A kalman filter corrected irradiance prediction method and system

Abstract

The application provides a kind of irradiance prediction method and system of kalman filter correction, comprising: step S1, the current time of a target area is measured irradiance is collected, and the weather data of target area is obtained, the weather data is input into a weather prediction model to predict at least a future time of target area The predicted irradiance of;Step S2, the predicted irradiance and measured irradiance are input into a kalman filter correction model obtained by pre-training to obtain a corrected irradiance at future time, as the irradiance prediction result of target area.The beneficial effect is that the method and system establish the kalman filter correction model by collecting historical irradiance data to correct the predicted irradiance obtained by the weather prediction model, solve the problem of low prediction accuracy and poor prediction effect, which is of great significance to the smooth operation of power grid.

Description

Technical Field

[0001] This invention relates to the field of irradiance prediction technology, and in particular to an irradiance prediction method and system with Kalman filter correction. Background Technology

[0002] Solar irradiance is an important parameter for quantitatively describing and studying solar radiation, and it is also an important factor affecting photovoltaic power generation. Solar irradiance refers to the radiant energy per unit area per unit time that reaches the Earth's surface after the solar radiation has been absorbed, scattered, and reflected by the atmosphere.

[0003] Current irradiance forecasting methods are divided into statistical methods and physical methods. Statistical methods do not consider the impact of meteorological changes on irradiance, have relatively simple data structures, and are difficult to obtain accurate forecast results. Physical methods require a clear understanding of actual physical processes and the ability to reproduce them, but due to the uncertainty of parameters, they have large deviations. Summary of the Invention

[0004] To address the problems existing in the prior art, this invention provides a Kalman filter-corrected irradiance prediction method, specifically including the following steps:

[0005] Step S1: Collect a measured irradiance of a target area at the current time, and obtain meteorological data of the target area. Input the meteorological data into a weather forecast model to predict the predicted irradiance of the target area at least one future time.

[0006] Step S2: Input the predicted irradiance and the measured irradiance into a pre-trained Kalman filter correction model to obtain a corrected irradiance at the future time, which is used as the irradiance prediction result for the target area.

[0007] Preferably, a model training process is included before performing step S1, specifically including the following steps:

[0008] Step A1: Obtain multiple sets of historical irradiance data for multiple target areas. Each set of historical irradiance data includes a historical predicted irradiance data and a historical measured irradiance data. The historical predicted irradiance data is the predicted irradiance at a historical moment after the acquisition time, which is predicted at the acquisition time of the historical measured irradiance data.

[0009] Step A2: Based on the historical predicted irradiance data and the historical measured irradiance data, process the data to obtain a future predicted irradiance at the historical moment predicted at the acquisition time;

[0010] Step A3: Using the historical predicted irradiance data and the historical measured irradiance data as inputs, and the future predicted irradiance as output, the Kalman filter correction model is trained.

[0011] Preferably, before step A2, the method further includes:

[0012] Based on an iterative optimization strategy and an isolated forest algorithm, the historical irradiance data of each group are preprocessed to obtain preprocessed predicted irradiance data and preprocessed measured irradiance data.

[0013] In step A2, the future predicted irradiance at the historical time predicted at the acquisition time is obtained by processing the preprocessed predicted irradiance data and the preprocessed measured irradiance data.

[0014] Preferably, step A2 includes:

[0015] Step A21: For each set of historical irradiance data, process to obtain an irradiance difference between the historical predicted irradiance data and the historical measured irradiance data, and process according to the historical predicted irradiance data to obtain an observation matrix at the acquisition time;

[0016] Step A22: Based on the irradiance difference and the historical measured irradiance data, a state vector for the historical moment is obtained;

[0017] Step A23: Based on the observation matrix, the state vector, and a pre-set measurement noise processing, a revised difference is obtained. Based on the revised difference and the historical measured irradiance data, the future predicted irradiance at the historical time predicted at the acquisition time is obtained.

[0018] Preferably, the revised difference is obtained by processing the following formula:

[0019] y t =H t x t-1 +q t

[0020] in,

[0021] y t This represents the revised difference;

[0022] H t Represents the observation matrix;

[0023] x t-1 Represents the state vector;

[0024] q t This indicates the measurement noise.

[0025] Preferably, a Kalman filter-corrected irradiance prediction system, applied to the above-mentioned irradiance prediction method, includes:

[0026] A data acquisition module is used to collect a measured irradiance of a target area at the current time, acquire meteorological data of the target area, and input the meteorological data into a weather forecast model to predict the predicted irradiance of the target area at least one future time.

[0027] An irradiance correction module, connected to the data acquisition module, is used to input the predicted irradiance and the measured irradiance into a pre-trained Kalman filter correction model to obtain a corrected irradiance at the future time, which serves as the irradiance prediction result for the target area.

[0028] Preferably, it further includes a model training module connected to the irradiance correction module, comprising:

[0029] A data acquisition unit is used to acquire multiple sets of historical irradiance data for multiple target areas. Each set of historical irradiance data includes a historical predicted irradiance data and a historical measured irradiance data. The historical predicted irradiance data is the predicted irradiance at a historical moment after the acquisition time, which is predicted at the acquisition time of the historical measured irradiance data.

[0030] A processing unit, connected to the acquisition unit, is used to process the historical predicted irradiance data and the historical measured irradiance data to obtain a future predicted irradiance at the historical time predicted at the acquisition time.

[0031] A training unit, connected to the processing unit, is used to train the Kalman filter correction model by taking the historical predicted irradiance data and the historical measured irradiance data as inputs and the future predicted irradiance as output.

[0032] Preferably, the processing unit includes:

[0033] A first processing subunit is used to process each set of historical irradiance data to obtain an irradiance difference between the historical predicted irradiance data and the historical measured irradiance data, and to process the historical predicted irradiance data to obtain an observation matrix at the acquisition time.

[0034] A second processing subunit is used to obtain a state vector for the historical moment based on the irradiance difference and the historical measured irradiance data.

[0035] A third processing subunit, connected to the first processing subunit and the second processing subunit respectively, is used to obtain a revised difference based on the observation matrix, the state vector and a pre-set measurement noise, and to obtain the future predicted irradiance at the historical time predicted at the acquisition time based on the revised difference and the historical measured irradiance data.

[0036] The above technical solution has the following advantages or beneficial effects: This method and system establish a Kalman filter correction model by collecting historical irradiance data to correct the predicted irradiance obtained by the weather forecast model, thereby solving the problems of low forecast accuracy and poor prediction effect, which is of great significance for the stable operation of the power grid. Attached Figure Description

[0037] Figure 1 The following is a flowchart of the steps in a preferred embodiment of the present invention;

[0038] Figure 2 A flowchart illustrating the steps of the model training process in a preferred embodiment of the present invention;

[0039] Figure 3 The following is a flowchart of step A2 in a preferred embodiment of the present invention;

[0040] Figure 4 The following is a schematic diagram of the system in a preferred embodiment of the present invention. Detailed Implementation

[0041] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. The present invention is not limited to this embodiment; other embodiments that conform to the spirit of the present invention may also fall within the scope of the present invention.

[0042] In a preferred embodiment of the present invention, based on the above-mentioned problems existing in the prior art, a Kalman filter-corrected irradiance prediction method is provided, such as... Figure 1 As shown, the specific steps include:

[0043] Step S1: Collect a measured irradiance of a target area at the current time, and obtain meteorological data of the target area. Input the meteorological data into a weather forecast model to predict the predicted irradiance of the target area at least one future time.

[0044] Step S2: Input the predicted irradiance and the measured irradiance into a pre-trained Kalman filter correction model to obtain a corrected irradiance for the future time, which is used as the irradiance prediction result for the target area.

[0045] Specifically, in this embodiment, meteorological data is provided by the GFS (Global Forecast System), a weather forecasting model created by the U.S. National Center for Environmental Prediction. This model can obtain dozens of atmospheric and terrestrial soil variables, including temperature, wind, precipitation, soil moisture, and atmospheric ozone concentration. The system couples four separate models (atmospheric model, ocean model, land / soil model, and sea ice model) together to accurately describe weather conditions. The GFS model simulates the entire globe, with a grid point resolution of 0.25° to 0.5° (28 km to 56 km), and can predict the weather for the next 16 days.

[0046] Specifically, in this embodiment, the weather forecast model is the WRF (Weather Research and Forecasting Model), a general meteorological model developed by U.S. research institutions such as the U.S. Environmental Prediction Center and the U.S. National Center for Atmospheric Research. This model is a mesoscale numerical weather prediction system designed for atmospheric research and forecasting, and can be used for weather forecasting in the range of thousands of meters to thousands of kilometers.

[0047] Preferably, meteorological data is used as input, and a WRF model is run in the target area to obtain predicted irradiance.

[0048] In a preferred embodiment of the present invention, a model training process is included before performing step S1, such as... Figure 2 As shown, the specific steps include:

[0049] Step A1: Obtain multiple sets of historical irradiance data for multiple target areas. Each set of historical irradiance data includes a historical predicted irradiance data and a historical measured irradiance data. The historical predicted irradiance data is the predicted irradiance at a historical moment after the acquisition time, which is predicted at the acquisition time of the historical measured irradiance data.

[0050] Step A2: Based on historical predicted irradiance data and historical measured irradiance data, process the data to obtain a future predicted irradiance at the historical moment predicted at the time of data acquisition.

[0051] Step A3: Using historical predicted irradiance data and historical measured irradiance data as inputs, and future predicted irradiance as output, train the Kalman filter correction model.

[0052] In a preferred embodiment of the present invention, before step A2, the method further includes:

[0053] Based on an iterative optimization strategy and an isolated forest algorithm, the historical irradiance data of each group are preprocessed to obtain preprocessed predicted irradiance data and preprocessed measured irradiance data.

[0054] In step A2, the future predicted irradiance at the historical time predicted at the time of acquisition is obtained by processing the preprocessed predicted irradiance data and the preprocessed measured irradiance data.

[0055] Specifically, in this embodiment, in order to improve the parameter search speed of the Kalman filter correction model, the historical irradiance data is standardized. To ensure data quality, the EM algorithm (Expectation-Maximization Algorithm, an iterative optimization strategy) is used to fill in missing values. (The EM algorithm is an iterative optimization strategy. Its basic idea is: first, estimate the values ​​of the model parameters based on the given observation data; then, estimate the values ​​of the missing data based on the parameter values ​​estimated in the previous step; then, estimate the parameter values ​​again based on the estimated missing data plus the previously observed data; and then iterate repeatedly until convergence and the iteration ends.)

[0056] Preferably, in order to ensure the stability of the model, the IFOrest algorithm (Isolation Forest Algorithm) is used to process the data. The IFOrest algorithm is an algorithm for mining abnormal data. It is simple and efficient. The algorithm is based on two characteristics of abnormal data: (1) abnormal data accounts for only a small amount; (2) abnormal data features are very different from normal data. The abnormal data features are detected when they are very different from normal data. The KNN algorithm (K Nearest Neighbors Algorithm) is used to classify and correct the abnormal data by measuring the distance between different feature values.

[0057] In a preferred embodiment of the present invention, as shown in the figure, step A2 includes:

[0058] Step A21: For each set of historical irradiance data, process to obtain an irradiance difference between the historical predicted irradiance data and the historical measured irradiance data, and process according to the historical predicted irradiance data to obtain an observation matrix at the time of acquisition.

[0059] Step A22: Based on the irradiance difference and the historical measured irradiance data, a state vector for a historical moment is obtained;

[0060] Step A23: Based on the observation matrix, state vector and a pre-set measurement noise processing, a revised difference is obtained. Based on the revised difference and historical measured irradiance data, the future predicted irradiance at the historical time predicted at the acquisition time is obtained.

[0061] Specifically, in this embodiment, the observation matrix and state vector are obtained through the following relational expression:

[0062]

[0063] in,

[0064] y′ t Indicates the difference in irradiance;

[0065] r t This represents historical predicted irradiance data;

[0066] q t Indicates measurement noise;

[0067] x 0，t-1 x 1.t-1 x 2,t-1 x 3,t-1 This indicates a variable parameter.

[0068] Preferably, the observation matrix can be obtained through the relational expression. State vector x t-1 =[x 0，t-1 x 1.t-1 x 2,t-1 x 3,t-1 ] T .

[0069] Preferably, x in the above relational expression can be obtained by using the Kalman filter algorithm. 0，t-1 x 1.t-1 x 2,t-1 x 3,t-1 .

[0070] In a preferred embodiment of the present invention, the revised difference is obtained by processing the following calculation formula:

[0071] y t =H t x t-1 +q t

[0072] in,

[0073] y t Indicates the difference after revision;

[0074] H t Represents the observation matrix;

[0075] x t-1 Represents the state vector;

[0076] q t This indicates measurement noise.

[0077] Specifically, in this embodiment, when it is necessary to correct the predicted irradiance at time t, it is only necessary to input the measured irradiance at time t-1 and the predicted irradiance at time t into the Kalman filter correction model to obtain the corrected irradiance at time t.

[0078] Preferably, research has found that the difference between the state vectors of two adjacent time points is a system noise, and the system noise and measurement noise follow a Gaussian distribution.

[0079] In a preferred embodiment of the present invention, a Kalman filter-corrected irradiance prediction system, applied to the above-mentioned irradiance prediction method, includes:

[0080] A data acquisition module 1 is used to acquire a measured irradiance of a target area at the current time, obtain meteorological data of the target area, and input the meteorological data into a weather forecast model to predict the predicted irradiance of the target area at least one future time.

[0081] An irradiance correction module 2, connected to the data acquisition module 1, is used to input the predicted irradiance and the measured irradiance into a pre-trained Kalman filter correction model to obtain a corrected irradiance at a future time, which serves as the irradiance prediction result for the target area.

[0082] In a preferred embodiment of the present invention, a model training module 3 is further included, connected to the irradiance correction module 2, comprising:

[0083] A data acquisition unit 31 is used to acquire multiple sets of historical irradiance data for multiple target areas. Each set of historical irradiance data includes a historical predicted irradiance data and a historical measured irradiance data. The historical predicted irradiance data is the predicted irradiance at a historical moment after the acquisition time, which is predicted at the acquisition time of the historical measured irradiance data.

[0084] A processing unit 32, connected to the acquisition unit 31, is used to process historical predicted irradiance data and historical measured irradiance data to obtain a future predicted irradiance at the historical moment predicted at the acquisition time.

[0085] A training unit 33 is connected to a processing unit 32, which takes historical predicted irradiance data and historical measured irradiance data as inputs and future predicted irradiance as outputs to train a Kalman filter correction model.

[0086] In a preferred embodiment of the present invention, the processing unit 32 includes:

[0087] A first processing subunit 321 is used to process each set of historical irradiance data to obtain an irradiance difference between historical predicted irradiance data and historical measured irradiance data, and to process an observation matrix at the time of acquisition based on the historical predicted irradiance data.

[0088] A second processing subunit 322 is used to obtain a state vector at a historical moment based on the irradiance difference and the historical measured irradiance data.

[0089] A third processing subunit 323 is connected to the first processing subunit 321 and the second processing subunit 322 respectively. It is used to obtain a revised difference based on the observation matrix, the state vector and a pre-set measurement noise, and to obtain the future predicted irradiance at the historical time predicted at the acquisition time based on the revised difference and historical measured irradiance data.

[0090] The above description is merely a preferred embodiment of the present invention and does not limit the implementation and protection scope of the present invention. Those skilled in the art should realize that any equivalent substitutions and obvious changes made using the content of this specification and illustrations should be included within the protection scope of the present invention.

Claims

1. A Kalman filter corrected irradiance prediction method, characterized by, Specifically comprising the following steps: Step S1, collecting a measured irradiance of a target area at a current time, and obtaining meteorological data of the target area, and inputting the meteorological data into a weather forecast model to obtain a predicted irradiance of the target area at at least one future time; Step S2, inputting the predicted irradiance and the measured irradiance into a Kalman filter correction model trained in advance to obtain a corrected irradiance at the future time as the irradiance prediction result of the target area; Before performing the step S1, a model training process is included, specifically comprising the following steps: Step A1, obtaining a plurality of sets of historical irradiance data of a plurality of target areas, each set of the historical irradiance data comprising a historical predicted irradiance data and a historical measured irradiance data, the historical predicted irradiance data being a predicted irradiance at a historical time after the collection time of the historical measured irradiance data; Step A2, processing a future predicted irradiance at the historical time predicted at the collection time according to the historical predicted irradiance data and the historical measured irradiance data; Step A3, training the Kalman filter correction model by taking the historical predicted irradiance data and the historical measured irradiance data as input and taking the future predicted irradiance as output; The step A2 comprises: Step A21, for each set of the historical irradiance data, processing an irradiance difference between the historical predicted irradiance data and the historical measured irradiance data, and processing an observation matrix at the collection time according to the historical predicted irradiance data; Step A22, processing a state vector at the historical time according to the irradiance difference and the historical measured irradiance data; Step A23, processing a revised difference according to the observation matrix, the state vector and a preset measurement noise, and processing the future predicted irradiance at the historical time predicted at the collection time according to the revised difference and the historical measured irradiance data; The observation matrix and the state vector are obtained through the following relationship expression: ； Wherein, represents the irradiance difference; representative historical predicted irradiance data; denotes the measurement noise; represents a variable parameter; The observation matrix is expressed as: = [1, , , ]; The state vector is expressed as: .

2. The irradiance prediction method of claim 1, wherein Before the step A2, further comprising: According to an iterative optimization strategy and an isolation forest algorithm, pre-processing each set of the historical irradiance data to obtain pre-processed predicted irradiance data and pre-processed measured irradiance data; In the step A2, processing the future predicted irradiance at the historical time predicted at the collection time according to the pre-processed predicted irradiance data and the pre-processed measured irradiance data.

3. The irradiance prediction method of claim 1, wherein The revised difference is processed through the following calculation formula: ； Wherein, represents the revised difference; denotes the observation matrix; represents the state vector; represents the measurement noise.

4. A Kalman filter corrected irradiance forecasting system, characterized by, Applied to the irradiance prediction method in any one of claims 1-3, comprising: A data acquisition module for collecting a measured irradiance of a target area at a current time, and obtaining meteorological data of the target area, and inputting the meteorological data into a weather forecast model to obtain a predicted irradiance of the target area at at least one future time; An irradiance correction module connected to the data acquisition module, configured to input the predicted irradiance and the measured irradiance into a Kalman filter correction model trained in advance to obtain a corrected irradiance at the future time as the irradiance prediction result of the target region.

5. The irradiance forecasting system of claim 4, wherein, Further comprising a model training module connected to the irradiance correction module, comprising: An acquisition unit configured to acquire a plurality of sets of historical irradiance data of a plurality of target regions, each set of the historical irradiance data comprising a historical predicted irradiance data and a historical measured irradiance data, the historical predicted irradiance data being a predicted irradiance at a historical time after the acquisition time of the historical measured irradiance data; A processing unit connected to the acquisition unit and configured to obtain a future predicted irradiance at the historical time predicted at the acquisition time according to the historical predicted irradiance data and the historical measured irradiance data; A training unit connected to the processing unit and configured to train the Kalman filter correction model by taking the historical predicted irradiance data and the historical measured irradiance data as input and taking the future predicted irradiance as output.

6. The irradiance forecasting system of claim 5, wherein, The processing unit comprises: A first processing subunit configured to obtain, for each set of the historical irradiance data, an irradiance difference between the historical predicted irradiance data and the historical measured irradiance data, and obtain an observation matrix at the acquisition time according to the historical predicted irradiance data; A second processing subunit configured to obtain a state vector at the historical time according to the irradiance difference and the historical measured irradiance data; A third processing subunit respectively connected to the first processing subunit and the second processing subunit, configured to obtain a revised difference according to the observation matrix, the state vector and a preset measurement noise, and obtain the future predicted irradiance at the historical time predicted at the acquisition time according to the revised difference and the historical measured irradiance data.

Images