Photovoltaic power short-term probability prediction system and method

By screening meteorological factors, clustering weather types, and improving the temporal Autoformer network, the optical power probability distribution parameters are optimized, solving the problem of insufficient utilization of multi-source meteorological information in existing technologies and realizing efficient short-term probability prediction of optical power.

CN119005750AActive Publication Date: 2024-11-22TIANJIN UNIV

Patent Information

Application Number
CN202411082717.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-08
Publication Date
2024-11-22
Estimated Expiration
2044-08-08

AI Technical Summary

Technical Problem

Existing short-term probability prediction methods for optical power fail to fully utilize multi-source meteorological information and have weak generalization ability, making it difficult to take into account both global and local features, resulting in insufficient prediction accuracy and generalization ability.

Method used

The Pearson correlation coefficient method was used to screen meteorological factors, and a convolutional neural network and an improved temporal Autoformer network were established. Weather type clustering was performed through a self-organizing map neural network, a probability density estimation module was constructed, and the optical power probability distribution parameters were optimized.

Benefits of technology

It significantly improves the accuracy of optical power prediction, reduces the risk of false alarms from a single meteorological source, improves information utilization efficiency, and reduces time complexity through maximum likelihood estimation, thus avoiding the neural network from getting trapped in local optima.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119005750A_ABST
    Figure CN119005750A_ABST
Patent Text Reader

Abstract

The invention discloses a photovoltaic power short-term probability prediction system and method. The system comprises a convolutional neural network, a sequential network and a probability density estimation module. The time sequence network is composed of a self-organizing mapping neural network, a time sequence encoder and a time sequence decoder; each of the time sequence encoder and the time sequence decoder is formed by alternately connecting an autocorrelation mechanism unit, a time sequence decomposition unit and a feed-forward layer; the time sequence decoder is composed of an upper branch sub-layer and a lower branch sub-layer; the convolutional neural network screens meteorological factors to obtain a high-dimensional mapping relation between meteorological characteristics and optical power; the sequential network performs progressive decomposition and autocorrelation extraction on an optical power sequential sequence to obtain periodicity and tendency characteristics of the optical power sequence; the probability density estimation module optimizes and combines the high-dimensional mapping relation with periodicity and tendency characteristics of the optical power sequence to obtain optical power probability distribution parameter output; according to the method, the numerical weather forecast information and the weather type forecast published by the meteorological observatory are integrated to realize optical power probability prediction, so that the risk of misinformation of a single meteorological source is effectively reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of photovoltaic prediction applications, and in particular relates to a photovoltaic power short-term probability prediction system and method. Background Art

[0002] In order to achieve the strategic goal of "dual carbon", the installed capacity and power generation of photovoltaic stations have increased significantly. Large-scale centralized photovoltaic stations account for an increasing proportion of newly installed photovoltaic capacity because they can achieve unified development and efficient management of light resources. However, the centralized photovoltaic grid connection has increased the uncertainty of the power system. The power grid needs to arrange conventional units to provide spare capacity to reduce its impact on safe and stable operation. Therefore, the accurate quantification of the uncertainty of centralized photovoltaic stations plays an important role in the economic and stable operation of the power system. Day-ahead forecasting is an important basis for power generation planning and power market transactions. Day-ahead forecasting can be divided into deterministic forecasting and probabilistic forecasting according to its forecasting form. Deterministic forecasting outputs the expected value of a single point at a certain moment in the future of the forecast object. Due to the chaotic characteristics of the atmospheric system, the forecast error is difficult to eliminate, and the forecast results often have high uncertainty. Probabilistic forecasting technology estimates the probability distribution of forecast uncertainty and realizes the quantitative analysis of forecast uncertainty, which can be better applied to power system optimization scheduling, market transactions and other aspects.

[0003] At present, many studies have been conducted at home and abroad on probabilistic prediction methods, including statistical methods and deep learning methods. Statistical methods require a relatively high amount of historical prediction data accumulation, and it is difficult to model the coupling effects of multiple factors. Recurrent neural networks and their variants in deep learning are widely used in power prediction, but the structure of the recurrent network severely limits the depth of the model and its ability to learn data. Based on the above two methods, domestic and foreign scholars have achieved rich research results in the study of short-term probabilistic prediction of optical power, but there are still defects in taking into account global and local features, breaking the bottleneck of information utilization, and improving the accuracy and generalization ability of the model. Summary of the invention

[0004] In view of the above problems, the purpose of the present invention is to overcome the shortcomings of the existing optical power short-term probability prediction method that fails to make full use of multi-source meteorological information and has weak generalization ability. Firstly, the Pearson correlation coefficient method is used to screen the meteorological factors related to the optical power, and a high-dimensional mapping relationship between meteorological characteristics and optical power is established based on a convolutional neural network; an improved time series Autoformer network is constructed to perform progressive decomposition and autocorrelation extraction on the optical power time series, and analyze its internal structure and inherent laws; finally, the hidden layer information output by the CNN-Autoformer network is input into the probability density estimation layer, and the optical power probability distribution parameters are obtained through the output of the fully connected layer.

[0005] In order to solve the problems of the prior art, the present invention adopts the following technical solutions:

[0006] A photovoltaic power short-term probability prediction system, characterized in that the prediction system comprises: a convolutional neural network, a timing network and a probability density estimation module; the timing network is composed of a self-organizing map neural network, a timing encoder and a timing decoder; the timing encoder and the timing decoder are both composed of an autocorrelation mechanism unit, a timing decomposition unit and a feedforward layer alternately connected; the timing decoder is composed of an upper branch sublayer and a lower branch sublayer; wherein:

[0007] The convolutional neural network screens meteorological factors to obtain a high-dimensional mapping relationship between meteorological characteristics and optical power;

[0008] The timing network performs progressive decomposition and autocorrelation extraction on the optical power timing sequence to obtain the periodicity and trend characteristics of the optical power sequence; wherein:

[0009] The self-organizing map neural network clusters the weather forecast types to obtain generalized weather types, and encodes them in a time series encoder as discrete features of the optical power daily series;

[0010] The timing encoder alternately decomposes the optical power daily sequence to obtain periodic information;

[0011] The timing decoder refines the periodic information through an autocorrelation mechanism unit; comprising:

[0012] The upper branch sub-layer extracts periodic features, and the lower branch sub-layer uses weighted addition to extract trend features.

[0013] The probability density estimation module optimizes and combines the high-dimensional mapping relationship with the periodicity and trend characteristics of the optical power sequence to obtain the optical power probability distribution parameter output.

[0014] Furthermore, the timing network performs progressive decomposition and autocorrelation extraction on the optical power timing sequence to obtain the periodicity and trend characteristics of the optical power sequence; including:

[0015] Through the discrete optical power time series {X t}Establish timing dependency according to the following formula:

[0016]

[0017] Among them, L represents the sequence length, and the autocorrelation coefficient R xx (τ) represents the sequence {X t} and its τ delay {X t-τ}, after the calculation is completed, the most likely top k time series τ1, τ2, ..., τk ;in:

[0018] Through the discrete optical power time series {X t}According to the following formula, the timing dependency is efficiently solved by Fourier transform:

[0019]

[0020] Among them, S xx (f) represents the autocorrelation coefficient in the frequency domain, and represent the fast Fourier transform and its inverse transform, respectively. express The conjugate part in the frequency domain calculation; τ∈{1,2,...,L};

[0021] The delay information aggregation relationship is established through the timing dependency according to the following formula:

[0022]

[0023] Among them, argTopK(·) aims to obtain the parameter set of the top k autocorrelations; c is a hyperparameter; R Q,K is the autocorrelation result of the query Q and the key K after the softmax operation; Roll(X,τ) represents the calculation process of X with a time delay of τ, and Auto-Correlation represents the calculation process of the autocorrelation coefficient.

[0024] Furthermore, the self-organizing map neural network clusters the weather forecast types to obtain generalized weather types, and encodes them in a time series encoder as discrete features of the optical power daily series, including:

[0025] The historical optical power daily series is subjected to wavelet transform to obtain the approximate signal representing the change trend as the clear sky series u of the day, and the detail signal representing the random fluctuation as the fluctuation daily series v of the day; on this basis, the amplitude and fluctuation characteristics of the optical power series are obtained:

[0026] The optical power amplitude adopts the average value of the clear sky sequence u To measure, the calculation is as follows:

[0027]

[0028] In the formula, u i is the i-th sampling value of u; n1 and n2 are the sampling start and end times respectively;

[0029] The optical power fluctuation amplitude adopts the first-order difference index of the fluctuation sequence v △ f express:

[0030]

[0031] In the formula, v i is the i-th sampling value of v;

[0032] The frequency of optical power fluctuation requires fast Fourier transform of the fluctuation sequence v to determine its central frequency; the amplitude characteristics and fluctuation characteristics of the optical power daily series are used as input variables, and the SOM neural network clustering algorithm is used for step-by-step binary division to obtain D types of generalized weather types, and the number of D is determined by the Davidson-Botting index;

[0033] The classification of optical power daily series under historical weather forecast types is statistically analyzed, and the most frequently occurring generalized weather types correspond to the original weather forecast types, so as to achieve the reduction and classification of weather forecast types;

[0034] One-hot encoding is used to encode the generalized weather type into discrete features of the daily series of optical power in the time series encoder.

[0035] Furthermore, the timing encoder alternately decomposes the optical power day sequence to obtain period information; including:

[0036] The temporal encoder is composed of N encoding layers, where the overall equation of the lth encoder layer can be expressed as:

[0037]

[0038] The temporal encoder gradually eliminates the trend component according to the following formula. The input of the encoder module is a length of L seq Optical power timing sequence (d represents dimension); get the periodic component

[0039]

[0040] Among them, “_” represents the trend component that is eliminated. represents the output of the encoder at layer l, They respectively represent the periodic components output by the i-th sequence decomposition unit in the l-th layer.

[0041] Furthermore, the timing decoder refines the periodic information through an autocorrelation mechanism unit; including:

[0042] The decoder is composed of M decoding layers, where the overall equation of the lth decoder layer can be expressed as:

[0043]

[0044] The upper branch processing unit performs alternate decomposition according to the following formula to obtain periodic components;

[0045]

[0046] In the formula, represents the output of the l-th layer decoder, They represent the period component and trend component of the output of the i-th sequence decomposition unit in the l-th layer, respectively. SeriesDecomp represents the time series decomposition module, and FeedForward represents the feedforward module.

[0047] The lower branch sublayer obtains the trend component by weighted addition based on the output of the upper branch sublayer

[0048]

[0049] In the formula, w l,i , i∈{1,2,3} represents the trend component extracted for the i-th time Projection.

[0050] Furthermore, the probability density estimation module optimizes the high-dimensional mapping relationship and the periodicity and trend characteristics of the optical power sequence to obtain the optical power probability distribution parameter output, including:

[0051] According to the optimization combination of high-dimensional mapping relationship and periodicity and trend characteristics of optical power sequence, assuming that the observed data is x, the maximum likelihood function of Gaussian distribution parameters is constructed according to the following formula:

[0052]

[0053] In the formula, μ and σ represent the expected value and standard deviation of Gaussian distribution, n represents the number of samples, argmax represents the maximum likelihood estimate, and X i represents the i-th sample;

[0054] According to the maximum likelihood function, the optical power probability distribution parameters are obtained according to the following formula:

[0055]

[0056] Where: p is the output power data in the training data, and is the estimated parameters output by the probability density estimation layer.

[0057] The present invention is also implemented by the following technical solution:

[0058] A photovoltaic power short-term probability prediction method comprises the following steps:

[0059] The meteorological factors are screened through convolutional neural networks to obtain a high-dimensional mapping relationship between meteorological characteristics and optical power;

[0060] The optical power timing sequence is progressively decomposed and autocorrelation extracted through the timing network to obtain the optical power sequence level link relationship; wherein:

[0061] The discrete features of the daily series of optical power are obtained by clustering the generalized weather types through a self-organizing map neural network.

[0062] The optical power daily series is alternately decomposed by a time encoder to obtain periodic information;

[0063] The periodic information is refined by a timing decoder; wherein:

[0064] The upper branch sub-layer extracts periodic features, and the lower branch sub-layer uses weighted addition to extract trend features.

[0065] The high-dimensional mapping relationship and the optical power sequence-level link relationship are optimized and combined through the probability density estimation module to obtain the optical power probability distribution parameter output.

[0066] Beneficial Effects

[0067] Compared with the prior art, the present invention has the following advantages:

[0068] (1) The proposed short-term probability prediction method for optical power based on the improved CNN-Autoformer network takes into account the impact of weather types on optical power output, classifies weather types and discretizes them into optical power daily series features. The integrated numerical weather forecast information is used to realize optical power probability prediction through the Autoformer network, thereby effectively reducing the risk of false alarms from a single meteorological source and significantly improving the prediction accuracy of the model.

[0069] (2) A progressive decomposition architecture was established to achieve the goal that during the prediction process, the model alternately optimizes the prediction results and decomposes the sequence. In addition, the autocorrelation mechanism was combined to capture the dependencies based on different periods, realize the optical power sequence-level linking, and improve the efficiency of information utilization.

[0070] (3) By combining maximum likelihood estimation with gradient optimization and outputting optical power probability distribution parameters, the optical power interval prediction results and deterministic prediction results can be obtained simultaneously. Compared with the heuristic-based optimization algorithm, the time complexity is reduced and the neural network is effectively prevented from falling into the local optimum. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 It is a short-term probability prediction framework of optical power in a specific implementation mode of the present invention;

[0072] Figure 2It is a convolutional neural network structure in a specific embodiment of the present invention;

[0073] Figure 3 It is the autocorrelation mechanism structure in the specific implementation mode of the present invention;

[0074] Figure 4 is the DBI under different numbers of categories in the embodiment of the present invention;

[0075] Figure 5 is the 90% confidence interval prediction result of each model in the embodiment of the present invention;

[0076] Figure 6 It is the deterministic prediction result of each model in the embodiment of the present invention.

[0077] Table 1 is the model setting parameters in the embodiment of the present invention

[0078] Table 2 is the monthly average values ​​of the evaluation indicators of each model in the embodiment of the present invention. DETAILED DESCRIPTION

[0079] The following is combined with Figure 1 The present invention is described as follows:

[0080] The present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0081] The present invention constructs a short-term optical power probability prediction model based on an improved CNN-Autoformer under a "deep decomposition architecture", and its overall architecture is as follows: Figure 1 As shown, it mainly includes 3 parts:

[0082] 1) High-dimensional meteorological feature extraction for numerical weather forecast based on convolutional neural networks (CNN): The Pearson correlation coefficient method is used to screen meteorological factors related to optical power, such as irradiance, temperature, humidity, etc. Meteorological factors with strong correlation are used as the input of CNN, and the temporal characteristics of a single meteorological factor and the mutual influence between different meteorological factors are captured through the convolution layer; the dimension of the output features of the convolution layer is reduced through the pooling layer, the number of parameters is reduced, and important features are retained to prevent model overfitting; the features extracted by the convolution layer and the pooling layer are integrated and mapped through the fully connected layer, thereby establishing a high-dimensional mapping relationship between the extracted meteorological features and optical power.

[0083] 2) Progressive decomposition and autocorrelation feature extraction based on improved time series autoformer network: In order to deeply extract the periodicity, trend and other characteristics of the optical power time series and reduce the influence of random noise, the improved time series autoformer network is used for feature extraction. In order to extract the changing rules of different periods and analyze the internal structure and inherent rules of the time series, the optical power time series needs to be decomposed.

[0084] 3) Probability distribution parameter optimization based on maximum likelihood estimation: Combine maximum likelihood estimation with the gradient optimization method in machine learning, input the hidden layer information output by the CNN-Autoformer network into the probability density estimation layer, and obtain the optical power probability distribution parameters through the fully connected layer output.

[0085] 1. Meteorological feature extraction

[0086] The typical structure of a CNN network is as follows: Figure 2 As shown in the figure. The main function of the convolution layer is to calculate the feature vector by sliding the convolution kernel at different positions of the input data, thereby deeply extracting the features of the meteorological data. The pooling layer usually follows the convolution layer to reduce the feature dimension while retaining the most significant features, making the CNN network converge faster. At the end of the CNN network is the fully connected layer, which flattens the features extracted by the convolution layer and the pooling layer and outputs the high-dimensional meteorological features of the numerical weather forecast.

[0087] 2. Temporal Autoformer Network

[0088] (1) Discrete feature coding of weather types

[0089] In order to improve the learning efficiency of the model and avoid the underfitting problem caused by too few samples under a single feature, it is necessary to reduce and classify the weather forecast types to obtain generalized weather types. The SOM neural network is used to cluster the optical power daily series, and the generalized weather types are encoded into discrete features of the optical power daily series through one-hot encoding to improve the Autoformer network's ability to identify different weather types.

[0090] (2) Deep Decomposition Architecture

[0091] The present invention uses sequence decomposition as an internal unit of Autoformer and embeds it into the encoder-decoder. In the prediction process, the model can alternately optimize the prediction results and decompose the sequence, gradually separate the trend term and the periodic term from the latent variable, and realize progressive decomposition.

[0092] In the encoder part, assuming there are N encoder layers, the overall equation of the lth encoder layer can be expressed as: The encoder obtains the periodic component by gradually eliminating the trend component

[0093]

[0094] In the formula, “_” represents the trend component that is eliminated. represents the output of the encoder at layer l, They respectively represent the periodic components output by the i-th sequence decomposition unit in the l-th layer.

[0095] In the decoder part, assuming there are M decoder layers, the overall equation of the lth decoder layer can be expressed as: Each decoder layer contains an internal autocorrelation mechanism and an encoder-decoder autocorrelation mechanism, which can respectively refine the prediction and utilize the decoder's periodic component information. The decoder adopts a dual-path processing mode, where the upper branch processes the periodic component and the lower branch uses weighted addition to obtain the trend component based on the output of the upper branch sublayer. The specific process is as follows:

[0096]

[0097] In the formula, represents the output of the l-th layer decoder; They represent the periodic component and trend component of the output of the i-th sequence decomposition unit in the l-th layer, respectively, and w l,i , i∈{1,2,3} represents the trend component extracted for the i-th time Projection.

[0098] (3) Autocorrelation mechanism

[0099] Optical power sequences usually exhibit similar sub-processes under similar phases. Therefore, the autocorrelation mechanism of the present invention replaces the traditional self-attention mechanism to enhance the feature extraction capability of similar weather processes and adjacent days.

[0100] Time-series dependency: The modeling method of the present invention is based on autonomous modeling of historical data, where the current trend is highly correlated with the historical time series. The current time pattern can be frequently observed in the historical subsequence. For a discrete optical power time series {X t}, in order to model the correlation between its time series, the autocorrelation coefficient R is proposed xx (τ), can be calculated by the following formula:

[0101]

[0102] In the formula, the autocorrelation coefficient R xx (τ) represents the sequence {X t} and its τ delay {X t-τ}, after the calculation is completed, the most likely top k time series τ1, τ2, ..., τ k .

[0103] Delay information aggregation: The proposed time series dependency feature can quantify the correlation between the historical subsequence and the current estimated sequence. On this basis, the delay information aggregation module is used to select the first k delay sequences τ1, τ2, ..., τ k Roll the sequence with a time delay of τ to the current phase period, and finally, aggregate and merge all subsequences to perform a softmax operation. The whole process is different from the traditional dot product aggregation mechanism based on the self-attention mechanism.

[0104] In order to calculate the autocorrelation coefficient, similar to self-attention, a multi-head mechanism is also used. For one head, assuming a time series of length L, similar to the self-attention mechanism, after the high-dimensional projection process, Q, K, and V can be obtained. Self-attention can be replaced, and the autocorrelation can be expressed as follows:

[0105]

[0106] Where argTopK(·) is intended to obtain the parameter set of the top k autocorrelations. In the present invention, Where c is a hyperparameter. Q,K is the autocorrelation result of the query Q and the key K after the softmax operation. Roll(X,τ) represents the calculation process of X with a delay of τ, where the current sequence being processed will move one time step as the similarity is calculated until the sequence moves to the τ lag time step.

[0107] To speed up the calculation process, the time lag information of Roll(X,τ) will be stored in a matrix, and the similarity calculation is only performed once through the matrix dot product. As for the autocorrelation between the encoder and decoder, similar to self-attention, the key K and value V are the encoder The output of the query Q comes from the previous block of the decoder.

[0108] Therefore, suppose the query, key, and value of the i-th head are where i∈{1,2,...,h}, L seq represents the sequence length, with d model The multi-head autocorrelation of channels and h heads is calculated as follows:

[0109] Muti-Head(Q,K,V)=W output *Concat(head1,head2,...,head n ) (9)

[0110] Among them, W output Indicates output projection, Concat indicates a join operation, head i Calculate according to the following formula:

[0111] head i =Auto-Correlation(Q i ,K i ,V i ) (10)

[0112] Fast calculation: To improve the computational efficiency, as shown in equations (11)-(12), the proposed autocorrelation mechanism and Fast Fourier Transforms (FFT) calculation mechanism can reduce the computational complexity to O(LlogL). Similarly, a memory cost of O(LlogL) can be achieved. FFT can transform continuous data in the time domain into simpler discrete frequency features in the frequency domain, thus allowing the Autoformer to process long time series. For the time series {X t}, its autocorrelation coefficient R xx (τ) can be expressed as Figure 3 The calculation is performed through FFT, and the process is as follows:

[0113]

[0114] Among them, S xx (f) represents the autocorrelation coefficient in the frequency domain, and represent the fast Fourier transform and its inverse transform, respectively. express The conjugate part in the frequency domain calculation. τ∈{1,2,...,L}. In addition, due to the matrix dot product, the autocorrelation process only needs to be calculated once by using the Fast Fourier Transform (FFT). Therefore, the complexity of the autocorrelation mechanism is O(LlogL).

[0115] 3. Probability Density Estimation Layer

[0116] The present invention combines maximum likelihood estimation with gradient optimization of deep learning. By training sample observations, the most likely parameter values ​​of the probability density distribution function are inferred. The prediction interval and deterministic prediction value under the specified coverage probability can be obtained through the probability density distribution function.

[0117] Taking Gaussian distribution as an example, the observed data is x, and the maximum likelihood function of Gaussian distribution parameters can be constructed as:

[0118]

[0119] In the formula, μ and σ represent the expected value and standard deviation of Gaussian distribution, n represents the number of samples, argmax represents the maximum likelihood estimate, and X i represents the i-th sample;

[0120] By finding the extreme value of the above formula, we can obtain the optimal Gaussian distribution parameters under the sample observation value. After substituting the probability density function into (13), we take the logarithm of both sides to obtain the final loss function:

[0121]

[0122] Where p is the output power data in the training data, and is the estimated parameters output by the probability density estimation layer.

[0123] 4. Examples

[0124] This example is based on a photovoltaic power station. The power station energy management system contains optical power data, numerical weather forecast data, and weather type forecast data from January 1, 2022 to December 31, 2023. The first 70% of the above data set is used as a training set, the next 20% is used as a validation set, and the remaining 10% is used as a test set. The parameter settings are shown in Table 1.

[0125] Table 1

[0126]

[0127] (1) General classification of weather types

[0128] The characteristics of the optical power daily series are extracted and the SOM neural network clustering algorithm is used for step-by-step classification. The number of classifications N is determined by DBI. The smaller the DBI, the better the classification effect. Figure 5 shown.

[0129] Theoretically, the more types of generalized weather types there are, the more sophisticated the classification model will be, but at the same time, the complexity of the model will increase. Figure 4 It can be seen that when N=5, DBI has a significant inflection point, so the weather forecast weather types are further divided into five generalized weather types.

[0130] (2) Model evaluation

[0131] Two models are used for comparative experiments, namely LSTM and GRU.

[0132] Interval prediction usually uses the prediction interval coverage probability (PICP) to measure the coverage quality of the prediction interval, which means the probability that the measured value falls within the interval, and is defined as follows:

[0133]

[0134] PICE=|PINC-PICP| (16)

[0135] In the formula, n represents the number of calculation points, [L i ,U i ] represents the predicted interval of the i-th point, y i represents the measured value of the i-th point, PINC is the given confidence level, PI CE is the interval coverage error, and the smaller the PI CE is, the better the interval prediction effect is.

[0136] The normalized average width (PI NAW) is used to measure the quality of interval width, which is defined as follows:

[0137]

[0138] In the formula, R represents the difference between the maximum and minimum power values ​​measured on that day;

[0139] In addition, the normalized root mean square error (NRMSE) and normalized mean absolute error (NMAE) are used to evaluate the output deterministic prediction results:

[0140]

[0141] In the above formula, P i is the actual power, Y i is the predicted power, P capcity The optical power data is normalized to the installed capacity, P capcity Take it as 1.

[0142] (3) Comparative experimental results

[0143] Compared with the traditional interval prediction method, the probability density estimation method proposed in the present invention can not only obtain the probability density distribution, but also obtain the deterministic prediction result through the predicted expected value. The CNN-Autoformer model proposed in the present invention is used with the GRU model and the LSTM model to perform 90% confidence interval prediction and deterministic prediction for one month, and one day is randomly selected for visual analysis.

[0144] Figure 5The 90% confidence interval prediction results of the three models are shown. Figure 6 The deterministic prediction results of three models are shown.

[0145] from Figure 5 and Figure 6 It can be seen that the CNN-Autoformer model has the narrowest prediction width and the best following effect in the photovoltaic climbing and landslide stages; in the photovoltaic stable output stage, the 90% confidence interval width of the GRU model is the narrowest, followed by the CNN-Autoformer model, and the widest for the LSTM model. However, the narrow interval width of the GRU model caused the measured optical power value to exceed the lower limit of the 90% confidence interval during 11:30-13:00. The measured optical power of the CNN-Autoformer model briefly exceeded the lower limit of the 90% confidence interval at 11:30 and 12:15, while the LSTM model did not exceed the 90% confidence interval during the noon period because the interval width remained the widest. After 14:00, the photovoltaic output began to fall slowly. The deterministic prediction result of the CNN-Autoformer model was the best, and the measured optical power closely followed the deterministic prediction result. On the contrary, the measured optical power curves of the GRU model and the LSTM model have exceeded the lower limit of the 90% confidence interval. The reason is that, on the one hand, the CNN-Autoformer model captures the law of irradiance drop in the evening period in meteorological information through CNN, and on the other hand, it learns the law of light power change on adjacent days and similar days through the autocorrelation mechanism, thereby improving the prediction performance of the model.

[0146] Table 2 shows the monthly root mean square error, monthly average absolute error, and monthly average coverage error and monthly average coverage width of the deterministic prediction of the model proposed in the present invention and the two comparative models within one month. It can be seen from Table 2 that in the comparison of the four evaluation indicators, the model proposed in the present invention performs the best, followed by the GRU model, and the LSTM model performs the worst. In terms of deterministic prediction, the monthly root mean square error of the model proposed in the present invention is reduced by 3.9% compared with the GRU model, and the monthly average absolute error is reduced by 3.91%; in terms of 90% confidence interval prediction, the monthly average coverage error of the model proposed in the present invention is reduced by 3.55% compared with the GRU model, and the monthly average coverage width is reduced by 11.32%. As shown below:

[0147] Table 2

[0148]

[0149]

[0150] (4) Conclusion

[0151] The case results show that:

[0152] 1) The proposed short-term probability prediction method for optical power based on the improved CNN-Autoformer network takes into account the impact of weather types on optical power output, classifies weather types and discretizes them into optical power daily series features, integrates numerical weather forecast information, effectively reduces the risk of false alarms from a single meteorological source, and significantly improves the prediction accuracy of the model.

[0153] 2) The simulation results of 30 days show that compared with the GRU model, the proposed method improves the NRMSE, NMAE, PICE, and PINAW evaluation indicators by 3.90%, 3.91%, 3.55%, and 11.32%, respectively; compared with the LSTM model, the proposed method improves the NRMSE, NMAE, PICE, and PINAW evaluation indicators by 5.08%, 5.06%, 18.99%, and 19.33%, respectively, which verifies the effectiveness of the proposed method.

[0154] 3) The method proposed in the present invention can be used for short-term power prediction of centralized photovoltaic power stations.

[0155] Although the present invention has been described above, the present invention is not limited to the above-mentioned specific embodiments. The above-mentioned specific embodiments are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can make many modifications without departing from the purpose of the present invention, which are all within the protection of the present invention.

Claims

1. A photovoltaic power short-term probability prediction system, characterized in that: The prediction system comprises: a convolutional neural network, a time series network and a probability density estimation module; the time series network is composed of a self-organizing map neural network, a time series encoder and a time series decoder; the time series encoder and the time series decoder are both composed of an autocorrelation mechanism unit, a time series decomposition unit and a feedforward layer alternately connected; the time series decoder is composed of an upper branch sublayer and a lower branch sublayer; wherein: The convolutional neural network screens meteorological factors to obtain a high-dimensional mapping relationship between meteorological characteristics and optical power; The timing network performs progressive decomposition and autocorrelation extraction on the optical power timing sequence to obtain the periodicity and trend characteristics of the optical power sequence; wherein: The self-organizing map neural network clusters the weather forecast types to obtain generalized weather types, and encodes them in a time series encoder as discrete features of the optical power daily series; The timing encoder alternately decomposes the optical power daily sequence to obtain periodic information; The timing decoder refines the periodic information through an autocorrelation mechanism unit; comprising: The upper branch sub-layer extracts periodic features, and the lower branch sub-layer uses weighted addition to extract trend features. The probability density estimation module optimizes and combines the high-dimensional mapping relationship with the periodicity and trend characteristics of the optical power sequence to obtain the optical power probability distribution parameter output.

2. A photovoltaic power short-term probability prediction system according to claim 1, characterized in that: The timing network performs progressive decomposition and autocorrelation extraction on the optical power timing sequence to obtain the periodicity and trend characteristics of the optical power sequence; including: Through the discrete optical power time series {X t }Establish timing dependency according to the following formula: Among them, L represents the sequence length, and the autocorrelation coefficient R xx (τ) represents the sequence {X t } and its τ delay {X t-τ }, after the calculation is completed, the most likely top k time series τ1, τ2, ..., τ k ;in: Through the discrete optical power time series {X t }According to the following formula, the timing dependency is efficiently solved by Fourier transform: Among them, S xx (f) represents the autocorrelation coefficient in the frequency domain, and represent the fast Fourier transform and its inverse transform, respectively. express The conjugate part in the frequency domain calculation; τ∈{1,2,...,L}; The delay information aggregation relationship is established through the timing dependency according to the following formula: Among them, argTopK(·) aims to obtain the parameter set of the top k autocorrelations; c is a hyperparameter; R Q,K is the autocorrelation result of the query Q and the key K after the softmax operation; Roll(X,τ) represents the calculation process of X with a time delay of τ, and Auto-Correlation represents the calculation process of the autocorrelation coefficient.

3. A photovoltaic power short-term probability prediction system according to claim 1, characterized in that: The self-organizing map neural network clusters the weather forecast types to obtain generalized weather types, and encodes them in a time series encoder as discrete features of the optical power daily sequence, including: The historical optical power daily series is subjected to wavelet transform to obtain the approximate signal representing the change trend as the clear sky series u of the day, and the detail signal representing the random fluctuation as the fluctuation daily series v of the day; on this basis, the amplitude and fluctuation characteristics of the optical power series are obtained: The optical power amplitude adopts the average value of the clear sky sequence u To measure, the calculation is as follows: In the formula, u i is the i-th sampling value of u; n1 and n2 are the sampling start and end times respectively; The optical power fluctuation amplitude adopts the first-order difference index of the fluctuation sequence v △ f express: In the formula, v i is the i-th sampling value of v; The frequency of optical power fluctuation requires fast Fourier transform of the fluctuation sequence v to determine its central frequency; the amplitude characteristics and fluctuation characteristics of the optical power daily series are used as input variables, and the SOM neural network clustering algorithm is used for step-by-step binary division to obtain D types of generalized weather types, and the number of D is determined by the Davidson-Botting index; The classification of optical power daily series under historical weather forecast types is statistically analyzed, and the most frequently occurring generalized weather types correspond to the original weather forecast types, so as to achieve the reduction and classification of weather forecast types; One-hot encoding is used to encode the generalized weather type into discrete features of the daily series of optical power in the time series encoder.

4. A photovoltaic power short-term probability prediction system according to claim 1, characterized in that: The timing encoder alternately decomposes the optical power day sequence to obtain period information; including: The temporal encoder is composed of N encoding layers, where the overall equation of the lth encoder layer can be expressed as: The temporal encoder gradually eliminates the trend component according to the following formula. The input of the encoder module is a length of L seq Optical power timing sequence Get the periodic component Among them, "_" represents the trend component that is eliminated. represents the output of the encoder at layer l, They respectively represent the periodic components output by the i-th sequence decomposition unit in the l-th layer; d represents the dimension.

5. A photovoltaic power short-term probability prediction system according to claim 1, characterized in that: The timing decoder refines the periodic information through the autocorrelation mechanism unit; including: The decoder is composed of M decoding layers, where the overall equation of the lth decoder layer can be expressed as: The upper branch processing unit performs alternate decomposition according to the following formula to obtain periodic components; In the formula, represents the output of the l-th layer decoder, They represent the period component and trend component of the output of the i-th sequence decomposition unit in the l-th layer, SeriesDecomp represents the time series decomposition module, and FeedForward represents the feedforward module; The lower branch sublayer obtains the trend component by weighted addition based on the output of the upper branch sublayer In the formula, w l,i , i∈{1,2,3} represents the trend component extracted for the i-th time Projection.

6. A photovoltaic power short-term probability prediction system according to claim 1, characterized in that: The probability density estimation module optimizes the high-dimensional mapping relationship and the periodicity and trend characteristics of the optical power sequence to obtain the optical power probability distribution parameter output, including: According to the optimization combination of high-dimensional mapping relationship and periodicity and trend characteristics of optical power sequence, assuming that the observed data is x, the maximum likelihood function of Gaussian distribution parameters is constructed according to the following formula: In the formula, μ and σ represent the expected value and standard deviation of Gaussian distribution, n represents the number of samples, argmax represents the maximum likelihood estimate, and X i represents the i-th sample; According to the maximum likelihood function, the optical power probability distribution parameters are obtained according to the following formula: Where: p is the output power data in the training data, and is the estimated parameters output by the probability density estimation layer.

7. A photovoltaic power short-term probability prediction method, characterized in that: The method is implemented based on the system of any one of claims 1 to 6, comprising the following steps: The meteorological factors are screened through convolutional neural networks to obtain a high-dimensional mapping relationship between meteorological characteristics and optical power; The optical power timing sequence is progressively decomposed and autocorrelation extracted through the timing network to obtain the optical power sequence level link relationship; wherein: The discrete features of the daily series of optical power are obtained by clustering the generalized weather types through a self-organizing map neural network. The optical power daily series is alternately decomposed by a time encoder to obtain periodic information; The periodic information is refined by a timing decoder; wherein: The upper branch sub-layer extracts periodic features, and the lower branch sub-layer uses weighted addition to extract trend features. The high-dimensional mapping relationship and the optical power sequence-level link relationship are optimized and combined through the probability density estimation module to obtain the optical power probability distribution parameter output.

Citation Information

Patent Citations

  • Photovoltaic power prediction method and photovoltaic power prediction model training method

    CN116050627A

  • Photovoltaic power prediction method based on improved MICN

    CN116776935A

  • Photovoltaic power probability prediction method and device based on ground-based cloud picture guided depth decomposition

    CN116777066A

  • Photovoltaic power ultra-short-term prediction method based on multi-mode decomposition and multi-branch input

    CN116995670A

  • Multi-scale entropy gated DWTform meteorological data time sequence prediction method and device

    CN117094431A

Cited By

  • Scene generation method and system based on deep learning source load medium and long term prediction

    CN119721455A

  • Network equipment light attenuation intelligent monitoring system and method based on AI algorithm

    CN119814140A

  • An intelligent monitoring system and method for optical attenuation of network equipment based on AI algorithm

    CN119814140B

  • Offshore wind power plant multi-unit probability interval prediction aggregation method, system and equipment based on fast Fourier transform and medium

    CN120409850A