Micrometeorological irradiance prediction method and system and medium

Abstract

The invention discloses a micrometeorological irradiance prediction method and system and a medium, and the method comprises the following steps: obtaining a historical meteorological data set, and carrying out the preprocessing of the historical meteorological data set; introducing weather types, respectively constructing corresponding micrometeorological prediction models, and carrying out irradiance prediction; according to the method, the predicted value of the irradiance obtained through the micrometeorological prediction model is corrected, a final irradiance prediction result is obtained, the irradiance of the fixed point area is effectively and accurately predicted, and the accuracy of meteorological prediction information is improved.

Description

Technical Field

[0001] The present invention relates to solar irradiance prediction, and in particular to a micro-meteorological irradiance prediction method, system and medium. Background Art

[0002] As a clean and renewable energy source, solar energy can not only meet the global electricity demand, but also slow down the global warming caused by traditional energy sources. However, the temporal and spatial variations of solar irradiance are large. When large-scale photovoltaic power generation is integrated into the power grid, it will lead to defects in the reliability and stability of the power grid, and increase the complexity of power grid management. Therefore, in the power system, the prediction accuracy of solar irradiance is increasingly required to ensure reliable photovoltaic power generation and provide a more reliable scheduling solution for power grid operation. At the same time, the corresponding irradiance-related information cannot be obtained through the regional meteorological forecast system, or the prediction granularity is coarse or the range given is large rather than the irradiance of a fixed area, which cannot meet the needs of the new power system. Summary of the invention

[0003] Purpose of the invention: The present invention aims to provide a micrometeorological irradiance prediction method that can perform fixed-point prediction. Another purpose of the present invention is to provide a micrometeorological irradiance prediction system and medium.

[0004] Technical solution: The micro-meteorological irradiance prediction method of the present invention comprises the following steps:

[0005] Obtain historical meteorological data sets and preprocess them;

[0006] Introduce weather types, build corresponding micro-meteorological prediction models, and make irradiance predictions;

[0007] The predicted value of irradiance obtained by the micrometeorological prediction model is corrected to obtain the final irradiance prediction result.

[0008] Furthermore, the historical meteorological dataset includes wind speed, wind direction, temperature, humidity, atmospheric pressure, rainfall, irradiance, and collection time point.

[0009] Furthermore, the historical meteorological data set is preprocessed as follows:

[0010] Data alignment: align historical meteorological data according to the collection time point;

[0011] Data cleaning, calculation of historical meteorological data set deviations and default data adjustments;

[0012] Data classification: divide the historical meteorological data set into data packets according to N+T hours. The start time of adjacent data packets differs by M hours. According to the weather corresponding to N hours and T hours in the data packet, determine the data classification. The data classification includes sunny to sunny, sunny to rainy, sunny to cloudy, rainy to sunny, rainy to rainy, rainy to cloudy, cloudy to sunny, cloudy to rainy, and cloudy to cloudy;

[0013] Data conversion: if the rainy time exceeds 1 / 3 of the time period, the weather in this time period is rainy and recorded as 1; if the normal weather time exceeds 1 / 2 of the time period, the weather in this time period is sunny and recorded as 2; the rest of the time, the weather is cloudy and recorded as 3;

[0014] Numerical characterization of micro-meteorological information converts wind direction information, time information, season and month information into numerical characteristic information.

[0015] Furthermore, the data cleaning is specifically as follows:

[0016] For single missing numerical micrometeorological elements, the two-way exponential smoothing method is used to take the mean value to fill in the single missing numerical data;

[0017] For several missing numerical micro-meteorological elements, a sliding window with adaptive length combined with a two-way average method is used to fill in the missing numerical data.

[0018] For a single numerical outlier, a single exponential smoothing method is used to fill and replace the single numerical outlier;

[0019] For missing non-numerical micro-meteorological elements, element classification is performed and the method for missing numerical micro-meteorological elements is used to handle them.

[0020] Furthermore, the missing numerical micrometeorological element y at time t is obtained by using the two-way exponential smoothing method. t for

[0021]

[0022] Among them, α is the positive smoothing coefficient, y at is the forward smoothing value at time t, β is the reverse smoothing coefficient, is the reverse smoothing value at time t.

[0023] Furthermore, weather types are introduced and corresponding micro-meteorological prediction models are constructed to perform irradiance prediction. The xgboost algorithm is used to predict the weather conditions in the next T hours. According to the characteristic information of the current N hours of historical meteorological data and the weather conditions in the next T hours, the irradiance prediction is realized through the Informer model of the corresponding weather type.

[0024] Furthermore, when the weather type is sunny to rainy, rainy to rainy, or cloudy to rainy, the corresponding micrometeorological prediction model loss function loss r for

[0025]

[0026] Among them, n is the number of training set samples, y ti is the true value at time i, y pi is the predicted value at time i, y rmax is the maximum value on rainy days;

[0027] When the weather type is sunny to sunny, sunny to cloudy, rainy to sunny, rainy to cloudy, cloudy to sunny, or cloudy to cloudy, the corresponding micrometeorological prediction model loss function loss s for

[0028]

[0029] The micro-meteorological irradiance prediction system of the present invention comprises:

[0030] Data preprocessing module, used to obtain historical meteorological data sets and perform preprocessing;

[0031] The model building module is used to introduce weather types, build corresponding micro-meteorological prediction models, and perform irradiance prediction;

[0032] The correction module corrects the predicted value of irradiance obtained by the micro-meteorological prediction model to obtain the final irradiance prediction result.

[0033] The computer device of the present invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the steps of the above method are implemented when the processor executes the computer program.

[0034] The computer-readable storage medium of the present invention stores a computer program thereon, and the computer program implements the steps of the above method when executed by a processor.

[0035] Beneficial effect: Compared with the prior art, the present invention has the following significant advantages: the present invention introduces weather types, constructs corresponding micrometeorological prediction models, performs irradiance prediction, and corrects the predicted irradiance value to obtain the final irradiance prediction result, thereby effectively predicting the irradiance of a fixed-point area. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is a flow chart of the method of the present invention;

[0037] Figure 2This is the irradiance curve of micrometeorology under various weather conditions. DETAILED DESCRIPTION

[0038] The present invention will be further described below in conjunction with the accompanying drawings.

[0039] The micro-meteorological irradiance prediction method of the present invention comprises the following steps:

[0040] 1. Obtain historical meteorological data sets and preprocess them.

[0041] Historical meteorological data collection: Collect wind speed, wind direction, temperature, humidity, atmospheric pressure, rainfall, and irradiance collected by designated micro-meteorological monitoring equipment 6 hours before the start of irradiance data recording every day, and align the data by collection time point.

[0042] Micro-meteorological data classification: historical meteorological data is divided into data packets according to 8 hours, and the start time of adjacent data packets differs by 2 hours. At the same time, the standard for splitting each data packet is 6+2. The data classification is determined based on the weather corresponding to 6 hours and 2 hours in the data packet. The weather in the first 6 hours is sunny, rainy or cloudy, and the weather in the last 2 hours is sunny, rainy or cloudy, so the specific classification standards are sunny to sunny, sunny to rainy, sunny to cloudy, rainy to sunny, rainy to rainy, rainy to cloudy, cloudy to sunny, cloudy to rainy, and cloudy to cloudy.

[0043] Data cleaning: Filter out abnormal and missing data within the time period. For single missing data, use the two-way exponential smoothing method to replace it. The specific replacement formula is as follows:

[0044]

[0045] Where: α is the positive smoothing coefficient, y at is the forward smoothing value at time t, β is the reverse smoothing coefficient, is the reverse smoothing value at time t.

[0046] For the case where there are multiple consecutive missing numerical micro-meteorological elements, a sliding window with adaptive length combined with a two-way averaging method is used to supplement the missing values.

[0047] Data conversion: Weather with rainfall for more than 1 / 3 of the time period is defined as rainy day and recorded as 1; weather with normal irradiance for more than 1 / 2 of the statistical time period is defined as sunny day and recorded as 2; other situations are defined as cloudy day and recorded as 3.

[0048] Construct features: convert all non-numerical micro-meteorological information into numerical feature information; convert wind direction information into feature information between 0 and 1; construct feature information for time information; construct feature information for season and month information.

[0049] 2. Introduce weather types and construct corresponding micro-meteorological prediction models to perform irradiance prediction.

[0050] Weather classification model: Using the constructed feature data, the xgboost algorithm framework is used to predict the possible weather conditions (sunny, cloudy, rainy) in the next 2 hours. The corresponding parameters are shown in Table 1.

[0051] Table 1

[0052]

[0053] With the current model parameters, the classification results have an accuracy of 0.9104, a recall of 0.9258, and an f1 of 0.9181.

[0054] Irradiance prediction model: Using the constructed features and the prediction results of the weather classification model, Informer is used to build models for sunny to sunny, sunny to rainy, sunny to cloudy, rainy to sunny, rainy to rainy, rainy to cloudy, cloudy to sunny, cloudy to rainy, and cloudy to cloudy to predict the irradiance in the next 2 hours. The final value of the model loss function is shown in Table 2.

[0055] When the weather type is sunny to rainy, rainy to rainy, or cloudy to rainy, the corresponding micro-meteorological prediction model loss function loss r for

[0056]

[0057] Among them, n is the number of training set samples, y ti is the true value at time i, y pi is the predicted value at time i, y rmax is the maximum value on rainy days;

[0058] When the weather type is sunny to sunny, sunny to cloudy, rainy to sunny, rainy to cloudy, cloudy to sunny, or cloudy to cloudy, the corresponding micrometeorological prediction model loss function loss s for

[0059]

[0060] The irradiance on rainy days is relatively small. In order to prevent the irradiance prediction from having large deviations and resulting in failure to converge, the maximum irradiance on rainy days in historical statistics is used.

[0061] Table 2

[0062]

[0063] 3. Correct the predicted value of irradiance obtained by the micrometeorological prediction model to obtain the final irradiance prediction result.

[0064] Conduct data experiment statistics on the predicted irradiance according to different prediction models, and count the probability that the prediction result of each model is too large or too small, or the content in different time periods is too large or too small;

[0065] The specific method is: count the number of prediction results given by each model that are greater than the true value and the number of prediction results that are less than the true value.

[0066] If the number of statistical results that are greater than the true value exceeds 70%, the model calculated value is considered to be too large. The oscillation range of all the large data is calculated and divided into 10 interval segments. The value range with the most landing points is found, and the corresponding prediction result value is adjusted in reverse. At the same time, for the statistical interval, the minimum true value is calculated. If the reverse adjustment is lower than the true value, it is replaced by the lowest value of the statistical result of the corresponding time period.

[0067] For statistical results that are too small, the adjustment plan is to adjust the data in the opposite way to the calculation method for statistical results that are too large.

[0068] For model output results that are not too large or too small, the maximum and minimum values ​​within the event segment are counted, and the values ​​that exceed the maximum or are lower than the minimum are replaced by the maximum and minimum values.

[0069] The micro-meteorological irradiance prediction system of the present invention comprises:

[0070] Data preprocessing module, used to obtain historical meteorological data sets and perform preprocessing;

[0071] The model building module is used to introduce weather types, build corresponding micro-meteorological prediction models, and perform irradiance prediction;

[0072] The correction module corrects the predicted value of irradiance obtained by the micro-meteorological prediction model to obtain the final irradiance prediction result.

[0073] The computer device of the present invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the steps of the above method are implemented when the processor executes the computer program.

[0074] The computer-readable storage medium of the present invention stores a computer program thereon, and the computer program implements the steps of the above method when executed by a processor.

Claims

1. A micro-meteorological irradiance prediction method, characterized in that: The following steps are involved: Obtain historical meteorological data sets and preprocess them; Introduce weather types, build corresponding micro-meteorological prediction models, and make irradiance predictions; The predicted value of irradiance obtained by the micrometeorological prediction model is corrected to obtain the final irradiance prediction result.

2. The micrometeorological irradiance prediction method according to claim 1, characterized in that: The historical meteorological dataset includes wind speed, wind direction, temperature, humidity, atmospheric pressure, rainfall, irradiance, and collection time point.

3. The micrometeorological irradiance prediction method according to claim 2 is characterized in that: The historical meteorological data set is preprocessed as follows: Data alignment: align historical meteorological data according to the collection time point; Data cleaning, calculation of historical meteorological data set deviations and default data adjustments; Data classification: divide the historical meteorological data set into data packets according to N+T hours. The start time of adjacent data packets differs by M hours. According to the weather corresponding to N hours and T hours in the data packet, determine the data classification. The data classification includes sunny to sunny, sunny to rainy, sunny to cloudy, rainy to sunny, rainy to rainy, rainy to cloudy, cloudy to sunny, cloudy to rainy, and cloudy to cloudy; Data conversion: if the rainy time exceeds 1 / 3 of the time period, the weather in this time period is rainy and recorded as 1; if the normal weather time exceeds 1 / 2 of the time period, the weather in this time period is sunny and recorded as 2; the rest of the time, the weather is cloudy and recorded as 3; Numerical characterization of micro-meteorological information converts wind direction information, time information, season and month information into numerical characteristic information.

4. The micrometeorological irradiance prediction method according to claim 3 is characterized in that: The data cleaning is specifically as follows: For single missing numerical micrometeorological elements, the two-way exponential smoothing method is used to take the mean value to fill in the single missing numerical data; For several missing numerical micro-meteorological elements, a sliding window with adaptive length combined with a two-way average method is used to fill in the missing numerical data. For a single numerical outlier, a single exponential smoothing method is used to fill and replace the single numerical outlier; For missing non-numerical micro-meteorological elements, element classification is performed and the method for missing numerical micro-meteorological elements is used to handle them.

5. The micrometeorological irradiance prediction method according to claim 4, characterized in that: The missing numerical micrometeorological element y at time t obtained by using the two-way exponential smoothing method t for and t =[α×y αt +(1-α)y t-1 +β×y βt +(1-β)y t+1 ] / 2 Among them, α is the positive smoothing coefficient, y at is the forward smoothing value at time t, β is the reverse smoothing coefficient, y βt is the reverse smoothing value at time t.

6. The micrometeorological irradiance prediction method according to claim 3 is characterized in that: Weather types are introduced, and corresponding micro-meteorological prediction models are constructed to predict irradiance. The xgboost algorithm is used to predict the weather conditions in the next T hours. According to the characteristic information of the current N hours of historical meteorological data and the weather conditions in the next T hours, the irradiance prediction is realized through the Informer model of the corresponding weather type.

7. The micrometeorological irradiance prediction method according to claim 6, characterized in that: When the weather type is sunny to rainy, rainy to rainy, or cloudy to rainy, the corresponding micro-meteorological prediction model loss function loss r for Among them, n is the number of training set samples, y ti is the true value at time i, y pi is the predicted value at time i, y rmax is the maximum value on rainy days; When the weather type is sunny to sunny, sunny to cloudy, rainy to sunny, rainy to cloudy, cloudy to sunny, or cloudy to cloudy, the corresponding micrometeorological prediction model loss function loss s for 8. A micro-meteorological irradiance prediction system, characterized in that: include Data preprocessing module, used to obtain historical meteorological data sets and perform preprocessing; The model building module is used to introduce weather types, build corresponding micro-meteorological prediction models, and perform irradiance prediction; The correction module corrects the predicted value of irradiance obtained by the micro-meteorological prediction model to obtain the final irradiance prediction result.

9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images