution curve of the solar radiation parameter, the current ambient temperature, the current ambient humidity and the current wind speed according to the time series; The S200 includes: S201, calculating the slope of each monitoring time point on each sequence distribution curve, obtaining all slopes on each sequence distribution curve, and constructing a current environment parameter-slope set; the slope is the rate of change of the parameter; the current environment parameter-slope set is the set of slopes of all points on the sequence distribution curve corresponding to each current environment parameter; S202, determining whether the slope in each current environment parameter-slope set is greater than or equal to a set first threshold; each current environment parameter-slope set is set with a different first threshold; S203, screening out all slopes that are greater than or equal to a set first threshold, and current environmental parameters corresponding to the screened slopes; S204, determining a fluctuation factor according to the screened slope and corresponding parameters; The S204 includes: S2041, setting a first weight value of all parameters in the current environment parameters, wherein the relationship between the first weight value, the parameter and the output power is determined; S2042, performing a weighted sum calculation on the screened slope and the corresponding first weight value to obtain a weighted sum value of all the screened parameters of the current environmental parameters; S2043, taking the parameter whose weighted sum value is greater than or equal to the sum threshold as a fluctuation factor; The S300 includes: S301, constructing a fluctuation factor database, recording a prediction model corresponding to each fluctuation factor in the fluctuation factor database, the prediction model including: a solar radiation parameter-prediction model, a current ambient temperature-prediction model, a current ambient humidity-prediction model, and a current wind speed-prediction model; S302, determining whether the number of volatility factors is greater than or equal to 1, if greater than or equal to 1, finding greater than or equal to 1 prediction models to form a prediction model set; The S400 includes: S401, setting a second weight value for each prediction model in the prediction model set; S402, inputting the current environmental parameters into all prediction models in the prediction model set respectively, and obtaining multiple prediction results through different prediction models; S403, corresponding to the second weight values ​​of different prediction models and the corresponding prediction results, calculate and determine the final prediction result.

2. The photovoltaic power station day-ahead power prediction method according to claim 1, It is characterized in that The S400 further includes: constructing a prediction model; the method for constructing a prediction model includes: S404, obtaining historical big data, the historical big data including: solar radiation parameters, current ambient temperature, current ambient humidity and current wind speed, and the relationship between the solar radiation parameters, current ambient temperature, current ambient humidity and current wind speed and the output power respectively; S405, according to the solar radiation parameter, the current ambient temperature, the current ambient humidity and the current wind speed having large fluctuation data as a classification principle, different data sets with the solar radiation parameter, the current ambient temperature, the current ambient humidity and the current wind speed as large fluctuation factors are classified respectively; S406, for different data sets, training is performed using the relationship between solar irradiation parameters, current ambient temperature, current ambient humidity and current wind speed and output power, to obtain different prediction models using solar irradiation parameters, current ambient temperature, current ambient humidity and current wind speed as fluctuation factors.

3. The photovoltaic power station day-ahead power prediction method according to claim 2, It is characterized in that The S406 includes: S4061, with the monitoring time point as the horizontal coordinate, and the monitoring value of each solar radiation parameter, current ambient temperature, current ambient humidity and current wind speed at the corresponding monitoring time point as the vertical coordinate, four sequence distribution curves with solar radiation parameter, current ambient temperature, current ambient humidity and current wind speed as the vertical coordinate are formed respectively; the horizontal coordinates are equally spaced; S4062, for a data set with different fluctuation factors, using the monitoring values ​​of the parameters corresponding to the fluctuation factors as the basis for change, distributing the monitoring values ​​in equal intervals, converting the sequence distribution curve into a conversion distribution curve with a constant speed change of the monitoring values, and the abscissa of the conversion distribution curve is distributed in unequal intervals; S4063, the conversion distribution curve and the corresponding output power are used as output parameters and input into the recurrent neural network RNN ​​for training to obtain a prediction result.

4. The photovoltaic power station day-ahead power prediction method according to claim 1, It is characterized in that The step S100 further includes: After obtaining the current environmental parameters, the abnormal data of the current environmental parameters are screened to filter out the abnormal data; Replace the abnormal data and add the replaced data to the original data position to form complete data, and construct a sequence distribution curve with the complete data.

5. The method for predicting the day-ahead power of a photovoltaic power station according to claim 4, It is characterized in that The screening out of abnormal data includes: if the data in a certain time period is abnormal compared with the change pattern of the day, but the data before and after the current time are periodic, then the data is non-abnormal data; The replacing operation on the abnormal data and adding the replaced data to the original data position includes: The data before the abnormal data is set as the front data, and the data after the abnormal data is set as the back data; Based on the previous data and the next data, the data is input into the data filling model, and the replacement data of the abnormal data is output; Supplement the replacement data corresponding to the monitoring time point to the original data position; The data filling model is obtained by training a deep neural network.

Images