Block weather forecast method, system and device and storage medium

Abstract

The invention relates to a block weather forecast method, system and device and a storage medium. The method comprises the following steps: collecting multi-source meteorological data, preprocessing the multi-source meteorological data, and constructing a meteorological time sequence; segmenting the meteorological time sequence into a plurality of blocks through a block division algorithm; constructing a local weather prediction model, and predicting a weather forecast result corresponding to each block through the local weather prediction model; performing weighted summation on the prediction result of each block to obtain a final weather forecast result; according to the method, a block type weather forecast concept is introduced, the meteorological time sequence is adaptively divided into a plurality of blocks, an independent or differentiated local prediction model is established for each block, and dynamic and steady fusion prediction is realized through multi-factor weighting, so that a refined forecast effect is improved under complex and changeable weather conditions.

Description

Technical Field

[0001] The invention relates to a block weather forecast method, system, equipment and storage medium, belonging to the technical field of meteorological forecasting. Background Art

[0002] Existing weather forecasts mainly rely on methods such as numerical weather prediction (NWP) and time series modeling, but they often face problems of insufficient accuracy and adaptability: although NWP can simulate large-scale dynamic processes, it has high computational costs and uncertainties in the refined forecasts of local or extreme weather; time series methods based on linear assumptions are often difficult to effectively capture sudden and nonlinear meteorological changes due to strict stationarity requirements and neglect of multi-scale characteristics.

[0003] As the amount of meteorological data continues to grow, machine learning and deep learning (such as LSTM, Transformer, etc.) can capture nonlinear relationships to a certain extent, but they are highly dependent on the scale of training data and hyperparameters, and lack interpretability, and are still unable to adapt to local mutations or extreme weather.

[0004] On the other hand, it is difficult to take into account the meteorological differences in different seasons and regions by uniformly constructing a single model, which leads to large fluctuations in prediction accuracy during non-stationary phases or extreme events. The fusion method between models is mostly simple weighting or voting, lacking a detailed consideration of the advantages and disadvantages of different models in different time periods.

[0005] Finally, as weather systems continue to evolve, traditional models find it difficult to update their structures and parameters in a timely manner, and lack the ability to adapt to emerging patterns or emergencies. Summary of the invention

[0006] In order to solve the above problems existing in the prior art, the present invention proposes a block weather forecast method, system, device and storage medium.

[0007] The technical solution of the present invention is as follows:

[0008] In one aspect, the present invention provides a regionalized weather forecasting method, comprising the following steps:

[0009] Collect multi-source meteorological data and construct meteorological time series after preprocessing the multi-source meteorological data;

[0010] The meteorological time series is divided into several blocks through the block partitioning algorithm;

[0011] Build a local meteorological prediction model and use it to predict the weather forecast results for each block;

[0012] The final weather forecast result is obtained by weighted summing of the prediction results of each block.

[0013] As a preferred embodiment of the present invention, the specific steps of dividing the meteorological time series into several blocks by using the block partitioning algorithm are:

[0014] Assume that the length of the meteorological time series is N, as shown in the following formula:

[0015] X={x1,x2,…,x N}

[0016] The meteorological time series is divided into M blocks in advance, and a block cost function is constructed to calculate the cost of the current block scheme, as shown in the following formula:

[0017]

[0018] Where: P represents the current block scheme; J represents the block cost function; S m represents the mth block; σ(S m ) represents the sample data variance of the mth block; Δt(S m ) represents the length of the mth block; γ represents a hyperparameter; Ω(M) represents the regularization for limiting the number of blocks; λ represents the restriction coefficient;

[0019] When the J(P) value reaches the preset threshold, the optimal partitioning scheme is obtained, and the meteorological time series is partitioned by the optimal partitioning scheme.

[0020] As a preferred embodiment of the present invention, the local meteorological forecast model is constructed based on an autoregressive model, as shown in the following formula:

[0021]

[0022] in: Represents the prediction result of the mth block at time t+1; represents the autoregressive coefficient of the mth block; p m represents the autoregressive order of the mth block; z t represents the exogenous meteorological factor at time t; Θ m represents the weight of the exogenous meteorological factor of the mth block; W represents the nonlinear correction term based on wavelet frequency domain information; ω m represents the wavelet parameters of the mth block; represents the fitting residual of the mth block at time t.

[0023] As a preferred embodiment of the present invention, the steps for calculating the weight of each block are:

[0024] Calculate the cumulative error of each block;

[0025] Analyze the main frequency components of each block and the main frequency components of the global block and perform global correlation analysis;

[0026] Construct frequency domain similarity factors based on the global correlation analysis results of each block;

[0027] The weight of each block is calculated based on the accumulated error and frequency domain similarity factor, as shown in the following formula:

[0028]

[0029] Where: w m (t) represents the weight of the mth block at time t; E m (t) represents the cumulative error of the mth block up to time t; SP m (t) represents the frequency domain similarity factor of the mth block at time t;

[0030] The weather forecast result at the corresponding time is obtained by weighted summing of all blocks.

[0031] As a preferred embodiment of the present invention, the calculation steps of the accumulated error of each block are:

[0032] Calculate the instantaneous error of each block as shown below:

[0033]

[0034] Where: e m (t) represents the instantaneous error of the mth block at time t; x t Represents the actual value of the mth block at time t;

[0035] The accumulated error of each block is updated through exponential smoothing, as shown in the following formula:

[0036] E m (t) = ρ·e m (t)+(1-ρ)E m (t-1)

[0037] Where: E m (t) represents the cumulative error of the mth block up to time t; E m (t-1) represents the cumulative error of the mth block accumulated to time t-1; ρ represents the smoothing coefficient.

[0038] As a preferred embodiment of the present invention, the main frequency component of each block and the main frequency component of the global block are obtained by short-time Fourier transform analysis and global correlation analysis is performed, as shown in the following formula:

[0039]

[0040] Where: C m (t) represents the global correlation analysis of the mth block at time t; corr represents the correlation analysis function; represents the main frequency component of the mth block at time t; Represents the main frequency component of the global block at time t;

[0041] The frequency domain similarity factor is constructed based on the global correlation analysis results of each block, as shown in the following formula:

[0042]

[0043] Among them: SP m (t) represents the frequency domain similarity factor of the mth block at time t; v represents the similarity sensitivity factor.

[0044] On the other hand, the present invention also provides a block weather forecast system, including a data acquisition module, a block division module, a local weather forecast module and an overall weather forecast module;

[0045] The data acquisition module is used to collect multi-source meteorological data and construct a meteorological time series after pre-processing the multi-source meteorological data;

[0046] The block division module is used to divide the meteorological time series into a number of blocks through a block division algorithm;

[0047] The local meteorological prediction module is used to construct a local meteorological prediction model, and predict the weather forecast result corresponding to each block through the local meteorological prediction model;

[0048] The overall weather forecast module is used to perform weighted summation of the forecast results of each block to obtain the final weather forecast result.

[0049] On the other hand, the present invention further provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in any embodiment of the present invention when executing the program.

[0050] In yet another aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in any embodiment of the present invention.

[0051] The present invention has the following beneficial effects:

[0052] 1. The present invention automatically obtains a reasonable number of blocks and boundaries by optimizing the block cost function and combining the statistical characteristics within the block with the global regularization. It combines multiple information sources within the block to improve the ability to capture different meteorological patterns. At the same time, it considers historical errors and frequency domain similarity factors, so that both accuracy and pattern matching can be taken into account in the fusion stage, and the complementarity between blocks can be better utilized. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] It should be understood that the step numbers used in this document are only for convenience of description and are not intended to limit the order in which the steps are executed.

[0056] It should be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0057] The terms “include” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0058] The term "and / or" means and includes any and all possible combinations of one or more of the associated listed items.

[0059] Embodiment 1:

[0060] See also Figure 1 , a block weather forecasting method, comprising the following steps:

[0061] Collect multi-source meteorological data and construct meteorological time series after preprocessing the multi-source meteorological data;

[0062] The meteorological time series is divided into several blocks through the block partitioning algorithm;

[0063] Build a local meteorological prediction model and use it to predict the weather forecast results for each block;

[0064] The final weather forecast result is obtained by weighted summing of the prediction results of each block.

[0065] As a preferred implementation mode of this embodiment, the multi-source meteorological data includes meteorological data such as temperature, humidity, precipitation, etc. collected by multiple sources such as meteorological stations, satellite remote sensing or sensor networks; the preprocessing steps include missing value filling, outlier removal, interpolation smoothing, etc.

[0066] As a preferred implementation of this embodiment, the specific steps of dividing the meteorological time series into several blocks by using the block partitioning algorithm are:

[0067] Assume that the length of the meteorological time series is N, as shown in the following formula:

[0068] X={x1,x2,…,x N}

[0069] The meteorological time series is divided into M blocks in advance, and a block cost function is constructed to calculate the cost of the current block scheme, as shown in the following formula:

[0070]

[0071] Where: P represents the current block scheme; J represents the block cost function; S m represents the mth block; σ(S m ) represents the sample data variance of the mth block, which is used to measure the internal complexity of the block; Δt(S m ) represents the length of the mth block (i.e., the time span); γ represents a hyperparameter used to balance the trade-off between block length and internal complexity of the block; Ω(M) represents a regularization that limits the number of blocks; λ represents a limiting coefficient that prevents the division of too many or too few blocks;

[0072] When the J(P) value reaches the preset threshold, the optimal partitioning scheme is obtained, and the meteorological time series is partitioned by the optimal partitioning scheme.

[0073] As a preferred implementation of this embodiment, the local meteorological forecast model is constructed based on an autoregressive model, as shown in the following formula:

[0074]

[0075] in: Represents the prediction result of the mth block at time t+1; represents the autoregressive coefficient of the mth block; p m represents the autoregressive order of the mth block; zt Represents the external meteorological factors at time t (such as wind direction, wind speed, sea level pressure, etc.); Θ m represents the weight of the exogenous meteorological factor of the mth block; W represents the nonlinear correction term based on the wavelet frequency domain information. For example, the data in the block can be transformed into multi-scale discrete wavelet transform to obtain wavelet coefficients of several scales, which are then mapped to the predicted value through wavelet parameters (which can be regarded as a set of network or fitting coefficients); ω m represents the wavelet parameters of the mth block; represents the fitting residual of the mth block at time t;

[0076] Assume a wavelet component of a certain scale as the correction value ω t , then W(ω m ) is calculated as:

[0077]

[0078] in: It represents the frequency domain offset, which can be understood as a correction of "adding / subtracting a fixed amount to the meteorological quantity in a specific frequency band"; Indicates frequency domain amplification, which is used to adjust the influence of the high-frequency (or specific frequency band) component on the final prediction. Indicates amplifying the frequency domain component; Indicates attenuation of the frequency domain component; Indicates a correction direction opposite to the main model, and may also be used to capture "reverse oscillations".

[0079] As a preferred implementation of this embodiment, the steps for calculating the weight of each block are:

[0080] Calculate the cumulative error of each block;

[0081] Analyze the main frequency components of each block and the main frequency components of the global block and perform global correlation analysis;

[0082] The frequency domain similarity factor is constructed based on the global correlation analysis results of each block;

[0083] The weight of each block is calculated based on the accumulated error and frequency domain similarity factor, as shown in the following formula:

[0084]

[0085] Where: w m (t) represents the weight of the mth block at time t; E m (t) represents the accumulated error of the mth block up to time t; SP m (t) represents the frequency domain similarity factor of the mth block at time t;

[0086] The weather forecast result at the corresponding time is obtained by weighted summing of all blocks.

[0087] As a preferred implementation of this embodiment, the calculation steps of the accumulated error of each block are:

[0088] Calculate the instantaneous error of each block as shown below:

[0089]

[0090] Where: e m (t) represents the instantaneous error of the mth block at time t; x t Represents the actual value of the mth block at time t;

[0091] The accumulated error of each block is updated through exponential smoothing, as shown in the following formula:

[0092] E m (t) = ρ·e m (t)+(1-ρ)E m (t-1)

[0093] Where: E m (t) represents the cumulative error of the mth block up to time t; E m (t-1) represents the cumulative error of the mth block accumulated to time t-1; ρ represents the smoothing coefficient, ρ∈(0,1).

[0094] As a preferred implementation of this embodiment, the main frequency components of each block and the main frequency components of the global block are obtained by short-time Fourier transform analysis and global correlation analysis is performed, as shown in the following formula:

[0095]

[0096] Where: C m (t) represents the global correlation analysis of the mth block at time t; corr represents the correlation analysis function; represents the main frequency component of the mth block at time t; Represents the main frequency component of the global block at time t;

[0097]

[0098] in: represents the mean of the main frequency components of the mth block at all times; Represents the mean of the main frequency components of the global block at all times;

[0099] The frequency domain similarity factor is constructed based on the global correlation analysis results of each block, as shown in the following formula:

[0100]

[0101] Among them: SP m (t) represents the frequency domain similarity factor of the mth block at time t; v represents the similarity sensitivity factor.

[0102] Embodiment 2:

[0103] A block weather forecast system, comprising a data acquisition module, a block division module, a local weather forecast module and an overall weather forecast module;

[0104] The data acquisition module is used to collect multi-source meteorological data and construct a meteorological time series after pre-processing the multi-source meteorological data;

[0105] The block division module is used to divide the meteorological time series into a number of blocks through a block division algorithm;

[0106] The local meteorological prediction module is used to construct a local meteorological prediction model, and predict the weather forecast result corresponding to each block through the local meteorological prediction model;

[0107] The overall weather forecast module is used to perform weighted summation of the forecast results of each block to obtain the final weather forecast result.

[0108] This system is used to implement the method of embodiment 1, which will not be described in detail here.

[0109] Embodiment three:

[0110] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in any embodiment of the present invention when executing the program.

[0111] Embodiment 4:

[0112] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the method described in any embodiment of the present invention is implemented.

[0113] In the embodiments of the present application, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent the existence of A alone, the existence of A and B at the same time, and the existence of B alone. Among them, A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can be represented by: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, c can be single or multiple.

[0114] Those of ordinary skill in the art will appreciate that the various units and algorithm steps described in the embodiments disclosed herein can be implemented in a combination of electronic hardware, computer software, and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0115] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0116] In several embodiments provided in the present application, if any function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory; hereinafter referred to as: ROM), random access memory (Random Access Memory; hereinafter referred to as: RAM), disk or optical disk, and other media that can store program codes.

[0117] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A block weather forecast method, characterized in that: The following steps are involved: Collect multi-source meteorological data and construct meteorological time series after preprocessing the multi-source meteorological data; The meteorological time series is divided into several blocks through the block partitioning algorithm; Build a local meteorological prediction model and use it to predict the weather forecast results for each block; The final weather forecast result is obtained by weighted summing of the prediction results of each block.

2. A regionalized weather forecast method according to claim 1, characterized in that: The specific steps of dividing the meteorological time series into several blocks by the block partitioning algorithm are: Assume that the length of the meteorological time series is N, as shown in the following formula: X={x1,x2,…,x N } The meteorological time series is divided into M blocks in advance, and a block cost function is constructed to calculate the cost of the current block scheme, as shown in the following formula: Where: P represents the current block scheme; J represents the block cost function; S m represents the mth block; σ(S m ) represents the sample data variance of the mth block; Δt(S m ) represents the length of the mth block; γ represents a hyperparameter; Ω(M) represents the regularization for limiting the number of blocks; λ represents the restriction coefficient; When the J(P) value reaches the preset threshold, the optimal partitioning scheme is obtained, and the meteorological time series is partitioned by the optimal partitioning scheme.

3. A regionalized weather forecast method according to claim 1, characterized in that: The local meteorological forecast model is constructed based on the autoregressive model, as shown in the following formula: in: Represents the prediction result of the mth block at time t+1; represents the autoregressive coefficient of the mth block; p m represents the autoregressive order of the mth block; z t represents the exogenous meteorological factor at time t; Θ m represents the weight of the exogenous meteorological factor of the mth block; W represents the nonlinear correction term based on wavelet frequency domain information; ω m represents the wavelet parameters of the mth block; represents the fitting residual of the mth block at time t.

4. A regionalized weather forecasting method according to claim 3, characterized in that: The calculation steps of the weight of each block are: Calculate the cumulative error of each block; Analyze the main frequency components of each block and the main frequency components of the global block and perform global correlation analysis; Construct frequency domain similarity factors based on the global correlation analysis results of each block; The weight of each block is calculated based on the accumulated error and frequency domain similarity factor, as shown in the following formula: Where: w m (t) represents the weight of the mth block at time t; E m (t) represents the cumulative error of the mth block up to time t; SP m (t) represents the frequency domain similarity factor of the mth block at time t; The weather forecast result at the corresponding time is obtained by weighted summing of all blocks.

5. A regionalized weather forecast method according to claim 4, characterized in that: The calculation steps of the cumulative error of each block are: Calculate the instantaneous error of each block as shown below: Where: e m (t) represents the instantaneous error of the mth block at time t; x t Represents the actual value of the mth block at time t; The accumulated error of each block is updated through exponential smoothing, as shown in the following formula: E m (t)=ρ·e m (t)+(1-ρ)E m (t-1) Where: E m (t) represents the cumulative error of the mth block up to time t; E m (t-1) represents the cumulative error of the mth block accumulated to time t-1; ρ represents the smoothing coefficient.

6. A regionalized weather forecast method according to claim 4, characterized in that: The main frequency components of each block and the main frequency components of the global block are obtained through short-time Fourier transform analysis, and global correlation analysis is performed, as shown in the following formula: Where: C m (t) represents the global correlation analysis of the mth block at time t; corr represents the correlation analysis function; represents the main frequency component of the mth block at time t; Represents the main frequency component of the global block at time t; The frequency domain similarity factor is constructed based on the global correlation analysis results of each block, as shown in the following formula: Among them: SP m (t) represents the frequency domain similarity factor of the mth block at time t; v represents the similarity sensitivity factor.

7. A block weather forecast system, characterized in that: It includes data collection module, block division module, local weather forecast module and overall weather forecast module; The data acquisition module is used to collect multi-source meteorological data and construct a meteorological time series after pre-processing the multi-source meteorological data; The block division module is used to divide the meteorological time series into a number of blocks through a block division algorithm; The local meteorological prediction module is used to construct a local meteorological prediction model, and predict the weather forecast result corresponding to each block through the local meteorological prediction model; The overall weather forecast module is used to perform weighted summation of the forecast results of each block to obtain the final weather forecast result.

8. An electronic 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 program, the method according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images