A New Energy Power Prediction Method and System for Meteorological Information Simulation

Abstract

A new energy power prediction method and system for meteorological information simulation provided by the present invention include: obtaining meteorological data and preprocessing the data; constructing an intelligent meteorological large model, training the intelligent meteorological large model according to the meteorological data obtained in the first step to predict future meteorological conditions; constructing a new energy power generation prediction model; using historical measured data to perform backtesting on the intelligent meteorological large model and the energy power generation prediction model to evaluate the prediction accuracy. This method and system can significantly improve the accuracy and precision of new energy power generation prediction.

Description

Technical Field

[0001] The present invention belongs to the field of new energy power generation, and particularly relates to a new energy power prediction method and system for meteorological information simulation. Background Art

[0002] With the continuous increase in the global demand for renewable energy, the development and utilization of new energy wind power and photovoltaic power have become a trend. In this process, accurately predicting the power generation of new energy is crucial for the stable operation of the power grid and energy scheduling. However, there are still several limitations and challenges in the application of meteorological information and the prediction of new energy power generation in the existing technology. Limitations and deficiencies of the existing technology: 1. Meteorological conditions and power prediction accuracy: Existing prediction models usually rely on the application of traditional statistical methods or simple machine learning algorithms, and these algorithms often fail to achieve the high accuracy required by the power grid assessment when dealing with complex and non-linear meteorological data, especially the prediction performance is unstable under extreme weather conditions. 2. Data processing ability: Traditional models face problems of low efficiency and insufficient processing ability when dealing with large-scale and high-dimensional meteorological data, and the real-time processing and analysis ability of data is insufficient, resulting in delays in real-time prediction applications. 3. Model generalization ability: Existing technologies are often optimized for specific regions and specific types of meteorological conditions, lacking sufficient generalization ability, which limits their application effects in different geographical locations or different meteorological conditions. 4. Response ability to extreme events: The accuracy of existing prediction systems often drops significantly when extreme meteorological events occur, yet accurate prediction is most needed when such events occur.

[0003] Current situation and challenges of the application of meteorological information in new energy power generation:

[0004] 1. Availability and quality of meteorological data: Although meteorological data is widely available, the accuracy, timeliness, and integrity of the data are often limited. For example, the number of observation stations in some regions is scarce, resulting in the data obtained being unable to fully reflect the actual meteorological conditions.

[0005] 2. The power generation of wind power and photovoltaic power highly depends on meteorological conditions such as wind speed, solar radiation, etc. Any small deviation in meteorological prediction may lead to a large deviation in power generation prediction, affecting power generation efficiency and the safe and stable operation of the power grid.

[0006] 3. Integration and synchronization issues: Integrating meteorological data with the power generation system to achieve real-time synchronization and analysis of data is a technical challenge. Efficient data interfaces and processing algorithms are required to ensure the smooth flow and real-time nature of the data stream.

[0007] 4. Multi-scale time dependence: The power generation of new energy is affected by meteorological conditions on multiple time scales, ranging from hourly weather changes to seasonal and annual climate patterns. It is often difficult for existing technologies to effectively predict on these different time scales. Summary of the Invention

[0008] The technical problem to be solved by the present invention is to provide a new energy power prediction method and system for meteorological information simulation, which significantly improves the accuracy of new energy power generation prediction.

[0009] To solve the above technical problems, the technical solution adopted by the present invention is: A new energy power prediction method for meteorological information simulation, including the following steps:

[0010] Step 1: Obtain meteorological data and preprocess the meteorological data;

[0011] Step 2: Construct an intelligent meteorological large model, and train the intelligent meteorological large model according to the meteorological data obtained in Step 1 to predict future meteorological conditions;

[0012] Step 3: Construct a new energy power generation prediction model. The input variables of the new energy power generation prediction model are the wind speed and solar radiation intensity prediction data output by the intelligent meteorological large model in Step 2, and predict the new energy power generation.

[0013] In a preferred embodiment, in Step 1, the preprocessing of the meteorological data includes the following steps:

[0014] S101: Data cleaning: Remove outliers and fill in missing values;

[0015] S102: Data standardization: Normalize data with different dimensions.

[0016] In a preferred embodiment, in Step S102, the normalization formula is:

[0017] ;

[0018] where, is the original data, is the mean of the original data , is the standard deviation of the original data , is the normalization result of the original data .

[0019] In a preferred embodiment, in step 2, the intelligent meteorological large model adopts a neural network model based on physical information, and the model architecture includes: an input layer for inputting preprocessed meteorological data; a hidden layer with a multi-layer structure, where each layer contains multiple neurons; a physical constraint layer that introduces physical equations in meteorology as constraint conditions; and an output layer for outputting the output results of future meteorological conditions.

[0020] In a preferred embodiment, in step 2, the training process for training the intelligent meteorological large model includes the following steps:

[0021] S201. Data preparation: Construct a training set and a validation set based on the collected and preprocessed meteorological data.

[0022] S202. Initial training: Train the initial intelligent meteorological large model without physical constraints to obtain preliminary prediction results.

[0023] S203. Introduce physical constraints: Based on the initial intelligent meteorological large model, add physical equation constraints and train according to the loss function.

[0024] In a preferred embodiment, the loss function is the mean square error function, and the expression is:

[0025] ;

[0026] where represents the mean square error, is the actual value, is the predicted value, n is the number of samples, represents the sample serial number.

[0027] In a preferred embodiment, in step 3, the new energy power prediction model includes a wind power prediction model and a photovoltaic power prediction model, where:

[0028] Wind power prediction model: Based on the wind speed predict the wind power , and the expression is:

[0029] ;

[0030] where is the air density, A is the swept area of the wind turbine rotor, is the efficiency coefficient;

[0031] Photovoltaic power prediction model: Based on the solar radiation intensity predict the photovoltaic power , and the expression is:

[0032] ;

[0033] Among them, is the area of the photovoltaic panel, represents the efficiency coefficient of the photovoltaic panel.

[0034] In a preferred solution, it further includes Step 4 of using historical measured data to backtest the intelligent meteorological large model and the energy power prediction model, and evaluating the prediction accuracy of the intelligent meteorological large model and the energy power prediction model.

[0035] In a preferred solution, in Step 4, the mean square error function is used to evaluate the prediction accuracy of the intelligent meteorological large model and the energy power prediction model.

[0036] The present invention also provides a new energy power prediction system for meteorological information simulation, including a data analysis module for preprocessing meteorological data; a meteorological prediction module including an intelligent meteorological large model for meteorological prediction; and a power conversion module including a new energy power prediction model for wind power and photovoltaic power prediction.

[0037] A new energy power prediction method and system for meteorological information simulation provided by the present invention have the following beneficial effects:

[0038] 1. Improve the accuracy and real-time performance of meteorological data: Obtain more accurate and real-time meteorological information through multi-source data fusion and high-frequency acquisition.

[0039] 2. Optimize the meteorological large model and power prediction algorithm: Use deep learning models to process complex meteorological data, extract key features, construct a high-precision meteorological prediction model, and at the same time design an efficient power conversion algorithm to achieve accurate prediction from meteorological data to power generation.

[0040] 3. The intelligent meteorological large model combines meteorological data and physical equations. In the process of combining data-driven and physical-driven, the system can not only capture the complex time-varying characteristics in meteorological data, but also ensure that the prediction results conform to physical laws, significantly improving the accuracy of meteorological prediction.

[0041] 4. Integrate the meteorological large model and the power conversion module into an intelligent system, support real-time data processing and dynamic adjustment, and ensure the real-time performance and reliability of the prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0043] Figure 1 This is the flowchart of the present invention. Detailed implementation manners

[0044] In combination with Figure 1 The detailed description of the specific implementation manners of the present invention is further provided.

[0045] A clear and complete description is given. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0046] Embodiment 1:

[0047] A new energy power prediction method for meteorological information simulation includes the following steps:

[0048] Step 1: Obtain meteorological data and preprocess the meteorological data.

[0049] Sources and types of meteorological data: Meteorological data is the basis for predicting new energy power generation, and its accuracy and timeliness directly affect the prediction results. In this embodiment, data is mainly obtained from the following aspects:

[0050] 1. Ground observation station data: including conventional meteorological elements such as temperature, humidity, wind speed, wind direction, and air pressure.

[0051] The wind speed includes wind speed data at different heights, such as wind speed data at heights of 10 meters, 50 meters, 100 meters, etc., as well as instantaneous wind speed and average wind speed; the wind direction data includes wind direction data at different heights and the wind direction change trend; the air temperature data includes ground air temperature and air temperature at different heights, as well as the daily maximum air temperature and minimum air temperature; the air pressure data includes ground air pressure and air pressure at different heights, as well as the air pressure change trend; the humidity data includes relative humidity, absolute humidity, and humidity data at different heights.

[0052] 2. Satellite remote sensing data: cloud amount, radiation intensity, etc.

[0053] 3. Radar monitoring data: precipitation, wind field, etc. The precipitation data includes precipitation intensity, precipitation amount, and precipitation type, and also includes radar wind measurement data.

[0054] Specific data sources and collection frequencies:

[0055] Ground meteorological stations: Provide accurate ground meteorological data, usually updated hourly or every minute. Suitable for collecting data such as wind speed, wind direction, air temperature, air pressure, and humidity.

[0056] Upper-air meteorological sounding: Provide meteorological data at different altitudes, usually updated daily or every few days. Suitable for collecting data such as upper-air wind speed, wind direction, and air temperature.

[0057] Meteorological satellite: Provide meteorological data over a large area, usually updated every 15 minutes to 1 hour. Suitable for collecting data such as cloud cover, solar radiation, and radiation intensity.

[0058] Weather radar: Provide fine precipitation and wind field data, usually updated every 5 minutes. Suitable for collecting data such as precipitation intensity and radar wind measurement.

[0059] Meteorological data includes historical meteorological data and real-time meteorological data. Historical meteorological data is used for model training and validation, and real-time meteorological data is used for real-time prediction.

[0060] The preprocessing of meteorological data includes the following steps:

[0061] S101. Data cleaning: Remove outliers and fill in missing values.

[0062] The interquartile range method (IQR) can be used to remove outliers.

[0063] IQR = Q3 - Q1, where Q1 is the 25th percentile and Q3 is the 75th percentile.

[0064] Define the lower limit and the upper limit. Lower limit = Q1 - 1.5 × IQR, upper limit = Q3 + 1.5 × IQR.

[0065] Data points below the lower limit and above the upper limit are regarded as outliers.

[0066] The k-nearest neighbor filling method can be used to fill in missing values: First, standardize and normalize the data, remove outliers that cause sample missing, then for each sample containing missing values, calculate its Euclidean distance from other samples. According to the calculated distances, select the k nearest neighbors that are most similar to the current sample. Arrange the calculated Euclidean distances in ascending order of value. The first k with the smallest Euclidean distance are the k nearest neighbors that are most similar. Use the mean of the k neighbors to fill in the missing value.

[0067] S102. Data standardization: Normalize data with different dimensions.

[0068] The normalization formula is as follows:

[0069] ;

[0070] Where, is the original data, is the original data the mean of, is the original data , the standard deviation of , is the normalization result of the original data .

[0071] Step 2: Construct an intelligent meteorological large model, and train the intelligent meteorological large model based on the meteorological data obtained in Step 1 to predict future meteorological conditions.

[0072] The intelligent meteorological large model adopts a physics-informed neural network model (Physics-Informed Neural Networks, PINN). The PINN algorithm integrates physical laws into the deep learning model, enabling it to not only capture the statistical characteristics of meteorological data but also follow the physical laws in meteorology. The PINN model combines traditional deep neural networks and physical equations, and by constraining the training process of the neural network, makes its output conform to physical laws.

[0073] The PINN model architecture is as follows:

[0074] 1) Input layer: Includes all preprocessed meteorological data.

[0075] 2) Hidden layer: A multi-layer neural network structure, with each layer containing multiple neurons. The non-linear activation functions ReLu or Tanh activation functions are used to enhance the model's expressive ability.

[0076] 3) Physical constraint layer: Introduce physical equations in meteorology as constraint conditions.

[0077] In the meteorological large model, common physical equations include the atmospheric dynamics equation and the thermodynamics equation. The following are some key physical equations:

[0078] 1. The Navier-Stokes equation, used to describe fluid motion, such as wind speed and direction, with the expression

[0079] ;

[0080] where u is the velocity field, t is time, ρ is air density, p is pressure, ν is the viscosity coefficient, f is the external force, represents the gradient.

[0081] 2. The mass conservation equation or continuity equation is:

[0082] ;

[0083] This equation describes the incompressibility of the fluid.

[0084] 3. The energy conservation equation or thermodynamics equation:

[0085] ;

[0086] Among them, T is the temperature field, κ is the thermal diffusion coefficient, and Q is the heat source term.

[0087] 4) Output layer: Output the output result of future meteorological conditions.

[0088] The training process of training the intelligent meteorological large model is as follows:

[0089] S201. Data preparation: Construct a training set and a validation set according to the collected and preprocessed meteorological data.

[0090] Integrate all data into a large dataset. Divide the large dataset into different data segments by day. Each segment contains the meteorological data segments of the current day, the previous day, and the next day, a total of three days, and ensure that the data segments are ranked in chronological order. Divide according to a certain ratio. In this embodiment, 80% is used as the training set and 20% is used as the validation set.

[0091] S202. Initial training: Without physical constraints, use traditional neural network training methods to train the initial model to obtain preliminary prediction results.

[0092] Traditional neural networks can use Long Short-Term Memory (LSTM) or Convolutional Neural Network (CNN).

[0093] LSTM is suitable for processing and predicting time series data and can capture long-term dependencies; CNN is suitable for processing image data and can be used to capture local spatio-temporal features in meteorology.

[0094] S203. Introduce physical constraints: On the basis of the initial model, add physical equation constraints and train by optimizing the loss function to make the model output conform to physical laws.

[0095] The loss function uses the mean square error MSE, and the expression is as follows:

[0096] ;

[0097] Among them, is the actual value, is the predicted value, and n is the number of samples.

[0098] In this embodiment, the Adam optimizer and the L-BFGS optimizer are used for joint training. The former is used for fast convergence, and the latter is used for fine-tuning. The optimal training parameters are determined through cross-validation, and the early stopping mechanism Early Stopping is adopted to prevent overfitting. Physical constraint optimization: According to different meteorological conditions and regional characteristics, the parameters in the physical equation are dynamically adjusted to make the model more adaptable.

[0099] To ensure the accuracy of the model, the meteorological model needs to be updated regularly, including: rolling update: using the latest meteorological data to retrain the model at fixed time intervals; online learning: continuously adjusting the model parameters through real-time meteorological data to improve the real-time prediction ability.

[0100] The methods for real-time adjustment of model parameters in online learning include the following several types:

[0101] 1) Incremental training:

[0102] Incremental training means further training using new data on the basis of an existing model to update the model parameters.

[0103] The operation steps are as follows:

[0104] 1. Initial model training: First, train an initial model using the initial historical data.

[0105] 2. Obtain new data: Obtain new meteorological data in real time.

[0106] 3. Incremental training: Use the new data to further train the initial model and update the model parameters.

[0107] 2) Online gradient descent:

[0108] Online gradient descent is a technique for updating model parameters one by one. Each time, the gradient is calculated using one sample and the weights are updated.

[0109] The operation steps are as follows:

[0110] 1. Initial model training: Train an initial model using historical data.

[0111] 2. Update one by one: Process new samples one by one, calculate the gradient, and update the model parameters.

[0112] 3) Window method:

[0113] The window method is to use the data within a recent time window for training to ensure that the model has good adaptability to the latest data.

[0114] The operation steps are as follows:

[0115] 1. Define the time window: Set a time window, for example, the most recent 100 data points.

[0116] 2. Rolling update: Each time new data arrives, update the data within the time window and use this data to retrain the model.

[0117] 4) Learning rate adjustment:

[0118] Learning rate adjustment is a method to improve the online learning performance of the model by dynamically adjusting the learning rate.

[0119] Step 3: Build a new energy power generation prediction model. The input variables of the new energy power generation prediction model are the wind speed and solar radiation intensity prediction data output by the intelligent meteorological large model in Step 2.

[0120] The new energy power generation prediction model includes a wind power prediction model and a photovoltaic power prediction model, specifically as follows:

[0121] Wind power prediction model: Based on the wind speed Predict the wind power , and the expression is as follows:

[0122] ;

[0123] Where is the air density, A is the swept area of the wind turbine rotor, is the efficiency coefficient.

[0124] Photovoltaic power prediction model: Based on the solar radiation intensity Predict the photovoltaic power generation , and the expression is as follows:

[0125] ;

[0126] Where is the area of the photovoltaic panel, represents the conversion efficiency of the photovoltaic panel.

[0127] Step 4: Use historical measured data to perform backtesting on the intelligent meteorological large model and the energy power generation prediction model to evaluate the prediction accuracy.

[0128] In this embodiment, the mean square error is used for evaluation.

[0129] Taking the wind power generation prediction model as an example, the mean square error is used as the main evaluation index, and the specific implementation steps are as follows.

[0130] The mean square error MSE, and the expression is as follows:

[0131] ;

[0132] Among them, is the actual value, is the predicted value, and n is the number of samples.

[0133] Use the energy power generation prediction model to backtest the historical power generation power. is the simulation result given by the energy power generation prediction model for past events that have occurred. is the true value of the historical record of the actual new energy power station power generation. Substitute the two into the above formula to calculate MSE, which can be used to evaluate the prediction performance of the energy power generation prediction model. A lower MSE indicates a higher prediction accuracy of the model.

[0134] Embodiment 2:

[0135] A new energy power prediction system for meteorological information simulation includes: a data analysis module for preprocessing meteorological data; a meteorological prediction module containing an intelligent meteorological large model for meteorological prediction; and a power conversion module containing a new energy power generation prediction model for wind power and photovoltaic power prediction.

[0136] Each functional module is developed independently, which is convenient for system integration and maintenance.

[0137] Real-time data processing and preprocessing result output: 1. Implement data access, the system accesses the latest meteorological data in real time and performs data preprocessing. 2. Real-time prediction: Use the trained intelligent meteorological large model and power conversion module to implement and output the prediction of new energy power prediction results. 3. Result display and feedback: Intuitively display the new energy power prediction results in the form of charts, dashboards, etc., and receive user feedback to further optimize the integrated system.

[0138] The above embodiments are only the preferred technical solutions of the present invention and should not be regarded as limitations to the present invention. The protection scope of the present invention should be the technical solutions recorded in the claims, including equivalent replacement solutions of the technical features in the technical solutions recorded in the claims. For those of ordinary skill in the art, other different forms of changes or modifications can be made on the basis of the above description. It is not necessary and impossible to enumerate all the implementation manners here. That is, the equivalent replacement improvements within this range are also within the protection scope of the present invention.

Claims

1. A new energy power prediction method for meteorological information simulation, characterized in that: The following steps are involved: Step 1: Obtain meteorological data and pre-process the meteorological data; Step 2: Build an intelligent meteorological big model, train the intelligent meteorological big model according to the meteorological data obtained in step 1, and predict future meteorological conditions; The intelligent meteorological model uses a neural network model based on physical information. The model architecture includes: Input layer, used to input preprocessed meteorological data; Hidden layer, adopts a multi-layer structure, each layer contains multiple neurons; The physical constraint layer introduces the physical equations in meteorology as constraints. The physical equations include: 1) Navier-Stokes equation, used to describe fluid motion, is expressed as: ； Among them, u is the velocity field, t is the time, ρ is the air density, p is the pressure, ν is the viscosity coefficient, and f is the external force. represents the gradient; 2) The mass conservation equation or continuity equation is: ； This equation describes the incompressibility of the fluid; 3) Energy conservation equation or thermodynamic equation: ； Where T is the temperature field, κ is the thermal diffusion coefficient, and Q is the heat source term; The output layer is used to output the future meteorological conditions output results; Step 3: construct a new energy power generation prediction model. The input variables of the new energy power generation prediction model are the wind speed and solar radiation intensity prediction data output by the intelligent meteorological model in step 2, and the new energy power generation is predicted; The new energy power generation prediction model includes wind power prediction model and photovoltaic power prediction model, among which: Wind power prediction model: based on wind speed Predicting wind power , the expression is: ； in, is the air density, A is the wind wheel swept area, is the efficiency coefficient; Photovoltaic power prediction model: based on solar radiation intensity Predicting photovoltaic power generation , the expression is: ； in, is the photovoltaic panel area, Represents the efficiency coefficient of photovoltaic panels.

2. A new energy power prediction method for meteorological information simulation according to claim 1, characterized in that: In the step 1, preprocessing the meteorological data includes the following steps: S101, data cleaning: remove outliers and fill in missing values; Remove outliers using the interquartile range method: IQR = Q3-Q1, where Q1 is the 25th percentile and Q3 is the 75th percentile; Define the lower limit and upper limit, lower limit = Q1-1.5×IQR, upper limit = Q3+1.5×IQR; Data points below the lower limit and above the upper limit are considered outliers; Missing values ​​can be filled using the k-nearest neighbor filling method: first, the data is standardized and normalized to remove outliers that cause missing samples. Then, for each sample containing missing values, the Euclidean distance between it and other samples is calculated. Based on the calculated distance, the k neighbors that are most similar to the current sample are selected. The calculated Euclidean distances are arranged from small to large in numerical order. The first k neighbors with the smallest Euclidean distance are the most similar k neighbors, and the missing values ​​are filled with the mean of the k neighbors. S102. Data standardization: normalize data of different dimensions.

3. A new energy power prediction method for meteorological information simulation according to claim 2, characterized in that: In step S102, the normalized formula is: ； in, is the original data, For the original data The mean of It is the original data The standard deviation of For the original data The normalized result of .

4. The new energy power prediction method for meteorological information simulation according to claim 1 is characterized in that: In step 2, the training process of training the intelligent meteorological large model includes the following steps: S201, data preparation: construct training set and validation set based on collected and preprocessed meteorological data; S202, initial training: without physical constraints, training the initial intelligent meteorological model to obtain preliminary prediction results; S203. Introducing physical constraints: Based on the initial intelligent meteorological model, physical equation constraints are added and training is performed according to the loss function.

5. A new energy power prediction method for meteorological information simulation according to claim 4, characterized in that: The loss function is the mean square error function, and the expression is: ； in, represents the mean square error, is the actual value, is the predicted value, n is the number of samples, Indicates the sample number.

6. The new energy power prediction method for meteorological information simulation according to claim 1 is characterized in that: It also includes step 4, using historical measured data to test the intelligent meteorological big model and the energy power generation prediction model, and evaluating the prediction accuracy of the intelligent meteorological big model and the energy power generation prediction model.

7. A new energy power prediction method for meteorological information simulation according to claim 6, characterized in that: In the step 4, the mean square error function is used to evaluate the prediction accuracy of the intelligent meteorological model and the energy generation power prediction model.

8. A prediction system using a new energy power prediction method for meteorological information simulation according to any one of claims 1 to 7, characterized in that: It includes a data analysis module for preprocessing meteorological data; a meteorological prediction module, which includes an intelligent meteorological model for meteorological prediction; and a power conversion module, which includes a new energy power generation prediction model for wind power and photovoltaic power prediction.

Images