A wind power error correction method based on AMS-TCN
The AMS-TCN method is used to decompose and correct the prediction error of the wind power sequence, which solves the error problem in wind power prediction, improves the prediction accuracy and stability, and achieves more accurate wind power prediction.
Patent Information
- Application Number
- CN202111055074.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-09-09
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2041-09-09
AI Technical Summary
The existing wind power power prediction technology has prediction error problems, resulting in insufficient prediction accuracy and stability, failure to effectively analyze and correct prediction errors, affecting wind power consumption capacity and power system safety.
Using the wind power error correction method based on AMS-TCN, the original wind power power sequence is obtained for pre-processing and normalization, and the principal component sequence is decomposed to extract the time and spatial characteristics, and the time characteristics of the prediction error are extracted in combination with the TCN model. The preliminary prediction is corrected to form the final prediction result.
Effectively reduce the impact of wind power power noise, improve prediction accuracy and stability, and form more accurate and reliable wind power power prediction results.
Smart Images

Figure CN113947237B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a wind power error correction method based on AMS-TCN, belonging to the technical field of wind power prediction. Background Art
[0002] The continuous maturity of wind power generation technology has led to an increase in wind power generation capacity. However, due to the volatility and intermittency of wind farm output power, large-scale access to wind power will have a great impact on the power balance of the entire network. Wind power forecasting is an important means to grasp the randomness of wind power, which directly affects the effectiveness of the dispatch plan. Accurate wind power forecasting results help to avoid the adverse effects of random fluctuations in wind power and improve wind power absorption capacity. Due to the limitations of existing forecasting technology, forecasting errors are inevitable. Analyzing wind power forecasting errors and their distribution characteristics plays an important role in improving wind power absorption capacity and ensuring safe and reliable operation of power systems.
[0003] The wind power prediction methods widely used at present are mainly divided into statistical methods and physical methods. In addition, there are combined prediction methods that combine multiple methods. Among them, the physical method is a wind power prediction method based solely on wind power historical data and numerical weather forecast (NWP) data. The advantages of wind power prediction based on physical methods are that it relies on a small amount of data, a simple model, and convenient and fast prediction; but there is a large error between NWP data and real meteorological data, and there is a spatial difference between NWP data and wind farms, which makes the prediction accuracy of wind power by physical methods generally low. The statistical method is a wind power prediction method that updates and adjusts the model weights and parameters based on the error between the predicted value and the true value. This method is usually divided into probabilistic statistical models, machine learning models, and deep learning models. Probabilistic statistical models can better follow the changing trend of wind power than physical methods; the prediction accuracy of machine learning models is better than that of traditional probabilistic models; deep learning models can fully explore the time and space characteristics of the input sequence and further improve the accuracy of wind power prediction. Statistical methods can effectively extract the linear and nonlinear relationships of historical data and improve the accuracy of wind power prediction, but they have high requirements on the consistency of the changing patterns of historical data; at the same time, the modeling process of statistical methods has a "black box" nature; like the physical prediction method, there is no further analysis of the prediction error, and there is a problem of poor prediction stability.
[0004] The combined prediction method is a wind power prediction method with higher prediction accuracy and better stability formed by combining the advantages of each single prediction model. The combined model has better prediction performance than the single model, but like the single model, it does not consider the re-analysis of the prediction error, resulting in low prediction accuracy and unstable prediction.
[0005] In view of the above, existing research has the problem of prediction error in wind power prediction, and lacks analysis and correction of prediction error when performing power prediction. Summary of the invention
[0006] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a wind power error correction method based on AMS-TCN, which can reduce the wind power prediction error and improve the prediction accuracy and stability.
[0007] To achieve the above object, the present invention is implemented by adopting the following technical solutions:
[0008] In a first aspect, the present invention provides a wind power error correction method based on AMS-TCN, comprising:
[0009] Obtaining the original wind power sequence;
[0010] Preprocessing and normalization are performed based on the original wind power sequence;
[0011] Decompose the processed wind power sequence to obtain the principal component sequence;
[0012] Extract the time and space characteristics of the principal component sequence to obtain the preliminary prediction power;
[0013] The time characteristics of the prediction error are extracted, and the preliminary predicted power is corrected to obtain the final prediction result.
[0014] Further, the preprocessing includes:
[0015] The negative values are processed by directly setting them to zero, and the wind power values at the two moments before and after are used to correct the over-full power values and garbled codes, and fill in the missing values.
[0016] The normalization expression is:
[0017]
[0018] Where P i ″ is the wind power at time i after normalization, P i ' is the wind power at time i before normalization, P' is the wind power sequence after preprocessing, max and min refer to the maximum and minimum values respectively.
[0019] Furthermore, the processed wind power sequence is decomposed to obtain the main component sequence, including:
[0020] Perform constrained variation on historical wind power signals;
[0021] By introducing Lagrange multipliers and quadratic penalty functions, the constrained variational results are transformed into unconstrained variational problems.
[0022] The alternating direction multiplier iterative algorithm is combined with Parseval, Plancherel theorems and Fourier isometric transform to optimize the unconstrained variational problem and obtain the modal components and center frequencies.
[0023] Furthermore, the constraint condition of the constrained variation is that the sum of all modes is equal to the original signal, the decomposition sequence is a modal component with a finite bandwidth and a center frequency, and the sum of the estimated bandwidths of each mode is minimal. The constrained variation expression is:
[0024]
[0025] Where u m ={u1,u2,...,u m} is the modal components of the decomposed wind power signal, is the square expression of the two-norm, δ(t) is the impulse function, is the first-order partial derivative with respect to time t, f is the historical wind power signal, m is the number of components, is the Fourier transform basis, and J is the imaginary part.
[0026] Furthermore, the unconstrained variational problem is expressed as:
[0027]
[0028] In the formula, u m (t)={u1(t), u2(t),. . . ,u m (t)} is the modal components of the decomposed wind power signal at time t, u m = {u1, u2, ..., u m} is the center frequency of each modal component, is the square expression of the two-norm, δ(t) is the impulse function, is the first-order partial derivative with respect to time t, f(t) is the historical wind power signal, m is the number of components, α is the introduced quadratic penalty factor parameter, and λ(t) is the Lagrange multiplier at time t.
[0029] Furthermore, the expressions of the modal components and center frequencies are as follows:
[0030]
[0031]
[0032]
[0033] Where τ is the noise tolerance, which satisfies the fidelity of wind power signal decomposition. and They are u i The Fourier transform of (t), f(t) and λ(t), is the L2 distance between the current frequency and the center frequency obtained in the nth iteration, α is the introduced quadratic penalty factor parameter, is the center frequency of each modal component after the n+1th iteration, is the Lagrange multiplier after the n+1th iteration, f(t) is the historical wind power signal, u i (t) is the i-th modal component at time t, are the modal components after the n+1th iteration, and λ(t) is the Lagrange multiplier at time t.
[0034] Furthermore, the temporal and spatial characteristics of the principal component sequence are extracted to obtain a preliminary predicted power, including: the input sequence is decomposed into an IMFs sequence, the temporal and spatial characteristics between the input and output are mined through stacked dilated convolution, and the wind power value at the next moment is output, where the IMF at time t is t The dilated convolution is:
[0035]
[0036] Where S = {s1 s2,...,s T} is the input sequence, s i , i∈[1,T] is a column vector, d represents the expansion factor, K represents the filter size, and the filter is F={f1,f2,…,f K}, * is the convolution symbol, s t-d(K-k) : The value range of the td(Kk)th input sequence vector si, i is [tK, t).
[0037] Furthermore, the temporal characteristics of the prediction error are extracted, including:
[0038] The mean μ of the sequence in the moving window e and variance It is expressed as:
[0039]
[0040]
[0041] Wherein, the noise sequence E = {e1, e2, ..., e T}, T is the length of the sequence, N is the length of the moving window;
[0042] Using μ e and The smoothed noise signal at time t is expressed as:
[0043]
[0044] Among them, f(e i ,μ e ,σ) is the sum of the error values of the points that meet the 3σ principle within the sliding window range, flag(e i , μ e ,σ) is the sum of the number of points that meet the 3σ principle within the sliding window range. The specific expression is:
[0045]
[0046]
[0047] In a second aspect, a wind power error correction device based on AMS-TCN includes a processor and a storage medium;
[0048] The storage medium is used to store instructions;
[0049] The processor is used to operate according to the instructions to execute the steps of any of the methods described above.
[0050] In a third aspect, a computer-readable storage medium stores a computer program, which implements the steps of any of the above methods when executed by a processor.
[0051] Compared with the prior art, the present invention has the following beneficial effects:
[0052] The wind power error correction method based on AMS-TCN provided by the present invention can decompose and obtain a stable signal component with a symmetrical center frequency that is more conducive to prediction, can effectively reduce the influence of the original wind power noise on the prediction, and improve the prediction accuracy; it can effectively extract the time characteristics of the prediction error, and the formed combined prediction model can effectively improve the prediction accuracy and stability. BRIEF DESCRIPTION OF THE DRAWINGS
[0053] Figure 1 is an algorithm flow chart of an AMS-TCN wind power error correction model provided in the first embodiment of the present invention;
[0054] Figure 2 It is a residual module diagram of the TCN model provided in Example 1 of the present invention. DETAILED DESCRIPTION
[0055] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.
[0056] Embodiment 1:
[0057] A wind power error correction method based on AMS-TCN consists of the following steps in its specific implementation process:
[0058] Step 1: Input the original wind power sequence and perform preprocessing and normalization;
[0059] Step 2: Use VMD module to decompose the wind power sequence to obtain the principal component sequence;
[0060] Step 3: Use the TCN model to extract the time and space characteristics of the principal component sequence to obtain the preliminary prediction power;
[0061] Step 4: Use the AMS-TCN module to extract the time characteristics of the prediction error, correct the preliminary prediction, and obtain the final prediction result.
[0062] More specifically, the preprocessing of the input original wind power sequence proposed in step 1 further includes the following steps:
[0063] Preprocessing includes abnormal value processing (negative values, values exceeding full load power, and garbled codes) and filling of vacant values. Negative values are processed by directly setting to zero, and the wind power values at the two moments before and after are used to correct or fill in the over-full load values, garbled codes, and vacant values. Considering that the different installed capacities of various wind power stations lead to differences in their power generation, the experimental data are normalized. The specific formula is:
[0064]
[0065] Where P i ″ is the wind power at time i after normalization, P i ' is the wind power at time i before normalization, P' is the wind power sequence after preprocessing, max and min refer to the maximum and minimum values respectively;
[0066] More specifically, the step 2 of decomposing the wind power sequence to obtain the principal component sequence also includes the following steps:
[0067] Assuming that the historical wind power signal P is decomposed into m components, the decomposition sequence is guaranteed to be a modal component with a limited bandwidth and a center frequency, and the sum of the estimated bandwidths of each mode is minimized. The constraint condition is that the sum of all modes is equal to the original signal. The corresponding constrained variational expression is:
[0068]
[0069] Where u m ={u1,u2,...,u m} is the modal components of the decomposed wind power signal, is the square expression of the two-norm, δ(t) is the impulse function, is the first-order partial derivative with respect to time t, f is the historical wind power signal, m is the number of components, is the Fourier transform basis, and J is the imaginary part.
[0070] By introducing the Lagrange multiplier λ(t) and the quadratic penalty function α, it is converted into an unconstrained variational problem. The specific formula is as follows:
[0071]
[0072] In the formula, u m (t)={u1(t), u2(t),. . . ,u m (t)} is the modal components of the decomposed wind power signal at time t, u m = {u1, u2, ..., u m} is the center frequency of each modal component, is the square expression of the two-norm, δ(t) is the impulse function, is the first-order partial derivative with respect to time t, f(t) is the historical wind power signal, m is the number of components, α is the introduced quadratic penalty factor parameter, and λ(t) is the Lagrange multiplier at time t.
[0073] Then, the alternating direction multiplier method is used to iterate the Parseval, Plancherel theorem and Fourier isometric transform to optimize the modal components and center frequencies. m ,ω m The expressions for and λ are as follows:
[0074]
[0075]
[0076]
[0077] Where τ is the noise tolerance, which satisfies the fidelity of wind power signal decomposition. and They are u i The Fourier transform of (t), f(t) and λ(t), is the L2 distance between the current frequency and the center frequency obtained in the nth iteration, α is the introduced quadratic penalty factor parameter, is the center frequency of each modal component after the n+1th iteration, is the Lagrange multiplier after the n+1th iteration, f(t) is the historical wind power signal, u i (t) is the i-th modal component at time t, are the modal components after the n+1th iteration, and λ(t) is the Lagrange multiplier at time t.
[0078] More specifically, the proposed method of using the TCN model to extract the temporal and spatial characteristics of the principal component sequence in step 3 also includes: decomposing the input sequence into a decomposed IMFs sequence, mining the temporal and spatial characteristics between the input and output through stacked dilated convolutions, and outputting the wind power value at the next moment.
[0079] Let filter F = (f1, f2, ..., f K ), the output sequence information is Y = (y1, y2, ..., y T ), where the input is a stationary sequence with symmetric center frequency obtained after VMD decomposition: IMFs = {IMF1, IMF2, ..., IMF T}, where IMF i , i∈[1,T] is a column vector; the following formula gives the IMF at time t t Atrous convolution:
[0080]
[0081] Where S = {s1 s2,...,s T} is the input sequence, s i , i∈[1,T] is a column vector, d represents the expansion factor, K represents the filter size, and the filter is F={f1,f2,…,f K}, * is the convolution symbol, s t-d(K-k) : The value range of the td(Kk)th input sequence vector si, i is [tK, t). 。
[0082] like Figure 2 The residual module of the TCN model is given. The TCN model introduces the residual module to solve the problems of gradient explosion and network degradation in deep traditional neural networks. This residual structure can avoid losing too much information in the feature extraction process, retain information as much as possible, and improve the accuracy of the model.
[0083] More specifically, the method of using the AMS module to smooth the prediction error in step 4 also includes the following steps:
[0084] Assume that the noise sequence E = {e1, e2, ..., e T}, where T is the length of the sequence and the length of the moving window is N, then the mean μ of the sequence in the moving window is e and variance It can be expressed as:
[0085]
[0086]
[0087] Using μ e and The smoothed noise signal at time t is expressed as:
[0088]
[0089] Among them, f(e i ,μ e ,σ) is the sum of the error values of the points that meet the 3σ principle within the sliding window range, flag(e i , μ e ,σ) is the sum of the number of points that meet the 3σ principle within the sliding window range. The specific expression is:
[0090]
[0091]
[0092] The method of the present invention uses the measured data of the Liaoning Shuangzitai Wind Farm and the Inner Mongolia Keqi Wind Farm from December 2019 to October 2020, with a sampling interval of 15 minutes. There are more than 10,000 data in total, 80% of which are used as training data sets after shuffling, and 20% are used as test data sets.
[0093] In order to verify the superiority of the proposed TCN model in wind power forecasting, the power forecasting performance of LSTM, CNN-LSTM and TCN models at three different time scales (15min, 30min and 1h) is analyzed and compared.
[0094] Table 1 Prediction performance indicators of each single model
[0095]
[0096] It can be seen from Table 1 that the TCN model performs better than the other two models in terms of MAE and MRE indicators. The causal convolution kernel hole convolution introduced by the TCN model can give it a wider receptive field, which is more conducive to exploring the changing characteristics of wind power and improving the prediction accuracy of wind power.
[0097] In order to verify the effect of VMD model on improving the accuracy of wind power prediction, this paper uses VMD-LSTM, VMD-CNN-LSTM and VMD-TCN models to conduct comparative experiments with the three single models in 3.4. VMD-LSTM, VMD-CNN-LSTM and VMD-TCN are used as Model 1, and LSTM, CNN-LSTM and TCN models are used as Model 2.
[0098] Table 2 Prediction performance indicators of VMD combination model
[0099]
[0100]
[0101] By analyzing the experimental results obtained in Table 2, it can be seen that adding VMD decomposition can effectively reduce the MAE and MRE of the prediction results of LSTM, CNN-LSTM and TCN models at three different time scales. Compared with the single prediction model without VMD, the combined prediction model with VMD has a MAE improvement of more than 40.0%, and the MRE improvement is generally maintained at more than 15.0%. The huge improvement of VMD in prediction performance mainly depends on its ability to decompose the stable signal components with symmetrical center frequency that are more conducive to prediction, which can effectively reduce the impact of the original wind power noise on the prediction. In addition, the VMD-TCN model has better prediction performance than the VMD-LSTM and VMD-CNN-LSTM models. Under the two wind farms, the MAE means of the prediction results in three different time scales are 0.2750, 0.3891 and 0.7932, respectively, which are better than 0.8011, 1.2032 and 1.6029 of VMD-CNN-LSTM, and 0.4099, 0.5106 and 1.0325 of VMD-LSTM; the MRE means of the prediction results in three different time scales are 0.0852, 0.2991 and 0.6416, respectively, which are better than 0.2711, 0.9095 and 2.1076 of VMD-CNN-LSTM, and 0.6082, 0.4818 and 0.9238 of VMD-LSTM.
[0102] In order to verify the superiority of the proposed combined prediction model (VMD-TCN-AMS-TCN) in wind power prediction accuracy and stability, the VMD-TCN-AMS-TCN model was compared with other models, and the VMD-TCN-AMS-TCN model was used as Model 1, and the best performing VMD-TCN model was used as Model 2.
[0103] Table 3 Model prediction performance indicators
[0104]
[0105]
[0106] As shown in Table 3, after adding the AMS-TCN module to correct the prediction error, the MAE and MRE under the three time scales can be effectively reduced, and the improvement of MAE is more than 50.0%, and the improvement of MRE is more than 10.0%. The experimental results show that AMS-TCN can effectively extract the time characteristics of the prediction error, reduce the prediction error and increase the stability of the prediction.
[0107] The present invention uses variational mode decomposition (VMD) and TCN to extract the spatiotemporal characteristics of wind power and obtain preliminary prediction results; then the AMS model is used to adaptively smooth the prediction error sequence to reduce the volatility of the error; finally, the TCN model is used to extract the temporal characteristics of the prediction error, and the preliminary prediction results are corrected to improve the accuracy and stability of the prediction. The comparison of different prediction models verifies that the proposed VMD-TCN-AMS-TCN model effectively improves the accuracy and stability of the prediction.
[0108] The wind power error correction method based on AMS-TCN provided by the present invention, the VMD can decompose to obtain a stable signal component with a symmetrical center frequency that is more conducive to prediction, can effectively reduce the influence of the original wind power noise on the prediction, and improve the prediction accuracy; the established AMS-TCN model can effectively extract the time characteristics of the prediction error, and the formed combined prediction model (VMD-TCN-AMS-TCN) can effectively improve the prediction accuracy and stability.
[0109] Embodiment 2:
[0110] The embodiment of the present invention also provides a wind power error correction device based on AMS-TCN, including a processor and a storage medium;
[0111] The storage medium is used to store instructions;
[0112] The processor is used to operate according to the instructions to perform the steps of the following method:
[0113] Obtaining the original wind power sequence;
[0114] Preprocessing and normalization are performed based on the original wind power sequence;
[0115] Decompose the processed wind power sequence to obtain the principal component sequence;
[0116] Extract the time and space characteristics of the principal component sequence to obtain the preliminary prediction power;
[0117] The time characteristics of the prediction error are extracted, and the preliminary predicted power is corrected to obtain the final prediction result.
[0118] Embodiment three:
[0119] The embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of the following method are implemented:
[0120] Obtaining the original wind power sequence;
[0121] Preprocessing and normalization are performed based on the original wind power sequence;
[0122] Decompose the processed wind power sequence to obtain the principal component sequence;
[0123] Extract the time and space characteristics of the principal component sequence to obtain the preliminary prediction power;
[0124] The time characteristics of the prediction error are extracted, and the preliminary predicted power is corrected to obtain the final prediction result.
[0125] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.
[0126] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0127] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0128] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0129] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.
Claims
1. A wind power error correction method based on AMS-TCN, characterized in that: include: Obtaining the original wind power sequence; Preprocessing and normalization are performed based on the original wind power sequence; Decompose the processed wind power sequence to obtain the principal component sequence; Extract the time and space characteristics of the principal component sequence to obtain the preliminary prediction power; Extract the time characteristics of the prediction error, correct the preliminary prediction power, and obtain the final prediction result; The processed wind power sequence is decomposed to obtain the main component sequence, including: Perform constrained variation on historical wind power signals; By introducing Lagrange multipliers and quadratic penalty functions, the constrained variational results are transformed into unconstrained variational problems. The alternating direction multiplier iterative algorithm is combined with Parseval, Plancherel theorems and Fourier isometric transform to optimize the unconstrained variational problem and obtain the modal components and center frequencies. The constraint conditions of the constrained variation are that the sum of all modes is equal to the original signal, the decomposition sequence is a modal component with a finite bandwidth and a center frequency, and the sum of the estimated bandwidths of each mode is minimal. The constrained variation expression is: Where u m = {u1, u2, ..., u m } is the modal components of the decomposed wind power signal, is the square expression of the two-norm, δ(t) is the impulse function, is the first-order partial derivative with respect to time t, f is the historical wind power signal, m is the number of components, is the Fourier transform basis, J is the imaginary part; The unconstrained variational problem expression is: In the formula, u m (t)={u1(t),u2(t),…,u m (t)} is the modal components of the decomposed wind power signal at time t, f(t) is the historical wind power signal at time t, α is the introduced quadratic penalty factor parameter, and λ(t) is the Lagrange multiplier at time t; The expressions of the modal components and center frequencies are as follows: Where τ is the noise tolerance, which satisfies the fidelity of wind power signal decomposition. and They are u i The Fourier transform of (t), f(t) and λ(t), is the L2 distance between the current frequency and the center frequency obtained in the nth iteration, is the center frequency of each modal component after the n+1th iteration, is the Lagrange multiplier after the n+1th iteration, u i (t) is the i-th modal component at time t, are the modal components after the n+1th iteration; Extract the temporal and spatial characteristics of the principal component sequence to obtain the preliminary predicted power, including: decomposing the input sequence into an IMFs sequence, mining the temporal and spatial characteristics between the input and output through stacked dilated convolution, and outputting the wind power value at the next moment, where the IMF at time t is t The dilated convolution is: Where S = {s1 s2, ..., s T } is the input sequence, d is the expansion factor, K is the filter size, and the filter is F = {f1, f2, ..., f K }, * is the convolution symbol, s t-d(K-k) Indicates the td(Kk)th input sequence vector si, the value range of i is [tK, t); Extract the temporal characteristics of the forecast error, including: The mean μ of the sequence in the moving window e and variance It is expressed as: Wherein, the noise sequence E = {e1, e2, ..., e T }, T is the length of the sequence, N is the length of the moving window; Using μ e and The smoothed noise signal at time t is expressed as: Among them, f(e i , μ e ,σ) is the sum of the error values of the points that meet the 3σ principle within the sliding window range, flag(e i , μ e ,σ) is the sum of the number of points that meet the 3σ principle within the sliding window range. The specific expression is:
2. The wind power error correction method based on AMS-TCN according to claim 1 is characterized in that: The pre-processing comprises: The negative values are processed by directly setting them to zero, and the wind power values at the two moments before and after are used to correct the over-full power values and garbled codes, and fill in the missing values. The normalization expression is: Where P i ″ is the wind power at time i after normalization, P i ' is the wind power at time i before normalization, P' is the wind power sequence after preprocessing, max and min refer to the maximum and minimum values respectively.
3. A wind power error correction device based on AMS-TCN, characterized in that: including processor and storage medium; The storage medium is used to store instructions; The processor is used to operate according to the instructions to execute the steps of the method according to any one of claims 1 to 2.
4. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 2 are implemented.
