CN116451598B - Solar Irradiance Prediction Method Based on Denoising Diffusion Probability Model - Google Patents
Solar Irradiance Prediction Method Based on Denoising Diffusion Probability Model Download PDFInfo
- Publication number
- CN116451598B CN116451598B CN202310734951.0A CN202310734951A CN116451598B CN 116451598 B CN116451598 B CN 116451598B CN 202310734951 A CN202310734951 A CN 202310734951A CN 116451598 B CN116451598 B CN 116451598B
- Authority
- CN
- China
- Prior art keywords
- solar irradiance
- model
- denoising
- noise
- probability model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000009792 diffusion process Methods 0.000 title claims abstract description 69
- 238000000034 method Methods 0.000 title claims abstract description 62
- 230000008569 process Effects 0.000 claims abstract description 36
- 238000012549 training Methods 0.000 claims abstract description 34
- 238000012360 testing method Methods 0.000 claims description 17
- 230000014509 gene expression Effects 0.000 claims description 6
- 238000005315 distribution function Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 230000009467 reduction Effects 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000013135 deep learning Methods 0.000 description 6
- 238000011156 evaluation Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 4
- 238000010801 machine learning Methods 0.000 description 3
- YHXISWVBGDMDLQ-UHFFFAOYSA-N moclobemide Chemical compound C1=CC(Cl)=CC=C1C(=O)NCCN1CCOCC1 YHXISWVBGDMDLQ-UHFFFAOYSA-N 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 2
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 description 2
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000007477 logistic regression Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000003042 antagnostic effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000001932 seasonal effect Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000010183 spectrum analysis Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Medical Informatics (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Photovoltaic Devices (AREA)
Abstract
The application discloses a solar irradiance prediction method based on a denoising diffusion probability model, which is used for newly customizing the denoising diffusion probability model constructed by a transducer frame, a residual error network and an embedded network, and firstly providing that the solar irradiance is predicted by applying the denoising diffusion probability model, and compared with experimental data of various existing models, the solar irradiance is predicted based on the denoising diffusion probability model customized by the method, so that the solar irradiance prediction method has higher prediction precision. The solar irradiance is predicted by adopting the denoising diffusion probability model, and the 0 value in the solar irradiance time sequence data is not required to be removed in the model training and prediction process, so that the prediction is more convenient and quick, and the method is more suitable for predicting the solar irradiance for a long time.
Description
Technical Field
The application relates to the technical field of solar irradiance prediction, in particular to a solar irradiance prediction method based on a denoising diffusion probability model.
Background
In the prior art, the prediction of solar irradiance is generally short-term prediction, and mainly predicts the solar irradiance observation result of the next time period in a given time period. In terms of the prediction results, it has a high expressive force on a photovoltaic curve with strong regularity, but there is still a lag in the prediction (prediction lag: the reality of the photovoltaic is that the value of the next second tends to be similar or equal to the last second. Training of a model with such a large amount of data tends to make the output value of the model the same (since deep learning is the best solution, the model tends to give the output value according to the last value of the input data) but in the case of real prediction there is a fluctuation, this fluctuation will tend to be the so-called "lag" of the model prediction, it is impossible to predict the next fluctuation (deep learning is to find the best solution, the model tends to give the output value according to the last value of the input data, and thus the next fluctuation cannot be predicted). Meanwhile, the model cannot predict fluctuation, and another problem exists in data preprocessing and prediction of photovoltaic work (0 value exists in the photovoltaic data, if long-time prediction is performed, the model can always output the 0 value and does not accord with real logic), namely most of work can directly discard the 0 value region of the photovoltaic and directly predict from a non-0 region, which is very unscientific, and the prediction of the 0 value region is very important. However, multi-step predictions place high demands on the predictive capabilities of the model, i.e., the ability to effectively capture accurate long-range correlation coupling between output and input. However, it is not very ideal for predicting photovoltaic (multi-step prediction is generally only applicable to predicting average results and does not work well for wave data). For photovoltaic prediction, various works have proposed solutions to solve the problem of prediction hysteresis, such as: wavelet decomposition of input data, singular Spectrum Analysis (SSA), addition of other eigenvalues, and the like. However, despite the achievements already achieved, it will still be limited by the principle of logistic regression.
Related work has shown that generating an antagonistic network (GAN) can better avoid the limitations of logistic regression, but its uniqueness of input makes the diversity of results vanish (GAN always uses the previous day of data as input), and has problems of low analog expressivity and difficulty in training.
Disclosure of Invention
The application aims to improve the accuracy of solar irradiance prediction, and provides a solar irradiance prediction method based on a denoising diffusion probability model.
To achieve the purpose, the application adopts the following technical scheme:
the solar irradiance prediction method based on the denoising diffusion probability model comprises the following steps:
s1, performing model training by taking solar irradiance time series data of each region collected in a history mode as a training sample of a denoising diffusion probability model;
s2, taking solar irradiance time sequence data of a designated area as input of the denoising diffusion probability model, and outputting solar irradiance prediction results of the designated area by the model.
Preferably, in step S1, the method for training the denoising diffusion probability model includes the following steps:
s11, adding random noise to the original solar irradiance time series data of each region;
s12, restoring the random noise into the corresponding original solar irradiance time series data;
s13, calculating model prediction loss;
s14, verifying whether the model prediction loss reaches the expected value,
if yes, stopping model iterative training and outputting the denoising diffusion probability model;
if not, the model parameters are adjusted, and then the step S11 is returned to continue the model iterative training.
Preferably, the step S14 is performed after the yes judgment:
s15, evaluating the performance of the denoising diffusion probability model,
if the performance reaches the standard, finally outputting the denoising diffusion probability model;
if the performance does not reach the standard, returning to the step S11.
Preferably, in step S11, the method of adding random noise to the original solar irradiance time series data is expressed by the following formula (1):
in the formula (1)，/>，/>Representing the time series data of the original solar irradiance +.>Is>Adding noise for a time>Representing the current noise adding times;
representation pair->Make->The data after the secondary noise addition is represented;
for a pair ofThe process of adding random noise is a forward process of the denoising diffusion probability model, the forward process is a Markov process which will +.>And (3) adding noise for multiple times to finally obtain Gaussian noise, wherein the Markov process is expressed by the following formulas (2) - (3):
in the formulas (2) to (3),is an increasing sequence, and +.>Between 0 and 1;
t represents the total number of times of noise addition;
t represents the current noise adding times;
indicate->Minor pair->Adding noise;
indicating the last->Minor pair->Adding noise;
representation->To->Conditional probability expressions of (2);
representing the identity matrix;
indicating that this is a normal distribution function;
representing any one of the first to the T-th noise addition of X;
representation->And->Probability relation of (2);
when (when)，/>Tends to be Gaussian, since the diffusion process of X conforms to the Markov principle, for +.>Is reduced to the expression of equation (1) above.
Preferably, in step S12, the process of recovering the random noise is expressed by the following formula (4):
in the formula (4) of the present application,indicate->Minor pairOriginal solar irradiance time series data +.>A noise adding result of (2);
indicate->Minor pair->A noise adding result of (2);
representing the denoising diffusion probability model;
is->Is represented by the element;
representation model pair->Is predicted by->And t is the control condition of the model;
indicate->Reducing errors of the secondary noise adding result;
。
preferably, in step S13, the model predictive loss is calculated by a loss function expressed by the following formula (5):
in the formula (5) of the present application,indicate->Predicting loss of reduction by using a secondary noise adding result;
indicate->Noise of secondary noise addition;
an output representing the model;
representing the training modelAnd t is the result obtained after the input;
representation->Is represented by the element;
representing the time series data of the original solar irradiance +.>Make->The data after the secondary noise addition is represented.
Preferably, in step S15, the performance of the denoising diffusion probability model is evaluated by calculating the average absolute value error of the predicted value and the true value, and the calculation method is expressed by the following formula (6):
in the formula (6) of the present application,respectively representing the real solar irradiance value of the appointed area and the predicted solar irradiance value of the appointed area by the model;
respectively represent +.>True value and model of solar irradiance of the test specimen +.>Predicted solar irradiance values for each test specimen;
representing the length of the test set.
Preferably, in step S15, the performance of the denoising diffusion probability model is evaluated by calculating the average absolute percentage error of the predicted value and the true value, and the calculation method is expressed by the following formula (7):
in the formula (7) of the present application,respectively representing the real solar irradiance value of the appointed area and the predicted solar irradiance value of the appointed area by the model;
respectively represent +.>True value and model of solar irradiance of the test specimen +.>Predicted solar irradiance values for each test specimen;
representing the length of the test set.
Preferably, the elements of the solar irradiance time series data include a "0" value, the "0" value indicating that the element data content is null.
The application has the following beneficial effects:
1. the application newly designs a denoising diffusion probability model constructed by a transducer framework, a residual error network and an embedded network, and firstly proposes that the denoising diffusion probability model is applied to predict solar irradiance, and compared with experimental data of various existing models, the denoising diffusion probability model customized based on the application predicts solar irradiance with higher prediction precision.
2. The solar irradiance is predicted by adopting the denoising diffusion probability model, and the 0 value in the solar irradiance time sequence data is not required to be removed in the model training and prediction process, so that the prediction is more convenient and quick, and the method is more suitable for predicting the solar irradiance for a long time.
Drawings
In order to more clearly illustrate the technical solution of the embodiments of the present application, the drawings that are required to be used in the embodiments of the present application will be briefly described below. It is evident that the drawings described below are only some embodiments of the present application and that other drawings may be obtained from these drawings without inventive effort for a person of ordinary skill in the art.
FIG. 1 is a frame diagram of a prediction algorithm of a denoising diffusion probability model provided by an embodiment of the present application;
FIG. 2 is a network structure diagram of a deep learning network for training a denoising diffusion probability model;
FIG. 3 is a schematic diagram of the forward process of the denoising diffusion probability model;
FIG. 4 is a plot of test set samples for real data, experimental data, and control groups;
fig. 5 is a diagram of implementation steps of a solar irradiance prediction method based on a denoising diffusion probability model according to an embodiment of the present application.
Detailed Description
The technical scheme of the application is further described below by the specific embodiments with reference to the accompanying drawings.
Wherein the drawings are for illustrative purposes only and are shown in schematic, non-physical, and not intended to be limiting of the present patent; for the purpose of better illustrating embodiments of the application, certain elements of the drawings may be omitted, enlarged or reduced and do not represent the size of the actual product; it will be appreciated by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.
The same or similar reference numbers in the drawings of embodiments of the application correspond to the same or similar components; in the description of the present application, it should be understood that, if the terms "upper", "lower", "left", "right", "inner", "outer", etc. indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings, only for convenience in describing the present application and simplifying the description, rather than indicating or implying that the apparatus or elements being referred to must have a specific orientation, be constructed and operated in a specific orientation, so that the terms describing the positional relationships in the drawings are merely for exemplary illustration and should not be construed as limiting the present patent, and that the specific meaning of the terms described above may be understood by those of ordinary skill in the art according to specific circumstances.
In the description of the present application, unless explicitly stated and limited otherwise, the term "coupled" or the like should be interpreted broadly, as it may be fixedly coupled, detachably coupled, or integrally formed, as indicating the relationship of components; can be mechanically or electrically connected; can be directly connected or indirectly connected through an intermediate medium, and can be communication between the two parts or interaction relationship between the two parts. The specific meaning of the above terms in the present application will be understood in specific cases by those of ordinary skill in the art.
The solar irradiance prediction method based on the denoising diffusion probability model provided by the embodiment of the application, as shown in fig. 5, comprises the following steps:
s1, performing model training by taking solar irradiance time series data of each region collected in a history mode as a training sample of a denoising diffusion probability model;
s2, taking solar irradiance time sequence data of a designated area as input of a denoising diffusion probability model, and outputting a solar irradiance prediction result of the designated area by the model.
In step S1, the method for training the denoising diffusion probability model specifically includes the following steps:
s11, adding random noise to the original solar irradiance time series data of each region;
s12, reducing random noise into corresponding original solar irradiance time series data;
s13, calculating model prediction loss;
s14, verifying whether the model prediction loss reaches the expected value;
if yes, stopping model iterative training and outputting a denoising diffusion probability model;
if not, the model parameters are adjusted, and then the step S11 is returned to continue the model iterative training.
Specifically, the process of training the denoising diffusion probability model comprises the following links:
1) Collecting solar irradiance data of recent years (such as 3 years) in each region, and sequentially recording irradiance values in time sequence to obtain time sequence dataWherein->Represents the +.o. for a certain region acquisition>Solar irradiance data->Representing the number of solar irradiance data collected, +.>；
It should be noted here that the time seriesThe data in the system are historical solar irradiance data of the same area, time sequence +.>Preferably 10 minutes.
2) And carrying out missing value processing and normalization processing on the collected solar irradiance time series data of each region. For the time series data of each region, firstly carrying out missing value processing, wherein the processing method comprises the following steps: if the value at a certain time point in the sequence is missing, the value at the time point immediately preceding the certain time point is replaced by the value at the time point. Next, there are many existing methods of normalizing data for time series, and therefore, the method of normalizing data specifically adopted in this embodiment will not be described here. Finally, the preprocessed data is used for manufacturing training numbers put into a model training network according to a preset time stepAnd corresponding data tags. For example, if it is desired to predict solar irradiance on the following day using solar irradiance time series data on the preceding day, a dataset may be createdData of previous day +.>Data from the day after prediction +.>。
3) Dividing the preprocessed data in the step 2) into a training set, a verification set and a test set according to the original sequence of the data, wherein the dividing ratio is 3:1:1.
4) The divided data are input into a deep learning network for training a denoising diffusion probability model for model training, and the deep learning network is constructed by a transducer framework, a residual network and an embedded network as shown in fig. 2. The method for training the denoising diffusion probability model consists of 3 important stages: firstly, preparing network data as a forward process, wherein the forward process is a process of continuously adding Gaussian noise to the original solar irradiance time sequence data and finally generating random noise. The forward process is specifically shown in fig. 3, and numbers on the left and right sides in fig. 3 are the sampling results of the current step number. The left side of the figure is a comparison of the sampled results from the top 5 sets of 1, 20, 40, 60, 80 with the non-sampled results. The top four from top to bottom on the right of the figure are 200, 300, 400, 500 samples and the final results. The last plot is a comparison of the final result samples with the initial results. Secondly, the inverse process of the network simulation chain gradually restores random noise into original data through prediction noise, which is also a Markov process like a forward process, in the inverse process, the expected curve is obtained by gradually denoising from a Gaussian noise (the inverse process is specifically that firstly, a Gaussian noise curve is randomly initialized, then, denoising is carried out for a plurality of times through a model, and finally, a photovoltaic curve of a prediction target is obtained). Finally, a loss function of the model is calculated to optimize the model parameters.
The specific algorithm of the denoising diffusion probability model is shown in fig. 1, and comprises the following processes:
a) The forward process is a markov process that will multiply the raw data to ultimately yield a random noise, expressed by the following formulas (2) - (3):
in the formulas (2) to (3),is an increasing sequence, and +.>Between 0 and 1. For example, assume +.>Is an arithmetic increment and T is equal to 100, then +.>May be denoted as 0,0.01,0.02, … …,1;
t represents the total number of times of noise addition;
t represents the current noise adding times;
indicate->Minor pair->Adding noise;
indicating the last->Minor pair->Adding noise;
representation->To->Conditional probability expressions of (2);
representing the identity matrix;
indicating that this is a normal distribution function;
representing any one of the first to the T-th noise addition of X;
representation->And->Probability relation of (2);
when (when)，/>Tends to be Gaussian, since we assume that the diffusion process of X is in accordance with the Markov principle, for ∈>Is simplified to the expression of formula (1) below:
in the formula (1)，/>，/>Representing the time series data of the original solar irradiance +.>Is>Adding noise for a time>Representing the current noise adding times;
representation pair->Make->The data after the secondary noise addition is represented.
b) The model simulates the back propagation of the markov chain to simulate the back process: by predicting the noise, gradually reducing the random noise to the original data, which is also a markov process, in hopes of gradually denoising from a gaussian noise segment to get the desired curve, the inverse process can be expressed by the following equation (4):
in the formula (4) of the present application,indicate->Sub-pair raw solar irradiance time series data +.>A noise adding result of (2);
indicate->Minor pair->A noise adding result of (2);
representing the denoising diffusion probability model;
is->Is represented by the element;
representation model pair->Is predicted by->And t is the control condition of the model;
indicate->Reducing errors of the secondary noise adding result;
。
the reverse process described above can also be expressed by the following formulas (8) - (9):
the model prediction formula can be obtained based on conditional probability reduction as follows:
in the formulas (8) - (9),representing the mean value of the calculated t denoising process,/->For its variance->Representation->Normal distribution function of ∈10->Representation->Is a conditional probability function of (2);
the mean and variance calculated for the model respectively,representation of model predicted +.>Is used for the normal distribution function of (a),representation of model predicted +.>Conditional probability function of>For their condition.
C) The model predictive loss is calculated using a loss function expressed by the following equation (5):
in the formula (5) of the present application,indicate->Predicting loss of reduction by using a secondary noise adding result;
indicate->Noise of secondary noise addition;
an output representing the model;
representing the training modelAnd t is the result obtained after the input;
representation->Is represented by the element;
representing the time series data of the original solar irradiance +.>Make->The data after the secondary noise addition is represented.
The above formula (5) is simplified from the following formula (10):
in the formula (10) of the present application,is an increasing sequence, and +.>Between 0 and 1. For example we assume in experiments thatIs an arithmetic increment and T is equal to 100, then +.>May be denoted as 0,0.01,0.02, … …,1;
representing the error of each denoising;
representing the expected loss function of the model, the input value at the model is +.>At the time of (I)>Noise is distributed for a positive too much.
5) And adjusting the iteration times and model parameters of model training according to the error evaluation index, and finally training to obtain the denoising diffusion probability model for predicting solar irradiance.
In order to evaluate the prediction performance of the denoising diffusion probability model on solar irradiance, the present embodiment adopts an average absolute value error (MAE) and/or average absolute percentage error (MAPE) evaluation method, and the two evaluation methods are expressed by the following formulas (6) and (7), respectively:
in the formulas (6) to (7),respectively representing the real solar irradiance value of the appointed area and the predicted solar irradiance value of the appointed area by the model;
respectively represent +.>True value and model of solar irradiance of the test specimen +.>Predicted solar irradiance values for each test specimen;
representing the length of the test set.
6) The solar irradiance is predicted, specifically: and inputting the acquired solar irradiance time sequence data into a denoised diffusion probability model after training, and outputting a solar irradiance prediction result by the model through the forward process, the reverse process and the like.
In order to verify the performance of the denoising diffusion probability model for predicting solar irradiance provided by the embodiment, a machine learning model which is popular in the technical field of solar irradiance prediction at present is selected for comparison. These models include Informir (Informir model is the best paper for AAAI conference 2021. The main contributions are in long-sequence prediction), lstnet (Lstnet is a deep learning network specifically designed for time-sequence prediction), deep AR (deep AR model is a deep learning model for time-sequence prediction and generation. Time-sequence data with complex seasonal and trend is intended to be processed), AST (AST model is a model that combines GAN with a transducer), ARIMA (ARIMA model combines the characteristics of Autoregressive (AR) model, differential (I) and Moving Average (MA) model). Fig. 4 shows model performance comparison curves of the above 2 existing models (Informir and ARIMA) and the Denoising Diffusion Probability Model (DDPM) provided in this example, and Label in fig. 4 is a true value curve. In addition, the following tables 1 and 2 show the prediction performance comparison data of the above 5 existing models and the denoising diffusion probability model provided in this example, respectively. The prediction performance of the denoising diffusion probability model proposed in this example is significantly better than that of the comparative model from the fitted curve shown in fig. 4 and the performance evaluation indexes given in tables 1 and 2 below.
Table 1: MAPE evaluation index comparison of denoising diffusion probability model and other machine learning models
Table 2: MSE evaluation index comparison of denoising diffusion probability model and other machine learning models
In summary, the application has the following beneficial effects:
1. the application newly designs a denoising diffusion probability model constructed by a transducer framework, a residual error network and an embedded network, and firstly proposes that the denoising diffusion probability model is applied to predict solar irradiance, and compared with experimental data of various existing models, the denoising diffusion probability model customized based on the application predicts solar irradiance with higher prediction precision.
2. The solar irradiance is predicted by adopting the denoising diffusion probability model, and the 0 value in the solar irradiance time sequence data is not required to be removed in the model training and prediction process, so that the prediction is more convenient and quick, and the method is more suitable for predicting the solar irradiance for a long time.
It should be understood that the above description is only illustrative of the preferred embodiments of the present application and the technical principles employed. It will be apparent to those skilled in the art that various modifications, equivalents, variations, and the like can be made to the present application. However, such modifications are intended to fall within the scope of the present application without departing from the spirit of the present application. In addition, some terms used in the description and claims of the present application are not limiting, but are merely for convenience of description.
Claims (5)
1. A solar irradiance prediction method based on a denoising diffusion probability model is characterized by comprising the following steps:
s1, performing model training by taking solar irradiance time series data of each region collected in a history mode as a training sample of a denoising diffusion probability model;
s2, taking solar irradiance time sequence data of a designated area as input of the denoising diffusion probability model, and outputting a solar irradiance prediction result of the designated area by the model;
the elements of the solar irradiance time series data comprise '0' values, wherein '0' values represent that the element data content is empty;
in step S1, the method for training the denoising diffusion probability model includes the following steps:
s11, adding random noise to the original solar irradiance time series data of each region;
s12, restoring the random noise into the corresponding original solar irradiance time series data;
s13, calculating model prediction loss;
s14, verifying whether the model prediction loss reaches the expected value,
if yes, stopping model iterative training and outputting the denoising diffusion probability model;
if not, returning to the step S11 to continue model iterative training after adjusting the model parameters;
in step S11, the method of adding random noise to the original solar irradiance time series data is expressed by the following formula (1):
in the formula (1),，/>，/>representing the time series data of the original solar irradiance +.>Is>Adding noise for a time>Representing the current noise adding times;
representation pair->Make->The data after the secondary noise addition is represented;
for a pair ofThe process of adding random noise is a forward process of the denoising diffusion probability model, the forward process is a Markov process which will +.>And (3) adding noise for multiple times to finally obtain Gaussian noise, wherein the Markov process is expressed by the following formulas (2) - (3):
in the formulas (2) to (3),is an incremental sequence;
t represents the total number of times of noise addition;
t represents the current noise adding times;
indicate->Minor pair->Adding noise;
indicating the last->Minor pair->Adding noise;
representation->To->Conditional probability expressions of (2);
representing the identity matrix;
representing a normal distribution function;
representing any one of the first to the T-th noise addition of X;
representation->And->Probability relation of (2);
when (when)，/>Tends to be Gaussian, since the diffusion process of X conforms to the Markov principle, for +.>Is reduced to the expression of formula (1) above;
in step S12, the process of recovering the random noise is expressed by the following formula (4):
in the formula (4) of the present application,indicate->Sub-pair raw solar irradiance time series data +.>A noise adding result of (2);
indicate->Minor pair->A noise adding result of (2);
representing the denoising diffusion probability model;
is->Is represented by the element;
representation model pair->Is predicted by->And t is the control condition of the model;
indicate->Reducing errors of the secondary noise adding result;
。
2. the solar irradiance prediction method based on the denoising diffusion probability model according to claim 1, wherein the step S14 is performed after the yes judgment:
s15, evaluating the performance of the denoising diffusion probability model,
if the performance reaches the standard, finally outputting the denoising diffusion probability model;
if the performance does not reach the standard, returning to the step S11.
3. The solar irradiance prediction method based on the denoising diffusion probability model according to claim 1, wherein in step S13, model prediction loss is calculated by a loss function expressed by the following formula (5):
in the formula (5) of the present application,indicate->Predicting loss of reduction by using a secondary noise adding result;
indicate->Noise of secondary noise addition;
representation modelOutputting a model;
representing the training model will->And t is taken as the result obtained after input;
representation->Is represented by the element;
representing the time series data of the original solar irradiance +.>Make->The data after the secondary noise addition is represented.
4. The solar irradiance prediction method based on the denoising diffusion probability model according to claim 2, wherein in step S15, the performance of the denoising diffusion probability model is evaluated by calculating the average absolute value error of the predicted value and the true value, the calculation method being expressed by the following formula (6):
in the formula (6) of the present application,respectively represent the solar irradiance of the appointed regionA predicted value of the true value and the model for solar irradiance in the specified region;
respectively represent +.>True value and model of solar irradiance of the test specimen +.>Predicted solar irradiance values for each test specimen;
representing the length of the test set.
5. The solar irradiance prediction method based on the denoising diffusion probability model according to claim 2, wherein in step S15, the performance of the denoising diffusion probability model is evaluated by calculating the average absolute percentage error of the predicted value and the true value, the calculation method being expressed by the following formula (7):
in the formula (7) of the present application,respectively representing the real solar irradiance value of the appointed area and the predicted solar irradiance value of the appointed area by the model;
respectively represent +.>True value and model of solar irradiance of the test specimen +.>Predicted solar irradiance values for each test specimen;
representing the length of the test set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310734951.0A CN116451598B (en) | 2023-06-20 | 2023-06-20 | Solar Irradiance Prediction Method Based on Denoising Diffusion Probability Model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310734951.0A CN116451598B (en) | 2023-06-20 | 2023-06-20 | Solar Irradiance Prediction Method Based on Denoising Diffusion Probability Model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116451598A CN116451598A (en) | 2023-07-18 |
| CN116451598B true CN116451598B (en) | 2023-09-05 |
Family
ID=87130655
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310734951.0A Active CN116451598B (en) | 2023-06-20 | 2023-06-20 | Solar Irradiance Prediction Method Based on Denoising Diffusion Probability Model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116451598B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016210102A1 (en) * | 2015-06-23 | 2016-12-29 | Qatar Foundation For Education, Science And Community Development | Method of forecasting for solar-based power systems |
| WO2018065045A1 (en) * | 2016-10-05 | 2018-04-12 | Telecom Italia S.P.A. | Method and system for estimating energy generation based on solar irradiance forecasting |
| CN110322364A (en) * | 2019-06-19 | 2019-10-11 | 山东大学 | A kind of short-term photovoltaic power generation prediction technique and system |
| CN113780636A (en) * | 2021-08-26 | 2021-12-10 | 河北工业大学 | Solar radiation prediction method based on EMD-GRU-Attention |
| CN114201924A (en) * | 2022-02-16 | 2022-03-18 | 杭州经纬信息技术股份有限公司 | Solar irradiance prediction method and system based on transfer learning |
| CN116167465A (en) * | 2023-04-23 | 2023-05-26 | 杭州经纬信息技术股份有限公司 | Solar irradiance prediction method based on multivariate time series ensemble learning |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111932468A (en) * | 2020-07-20 | 2020-11-13 | 清华大学 | Bayesian image denoising method based on noise-containing image distribution constraint |
-
2023
- 2023-06-20 CN CN202310734951.0A patent/CN116451598B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016210102A1 (en) * | 2015-06-23 | 2016-12-29 | Qatar Foundation For Education, Science And Community Development | Method of forecasting for solar-based power systems |
| WO2018065045A1 (en) * | 2016-10-05 | 2018-04-12 | Telecom Italia S.P.A. | Method and system for estimating energy generation based on solar irradiance forecasting |
| CN110322364A (en) * | 2019-06-19 | 2019-10-11 | 山东大学 | A kind of short-term photovoltaic power generation prediction technique and system |
| CN113780636A (en) * | 2021-08-26 | 2021-12-10 | 河北工业大学 | Solar radiation prediction method based on EMD-GRU-Attention |
| CN114201924A (en) * | 2022-02-16 | 2022-03-18 | 杭州经纬信息技术股份有限公司 | Solar irradiance prediction method and system based on transfer learning |
| CN116167465A (en) * | 2023-04-23 | 2023-05-26 | 杭州经纬信息技术股份有限公司 | Solar irradiance prediction method based on multivariate time series ensemble learning |
Non-Patent Citations (1)
| Title |
|---|
| 海涛 ; 王路 ; 陈春华 ; 梁挺兴 ; 林波 ; 李朝伟 ; .基于RBF神经网络的太阳光辐照度预测.广西大学学报(自然科学版).2016,(第05期),全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116451598A (en) | 2023-07-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110705692B (en) | Nonlinear dynamic industrial process product prediction method of space-time attention network | |
| CN108804611B (en) | Dialog reply generation method and system based on self comment sequence learning | |
| CN111860785A (en) | Time sequence prediction method and system based on attention mechanism cyclic neural network | |
| CN108647226B (en) | Hybrid recommendation method based on variational automatic encoder | |
| CN110705743A (en) | New energy consumption electric quantity prediction method based on long-term and short-term memory neural network | |
| CN111917785B (en) | Industrial internet security situation prediction method based on DE-GWO-SVR | |
| CN113988449B (en) | Wind power prediction method based on transducer model | |
| CN115146700B (en) | Runoff prediction method based on transform sequence-to-sequence model | |
| CN112560948B (en) | Fundus image classification method and imaging method under data deviation | |
| CN115906954A (en) | Multivariate time sequence prediction method and device based on graph neural network | |
| Aue et al. | Delay times of sequential procedures for multiple time series regression models | |
| CN117786602A (en) | Long-period multi-element time sequence prediction method based on multi-element information interaction | |
| CN113609766B (en) | Soft measurement method based on depth probability hidden model | |
| CN115358838A (en) | Credit time series data modeling method and device based on convolutional neural network | |
| Hu et al. | An efficient Long Short-Term Memory model based on Laplacian Eigenmap in artificial neural networks | |
| CN114399101A (en) | TCN-BIGRU-based gas load prediction method and device | |
| CN108090266B (en) | Method for calculating related reliability of multiple failure modes of mechanical part | |
| Vo et al. | Harnessing attention mechanisms in a comprehensive deep learning approach for induction motor fault diagnosis using raw electrical signals | |
| CN116451598B (en) | Solar Irradiance Prediction Method Based on Denoising Diffusion Probability Model | |
| CN113469013A (en) | Motor fault prediction method and system based on transfer learning and time sequence | |
| CN117094431A (en) | DWTfar meteorological data time sequence prediction method and equipment for multi-scale entropy gating | |
| CN112348656A (en) | BA-WNN-based personal loan credit scoring method | |
| Kale et al. | Forecasting Indian stock market using artificial neural networks | |
| CN115758187A (en) | Coal mine mechanical motor state prediction method based on K-CNN-N-GRU | |
| Yang | Market Forecast using XGboost and Hyperparameters Optimized by TPE |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |