CN116470486A - Ultra-short-term photovoltaic power generation power prediction method, device and storage medium - Google Patents
Ultra-short-term photovoltaic power generation power prediction method, device and storage medium Download PDFInfo
- Publication number
- CN116470486A CN116470486A CN202310399296.8A CN202310399296A CN116470486A CN 116470486 A CN116470486 A CN 116470486A CN 202310399296 A CN202310399296 A CN 202310399296A CN 116470486 A CN116470486 A CN 116470486A
- Authority
- CN
- China
- Prior art keywords
- photovoltaic power
- data
- original
- prediction model
- data set
- 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.)
- Pending
Links
- 238000010248 power generation Methods 0.000 title claims abstract description 72
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000003860 storage Methods 0.000 title claims abstract description 18
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 62
- 238000012549 training Methods 0.000 claims abstract description 58
- 238000012545 processing Methods 0.000 claims abstract description 22
- 238000011156 evaluation Methods 0.000 claims description 37
- 238000012360 testing method Methods 0.000 claims description 27
- 230000006870 function Effects 0.000 claims description 17
- 238000004590 computer program Methods 0.000 claims description 13
- 238000012795 verification Methods 0.000 claims description 12
- 230000015654 memory Effects 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 6
- 230000009466 transformation Effects 0.000 claims description 3
- 238000010200 validation analysis Methods 0.000 claims description 3
- 238000013527 convolutional neural network Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 9
- 230000006872 improvement Effects 0.000 description 6
- 238000004422 calculation algorithm Methods 0.000 description 5
- 238000010606 normalization Methods 0.000 description 5
- 238000007781 pre-processing Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 230000005855 radiation Effects 0.000 description 3
- 238000013519 translation Methods 0.000 description 3
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 239000007789 gas Substances 0.000 description 1
- 239000005431 greenhouse gas Substances 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000001556 precipitation Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/004—Generation forecast, e.g. methods or systems for forecasting future energy generation
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/38—Arrangements for parallely feeding a single network by two or more generators, converters or transformers
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2300/00—Systems for supplying or distributing electric power characterised by decentralized, dispersed, or local generation
- H02J2300/20—The dispersed energy generation being of renewable origin
- H02J2300/22—The renewable source being solar energy
- H02J2300/24—The renewable source being solar energy of photovoltaic origin
-
- 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)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Power Engineering (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method, a device and a storage medium for predicting ultra-short-term photovoltaic power generation, which are used for generating a plurality of original data sequences by acquiring data related to a centralized photovoltaic power station, wherein the data comprise basic data, historical operation data, historical meteorological data and meteorological forecast data of the centralized photovoltaic power station; determining a preset wavelet decomposition layer number, performing wavelet decomposition processing on the original data sequence, and generating a sample data set; training and optimizing a preset original prediction model according to the sample data set to obtain a target prediction model; the target prediction model is of a CNN-LSTM network structure and is used for predicting photovoltaic power generation power of the centralized photovoltaic power station within a preset time length in the future. By adopting the method, the accuracy of the ultra-short-term photovoltaic power generation power prediction of the photovoltaic power station can be effectively improved, and the rationality of real-time dispatching of the power grid is improved.
Description
Technical Field
The invention relates to the technical field of power transmission networks, in particular to a method and a device for predicting ultra-short-term photovoltaic power generation power and a storage medium.
Background
Along with the rapid development of the economy and society, the production and consumption of energy resources are rapidly increased. The consumption of fossil energy not only leads to the exhaustion of resources, but also generates a great deal of greenhouse gases and harmful gases, endangering the ecological environment and human health. For this reason, the intensive development and utilization of renewable energy sources, in particular, new energy sources represented by wind power and photovoltaic power generation, has become an important strategy for sustainable development of the world economy and society. The grid connection of large-scale photovoltaic power generation enables the morphological structure and the tide distribution of a power system to be changed deeply, and the characteristic of random fluctuation of the output of the power system brings serious challenges to planning, running, scheduling and controlling of a power grid. The accurate ultra-short term photovoltaic power generation power prediction not only can provide basis for power grid dispatching decision, but also can provide support for multi-energy complementary coordination control of wind, light, water, fire and storage, and is one of key technologies for improving power grid consumption and large-scale new energy power generation.
At present, more research methods aiming at energy scheduling exist, photovoltaic power generation ultra-short-term prediction needs to be capable of providing photovoltaic power generation data within 4 hours, and compared with medium-long-term prediction and comparison, the predicted time period is shorter and is closer to the prediction time, and the predicted time period can be used as the basis of power grid real-time scheduling, so that the prediction precision, particularly the tracking prediction capability of rapid power fluctuation in a non-time long weather state, is higher. The current model algorithm commonly used in ultra-short-term prediction of photovoltaic power generation comprises the following steps: machine learning theory represented by artificial neural network and SVM, and time series prediction algorithm represented by autoregressive integral-sliding average model. However, the inventors found that the prior art has at least the following problems: the method is characterized in that the method is subjected to weather state change and is in shape response, irradiance and power data have various ultra-short-term fluctuation characteristics, the existing prediction algorithm has limitations on ultra-short-term photovoltaic power generation power prediction, and the prediction structure is not accurate enough.
Disclosure of Invention
The embodiment of the invention aims to provide a method, a device and a storage medium for predicting ultra-short-term photovoltaic power generation power, which can effectively improve the accuracy of the prediction of the ultra-short-term photovoltaic power generation power of a photovoltaic power station, thereby improving the rationality of real-time dispatching of a power grid.
In order to achieve the above object, an embodiment of the present invention provides a method for predicting ultra-short term photovoltaic power generation, including:
acquiring data related to a centralized photovoltaic power station, and generating a plurality of original data sequences; wherein the data related to the centralized photovoltaic power plant comprises base data, historical operating data, historical meteorological data and weather forecast data of the centralized photovoltaic power plant;
determining a preset wavelet decomposition layer number, performing wavelet decomposition processing on the original data sequence, and generating a sample data set;
training and optimizing a preset original prediction model according to the sample data set to obtain a target prediction model; the target prediction model is of a CNN-LSTM network structure and is used for predicting photovoltaic power generation power of the centralized photovoltaic power station within a preset time length in the future.
As an improvement of the above solution, training and optimizing a preset original prediction model according to the sample data set to obtain a target prediction model, specifically:
dividing the sample dataset into a training dataset, a validation dataset and a test dataset;
and training and optimizing a preset original prediction model according to the training data set, the verification data set and the test data set to obtain a target prediction model.
As an improvement of the above solution, training and optimizing a preset original prediction model according to the training data set, the verification data set and the test data set to obtain a target prediction model, specifically:
training a preset original prediction model according to the training data set;
optimizing the trained original prediction model according to the verification data set;
and performing precision evaluation on the optimized original prediction model by adopting the test data set to obtain a target prediction model meeting the precision evaluation index requirement.
As an improvement of the above solution, the determining the number of preset wavelet decomposition layers performs wavelet decomposition processing on the original data sequence to generate a sample data set, specifically:
determining a preset wavelet decomposition layer number k, and performing wavelet decomposition treatment on the original data sequence;
generating the sample data set according to each original data sequence and a stable first sub-data sequence and k high-fluctuation second sub-data sequences obtained by wavelet decomposition processing of each original data sequence; wherein k is greater than or equal to 1.
As an improvement of the above solution, training a preset original prediction model according to the training data set specifically includes:
respectively constructing corresponding sub-prediction models according to an original data sequence in the training data set, the first sub-data sequence corresponding to the original data sequence and k second sub-data sequences, and obtaining a prediction result of each sub-prediction model; the sub-prediction model is of a CNN-LSTM network structure;
performing discrete wavelet inverse transformation on the prediction results of all the sub-prediction models to obtain a final prediction result of the original data sequence;
and calculating a loss function according to a final prediction result of the original data sequence, and carrying out iterative adjustment on parameters of the original prediction model according to the loss function until the loss function meets a preset loss function threshold requirement so as to obtain the trained original prediction model.
As an improvement of the above solution, the method uses the test data set to perform precision evaluation on the optimized original prediction model, so as to obtain a target prediction model meeting the precision evaluation index requirement, which specifically includes:
determining an accuracy evaluation index; wherein, the precision evaluation index is a determination coefficient R2 and a root mean square error RMSE;
inputting the original data sequence in the test data set into the optimized original prediction model, and calculating to obtain a prediction result;
calculating the precision evaluation index according to the prediction result and the real result of the original data sequence in the test data set:
and taking the original prediction model corresponding to the precision evaluation index meeting the preset numerical requirement as the target prediction model meeting the precision evaluation index requirement.
As an improvement of the above-mentioned scheme, the calculation formula of the determination coefficient R2 in the precision evaluation index is:
the calculation formula of the Root Mean Square Error (RMSE) in the precision evaluation index is as follows:
wherein m is the number of samples of the original data sequence in the test dataset, y i As a result of the fact that,for prediction result +.>Is the mean value of the real result.
The embodiment of the invention also provides a device for predicting the ultra-short-term photovoltaic power generation power, which comprises the following steps:
the system comprises an original data sequence acquisition module, a data processing module and a data processing module, wherein the original data sequence acquisition module is used for acquiring data related to a centralized photovoltaic power station and generating a plurality of original data sequences; wherein the data related to the centralized photovoltaic power plant comprises base data, historical operating data, historical meteorological data and weather forecast data of the centralized photovoltaic power plant;
the wavelet decomposition module is used for determining the preset wavelet decomposition layer number, carrying out wavelet decomposition processing on the original data sequence and generating a sample data set;
the model training module is used for training and optimizing a preset original prediction model according to the sample data set to obtain a target prediction model; the target prediction model is of a CNN-LSTM network structure and is used for predicting photovoltaic power generation power of the centralized photovoltaic power station within a preset time length in the future.
The embodiment of the invention also provides a device for predicting the ultra-short-term photovoltaic power generation power, which comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, wherein the method for predicting the ultra-short-term photovoltaic power generation power is realized when the processor executes the computer program.
The embodiment of the invention also provides a computer readable storage medium, which comprises a stored computer program, wherein the computer program is used for controlling equipment where the computer readable storage medium is located to execute the method for predicting the ultra-short-term photovoltaic power generation power according to any one of the above.
Compared with the prior art, the ultra-short-term photovoltaic power generation power prediction method, the ultra-short-term photovoltaic power generation power prediction device and the storage medium disclosed by the invention generate a plurality of original data sequences by acquiring data related to a centralized photovoltaic power station; wherein the data related to the centralized photovoltaic power plant comprises base data, historical operating data, historical meteorological data and weather forecast data of the centralized photovoltaic power plant; determining a preset wavelet decomposition layer number, performing wavelet decomposition processing on the original data sequence, and generating a sample data set; training and optimizing a preset original prediction model according to the sample data set to obtain a target prediction model; the target prediction model is of a CNN-LSTM network structure and is used for predicting photovoltaic power generation power of the centralized photovoltaic power station within a preset time length in the future. By adopting the technical means of the embodiment of the invention, the multisource data is processed by wavelet decomposition through combining the photovoltaic power station parameter information, the historical power generation data and the weather forecast data, and then the ultra-short-term photovoltaic power generation power forecast value in a future period is obtained through the CNN-LSTM space-time neural network forecast model, so that the forecast precision is effectively improved. From the aspect of scheduling planning, scientific basis can be provided for real-time scheduling of a power grid, power generation planning in different time scales, unit combination optimization of a power system, equipment overhaul scheduling and the like; from the aspect of operation control, the photovoltaic power generation power prediction is matched with power grid dispatching to realize the maximum consumption of new energy power, and technical support can be provided for the coordinated control and the optimized operation of wind, light, water, fire, storage and other energy power generation; from the perspective of photovoltaic station operators, accurate power prediction can not only increase the number of station power generation hours and the capacity utilization rate and reduce economic penalty caused by prediction deviation, but also provide references for reasonably arranging maintenance and overhaul of power generation units and inverters, thereby improving the economic benefit and the return on investment of photovoltaic station operation.
Drawings
Fig. 1 is a schematic flow chart of a method for predicting ultra-short-term photovoltaic power generation power according to an embodiment of the present invention;
FIG. 2 is a flow chart of a preferred method for predicting ultra-short term photovoltaic power generation in accordance with an embodiment of the present invention;
FIG. 3 is a schematic representation of a wavelet decomposition tree in an embodiment of the present invention;
FIG. 4 is a schematic diagram of training principles based on a wavelet decomposition model in an embodiment of the present invention;
FIG. 5 is a schematic diagram of a prediction model of a CNN-LSTM network structure model in an embodiment of the invention;
fig. 6 is a schematic structural diagram of an apparatus for predicting ultra-short term photovoltaic power generation power according to an embodiment of the present invention;
fig. 7 is a schematic structural diagram of another device for predicting ultra-short-term photovoltaic power generation power according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1, a flow chart of a method for predicting ultra-short term photovoltaic power generation power according to an embodiment of the present invention is shown. The embodiment of the invention provides a method for predicting ultra-short-term photovoltaic power generation power, which is specifically implemented through steps S11 to S13:
s11, acquiring data related to a centralized photovoltaic power station, and generating a plurality of original data sequences; wherein the data related to the centralized photovoltaic power plant comprises base data, historical operating data, historical meteorological data and weather forecast data of the centralized photovoltaic power plant;
s12, determining the number of preset wavelet decomposition layers, and performing wavelet decomposition processing on the original data sequence to generate a sample data set;
s13, training and optimizing a preset original prediction model according to the sample data set to obtain a target prediction model; the target prediction model is of a CNN-LSTM network structure and is used for predicting photovoltaic power generation power of the centralized photovoltaic power station within a preset time length (4 h) in the future.
In the embodiment of the invention, photovoltaic data of a centralized photovoltaic power station is firstly obtained, the photovoltaic data comprises basic data and historical operation data, historical meteorological data and weather forecast data of an area where the centralized photovoltaic power station is located are obtained, an original data sequence is obtained through preprocessing, the original data sequence refers to complete and consistent time sequence data, and the time sequence data which can be directly input into a model is generated according to the construction of a model input step length.
And further, preprocessing an original model input data sequence through wavelet decomposition, constructing a training sample data set of an original prediction model, constructing, training and optimizing the original prediction model to obtain a final target photovoltaic prediction model, so that photovoltaic power generation in a future period of time is predicted in real time.
Optionally, the basic data of the chinese photovoltaic power plant includes: the centralized photovoltaic power station basic data comprises power station installed capacity (MW), photovoltaic power station geographic position (longitude and latitude), occupied area, photovoltaic alternating current combiner box, inverter parameters and the like. The historical operating data of the centralized photovoltaic power station comprises: the historical output power (MW) recorded by the photovoltaic power station is 15 minutes at intervals and is used as a target variable to be predicted. In addition, meteorological variables recorded by a small meteorological station of the photovoltaic power station are included, including radiation value (W/m 2), wind speed (m/s), air temperature (DEG C), air pressure (KPa) and relative humidity (%), and the time interval is 5 minutes. The historical meteorological data refers to: historical meteorological data uses ERA5 as part of the input variables for model training. ERA5 is a fifth generation atmospheric re-analysis product of global climate that combines model data with observations from all over the world to form a global complete, consistent data set. The time resolution is divided into hour level, day level, month level and the like, the space resolution is about 0.1-0.5 degrees, and the content of the time resolution comprises common meteorological data such as humidity, wind speed, temperature, precipitation and the like. According to research, ERA5 data with time resolution of 0.1 DEG per hour and spatial resolution are selected, and data variables comprise temperature, dew point temperature, air pressure and downlink short wave radiation. The weather forecast data refers to: the weather forecast data adopts GFS data as part of input data when model prediction is performed. The GFS distributes weather forecast data on the global scale for 4 times every day, forecast weather change conditions of 16 days in the future worldwide, the forecast spatial resolution can reach 0.25 degrees x 0.25 degrees, the time resolution of the first 5 days is 1 hour, and the resolution of the rest time is 3 hours. The variables selected include temperature, dew point temperature, air pressure and downlink short wave radiation.
As a preferred embodiment, after step S11, before step S12, the prediction method includes the steps of S11':
s11', preprocessing the data related to the centralized photovoltaic power station; wherein the preprocessing includes, but is not limited to, data fusion, outlier processing, feature construction and normalization processing.
Referring to fig. 2, a flow chart of a preferred method for predicting ultra-short term photovoltaic power generation power according to an embodiment of the present invention is shown, where the process of preprocessing data related to a centralized photovoltaic power station includes: according to the geographical position information and the occupied area of the centralized photovoltaic power stations, the centralized photovoltaic power stations in the area are clustered according to the distance, and the photovoltaic power stations in the same collection use the same weather forecast data, so that the collection of each photovoltaic station should not be larger than 0.25 degrees. Because ERA5 and GFS data are global meteorological data and are inconvenient to input as model parameters, the meteorological data of the area where the photovoltaic power station is located need to be cut out, and the average meteorological conditions of the area are represented by the average value in space. Because the original data has missing values, the time resolutions of the power data and the weather data are inconsistent, the time interval of the power data is 15 minutes, the time interval of the ERA5 historical data is 1 hour, and the time interval of the GFS weather forecast data is 1 hour or 3 hours, the original data needs to be preprocessed. Firstly, ERA5 and GFS meteorological data are downsampled to 15 minutes, the interpolation method adopts linear interpolation, the power data are also interpolated, null value conditions in the data are avoided, and the power data and the meteorological data with consistent time resolution are obtained. In order to enable the prediction model to learn the rule, the input variables are added with time variables including the days, months, days of each month, hours and minutes of each day of each year. Because the physical meaning and the numerical range of different types of input variables are different, the excessive difference can cause different influences of different features on model parameters, and even the model can not be converged, and therefore each feature needs to be independently normalized before the model is input. The most commonly used data normalization method is normalized by maximum and minimum values and mean variance, and the scheme adopts the maximum and minimum values for normalization. Firstly, statistics of histogram distribution conditions of various variables are carried out, and the maximum value and the minimum value are independently determined and then normalized by using the following formula:
V norm ＝(V-V min )/(V max -V min )；
wherein V is the value before normalization, V min And V max For minimum and maximum values, V norm Is the value after normalization.
Furthermore, after the complete and consistent time series data is obtained, the data which can be directly input into the model is generated by constructing and generating according to the input step length of the model, and the original data sequence is obtained.
As a preferred embodiment, in step S13, the training and optimizing the preset original prediction model according to the sample dataset to obtain a target prediction model, specifically:
dividing the sample dataset into a training dataset, a validation dataset and a test dataset;
and training and optimizing a preset original prediction model according to the training data set, the verification data set and the test data set to obtain a target prediction model.
Specifically, training and optimizing a preset original prediction model according to the training data set, the verification data set and the test data set to obtain a target prediction model, which specifically includes:
training a preset original prediction model according to the training data set;
optimizing the trained original prediction model according to the verification data set;
and performing precision evaluation on the optimized original prediction model by adopting the test data set to obtain a target prediction model meeting the precision evaluation index requirement.
It should be noted that, the training set is data containing reference answers, and is used for training labeled data of the model, building the model and finding rules; the verification set is a sample set independently reserved in the model training process and is used for adjusting the hyper-parameters of the model and carrying out preliminary evaluation on the capacity of the model; the test set is used for evaluating the generalization capability of the final model of the model, marked data, the common practice is to hide the marks, transmit the marks to the trained model, compare the result with the true marks, evaluate the learning capability of the model, and make the final precision evaluation.
Optionally, the sample data set is divided into a training data set, a verification data set and a test data set according to the embodiment of the present invention, and the sample data set is respectively 60%, 20% and 20%.
In a preferred embodiment, in step S12, the determining the number of preset wavelet decomposition layers performs wavelet decomposition processing on the original data sequence to generate a sample data set, specifically:
determining a preset wavelet decomposition layer number k, and performing wavelet decomposition treatment on the original data sequence;
generating the sample data set according to each original data sequence and a stable first sub-data sequence and k high-fluctuation second sub-data sequences obtained by wavelet decomposition processing of each original data sequence; wherein k is greater than or equal to 1.
On the basis, training a preset original prediction model according to the training data set, specifically:
respectively constructing corresponding sub-prediction models according to an original data sequence in the training data set, the first sub-data sequence corresponding to the original data sequence and k second sub-data sequences, and obtaining a prediction result of each sub-prediction model; the sub-prediction model is of a CNN-LSTM network structure;
performing discrete wavelet inverse transformation on the prediction results of all the sub-prediction models to obtain a final prediction result of the original data sequence;
and calculating a loss function according to a final prediction result of the original data sequence, and carrying out iterative adjustment on parameters of the original prediction model according to the loss function until the loss function meets a preset loss function threshold requirement so as to obtain the trained original prediction model.
In particular, wavelet theory has significant advantages in performing multi-scale information processing, which has evolved to provide an effective tool for analysis of complex data sequences. The original data sequence can be decomposed into two parts, an approximation component and a detail component, by discrete wavelet transform (diserete wavelet transform, DWT). Wherein the approximation component can be seen as a low frequency part sub-sequence in the original data sequence and the detail component as a high frequency part sub-sequence in the original data sequence, a process called wavelet decomposition. The two component sub-sequences obtained by wavelet decomposition may also be used to further extract the low frequency and high frequency component sub-sequences by wavelet decomposition. According to the principle, for photovoltaic power prediction under non-fixed weather conditions, an effective method for improving prediction accuracy is to preprocess an original model input data sequence through wavelet decomposition, so as to obtain a stable subsequence and a plurality of subsequences with high volatility. Compared with the original data sequence, the subsequences obtained after decomposition can better represent the characteristics of the original data from different dimensions, and more accurate prediction results can be obtained through targeted prediction modeling.
For a given mother wavelet function ψ (t) and the corresponding scale functionThe wavelet sequences ψj, k (t) and the scale function +.>
Wherein t is a time index; j and k represent scale and translation variables, respectively. The original training data sequence s (t) can be expressed as:
where cj, k represents the approximate component coefficients at a scale of j and a translation scale of k; dj, k represents the detail component coefficient when the scale is j and the translation scale is k; n is the length of the data sequence; j is the number of wavelet decomposition layers.
According to the fast discrete wavelet transform algorithm, the approximate component and the detail component under the characteristic layer wavelet decomposition can be obtained through a plurality of low-pass filters and high-pass filters. Referring to fig. 3, which is a schematic diagram of a wavelet decomposition tree according to an embodiment of the present invention, the original data sequence s is first decomposed into two parts by a high pass filter and a low pass filter: the approximation component A1 and the detail component D1 corresponding to the 1-layer wavelet decomposition are then further decomposed into a second-order approximation component A2 and a corresponding detail component D2, and the approximation component A1 may be further decomposed into A3 and D3, and so on.
Therefore, for a specific wavelet decomposition level k, the sub-sequence of the original data sequence s finally obtained through the wavelet decomposition process has a kth order approximation component AK, and detail components D1-DK. Once the number of wavelet decomposition layers k is determined, the sub-sequences Ak and D1-DK can be calculated by a series of discrete wavelet transforms on the original sequence. All of these subsequences may be predicted subsequently using a neural network model, which in embodiments of the present invention uses a CNN-LSTM network for training and prediction. And then carrying out inverse discrete wavelet transform on all the predicted result sequences to obtain predicted results corresponding to the original sequences.
Referring to fig. 4 and fig. 5, fig. 4 is a schematic diagram of training based on a wavelet decomposition model in the embodiment of the present invention, and fig. 5 is a schematic diagram of a prediction model structure of a CNN-LSTM network structure model in the embodiment of the present invention, according to the data prediction method based on k-layer wavelet decomposition provided in the embodiment of the present invention, the ultra-short term prediction of the photovoltaic power generation may be implemented by using multi-layer wavelet decomposition on an original data sequence, and 1-k-layer wavelet decomposition is performed on the original data sequence respectively, and a corresponding prediction model is constructed, so as to obtain a multiple parallel prediction result.
Taking k=5 as an example, a prediction model can be constructed for an original data sequence, a 1-layer wavelet decomposition subsequence, a 2-layer wavelet decomposition subsequence, a 3-layer wavelet decomposition subsequence, a 4-layer wavelet decomposition subsequence and a 5-layer wavelet decomposition subsequence, and the prediction model adopts a CNN-LSTM network structure. Convolutional Neural Networks (CNNs) may be used as encoders in encoder-decoder architectures. CNNs do not directly support sequence input, but one-dimensional CNNs are able to read sequence input and automatically learn salient features. These contents can then be interpreted by the LSTM decoder. The hybrid model of CNN and LSTM is called CNN-LSTM model, and is used together in the encoder-decoder architecture. CNN expects the input data to have the same 3D structure as the LSTM model, although multiple features are read as different channels, the effect is the same.
CNN and LSTM are the mainstream algorithms of current deep learning, and in contrast, CNN is more suitable for extracting local features of data and combining and abstracting the local features into high-level features, while LSTM is more suitable for time expansion, has long-term memory function, and is more suitable for processing time series data. Therefore, the embodiment of the invention combines CNN and LSTM in a mode of a mixed model to construct a photovoltaic prediction model based on CNN-LSTM, so that the model has the capability of extracting potential characteristics of time sequence data and predicting the time sequence data, and the structure of the CNN-LSTM prediction model is shown as figure 5 and comprises an input layer, a one-dimensional convolution layer, a bidirectional LSTM layer and three full-connection layers.
As a preferred embodiment, the performing precision evaluation on the optimized original prediction model by using the test data set to obtain a target prediction model meeting the precision evaluation index requirement, specifically:
determining an accuracy evaluation index; wherein, the precision evaluation index is a determination coefficient R2 and a root mean square error RMSE;
inputting the original data sequence in the test data set into the optimized original prediction model, and calculating to obtain a prediction result;
calculating the precision evaluation index according to the prediction result and the real result of the original data sequence in the test data set:
and taking the original prediction model corresponding to the precision evaluation index meeting the preset numerical requirement as the target prediction model meeting the precision evaluation index requirement.
Specifically, the calculation formula of the determination coefficient R2 in the precision evaluation index is as follows:
the calculation formula of the Root Mean Square Error (RMSE) in the precision evaluation index is as follows:
wherein m is the number of samples of the original data sequence in the test dataset, y i As a result of the fact that,for prediction result +.>Is the mean value of the real result.
In the embodiment of the invention, in order to improve the prediction precision, the model hyper-parameters are required to be adjusted by combining the training set, the verification set and the test set, and the optimal model is obtained through iterative training and comparison, so that the situation of over-fitting or under-fitting is avoided. Parameters that can be adjusted include: data step size, model structure, batch size, learning rate, epoch number, cost function, optimizer, etc. Among these, MSE (Mean Square Error) is the most common loss function for regression models, and the optimizer selects Adam (Adaptive moment estimation) the optimizers. The model is a regression prediction problem, so that a decision coefficient R2 and a Root Mean Square Error (RMSE) are selected as precision evaluation indexes to provide basis for parameter selection and final precision evaluation in the training process. The closer R2 is to 1, the smaller the RMSE, the higher the accuracy of the representation, and the smaller the error.
The embodiment of the invention provides a prediction method of ultra-short-term photovoltaic power generation power, which generates a plurality of original data sequences by acquiring data related to a centralized photovoltaic power station; wherein the data related to the centralized photovoltaic power plant comprises base data, historical operating data, historical meteorological data and weather forecast data of the centralized photovoltaic power plant; determining a preset wavelet decomposition layer number, performing wavelet decomposition processing on the original data sequence, and generating a sample data set; training and optimizing a preset original prediction model according to the sample data set to obtain a target prediction model; the target prediction model is of a CNN-LSTM network structure and is used for predicting photovoltaic power generation power of the centralized photovoltaic power station within a preset time length in the future. By adopting the technical means of the embodiment of the invention, the multisource data is processed by wavelet decomposition through combining the photovoltaic power station parameter information, the historical power generation data and the weather forecast data, and then the ultra-short-term photovoltaic power generation power forecast value in a future period is obtained through the CNN-LSTM space-time neural network forecast model, so that the forecast precision is effectively improved. From the aspect of scheduling planning, scientific basis can be provided for real-time scheduling of a power grid, power generation planning in different time scales, unit combination optimization of a power system, equipment overhaul scheduling and the like; from the aspect of operation control, the photovoltaic power generation power prediction is matched with power grid dispatching to realize the maximum consumption of new energy power, and technical support can be provided for the coordinated control and the optimized operation of wind, light, water, fire, storage and other energy power generation; from the perspective of photovoltaic station operators, accurate power prediction can not only increase the number of station power generation hours and the capacity utilization rate and reduce economic penalty caused by prediction deviation, but also provide references for reasonably arranging maintenance and overhaul of power generation units and inverters, thereby improving the economic benefit and the return on investment of photovoltaic station operation.
Referring to fig. 6, a schematic structural diagram of an apparatus for predicting ultra-short-term photovoltaic power provided by an embodiment of the present invention, an embodiment of the present invention provides an apparatus 20 for predicting ultra-short-term photovoltaic power, including:
the original data sequence acquisition module 21 is used for acquiring data related to the centralized photovoltaic power station and generating a plurality of original data sequences; wherein the data related to the centralized photovoltaic power plant comprises base data, historical operating data, historical meteorological data and weather forecast data of the centralized photovoltaic power plant;
the wavelet decomposition module 22 is configured to determine a preset number of wavelet decomposition layers, perform wavelet decomposition processing on the original data sequence, and generate a sample data set;
the model training module 23 is configured to train and optimize a preset original prediction model according to the sample data set, so as to obtain a target prediction model; the target prediction model is of a CNN-LSTM network structure and is used for predicting photovoltaic power generation power of the centralized photovoltaic power station within a preset time length in the future.
It should be noted that, the device for predicting the ultra-short-term photovoltaic power generation provided by the embodiment of the present invention is used for executing all the flow steps of the method for predicting the ultra-short-term photovoltaic power generation in the above embodiment, and the working principles and beneficial effects of the two correspond one to one, so that the description is omitted.
Referring to fig. 7, which is a schematic structural diagram of another device for predicting ultra-short-term photovoltaic power generation provided by an embodiment of the present invention, the embodiment of the present invention further provides a device 30 for predicting ultra-short-term photovoltaic power generation, which includes a processor 31, a memory 32, and a computer program stored in the memory and configured to be executed by the processor, where the processor executes the computer program to implement the method for predicting ultra-short-term photovoltaic power generation according to any one of the embodiments.
The embodiment of the invention also provides a computer readable storage medium, which comprises a stored computer program, wherein when the computer program runs, equipment where the computer readable storage medium is located is controlled to execute the method for predicting the ultra-short-term photovoltaic power generation power according to any one of the embodiments.
Those skilled in the art will appreciate that implementing all or part of the above-described methods in accordance with the embodiments may be accomplished by way of a computer program stored on a computer readable storage medium, which when executed may comprise the steps of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), or the like.
While the foregoing is directed to the preferred embodiments of the present invention, it will be appreciated by those skilled in the art that changes and modifications may be made without departing from the principles of the invention, such changes and modifications are also intended to be within the scope of the invention.
Claims (10)
1. The method for predicting the ultra-short-term photovoltaic power generation power is characterized by comprising the following steps of:
acquiring data related to a centralized photovoltaic power station, and generating a plurality of original data sequences; wherein the data related to the centralized photovoltaic power plant comprises base data, historical operating data, historical meteorological data and weather forecast data of the centralized photovoltaic power plant;
determining a preset wavelet decomposition layer number, performing wavelet decomposition processing on the original data sequence, and generating a sample data set;
training and optimizing a preset original prediction model according to the sample data set to obtain a target prediction model; the target prediction model is of a CNN-LSTM network structure and is used for predicting photovoltaic power generation power of the centralized photovoltaic power station within a preset time length in the future.
2. The method for predicting ultra-short term photovoltaic power generation according to claim 1, wherein the training and optimizing the preset original prediction model according to the sample dataset to obtain a target prediction model is specifically as follows:
dividing the sample dataset into a training dataset, a validation dataset and a test dataset;
and training and optimizing a preset original prediction model according to the training data set, the verification data set and the test data set to obtain a target prediction model.
3. The method for predicting ultra-short term photovoltaic power generation according to claim 2, wherein the training and optimizing the preset original prediction model according to the training data set, the verification data set and the test data set to obtain a target prediction model specifically comprises:
training a preset original prediction model according to the training data set;
optimizing the trained original prediction model according to the verification data set;
and performing precision evaluation on the optimized original prediction model by adopting the test data set to obtain a target prediction model meeting the precision evaluation index requirement.
4. The method for predicting ultra-short term photovoltaic power generation according to claim 3, wherein the determining the number of preset wavelet decomposition layers performs wavelet decomposition processing on the original data sequence to generate a sample data set, specifically:
determining a preset wavelet decomposition layer number k, and performing wavelet decomposition treatment on the original data sequence;
generating the sample data set according to each original data sequence and a stable first sub-data sequence and k high-fluctuation second sub-data sequences obtained by wavelet decomposition processing of each original data sequence; wherein k is greater than or equal to 1.
5. The method for predicting ultra-short term photovoltaic power generation according to claim 4, wherein training a preset original prediction model according to the training data set is specifically as follows:
respectively constructing corresponding sub-prediction models according to an original data sequence in the training data set, the first sub-data sequence corresponding to the original data sequence and k second sub-data sequences, and obtaining a prediction result of each sub-prediction model; the sub-prediction model is of a CNN-LSTM network structure;
performing discrete wavelet inverse transformation on the prediction results of all the sub-prediction models to obtain a final prediction result of the original data sequence;
and calculating a loss function according to a final prediction result of the original data sequence, and carrying out iterative adjustment on parameters of the original prediction model according to the loss function until the loss function meets a preset loss function threshold requirement so as to obtain the trained original prediction model.
6. The method for predicting ultra-short term photovoltaic power generation according to any one of claims 3 to 5, wherein the precision evaluation is performed on the optimized original prediction model by using the test data set to obtain a target prediction model meeting the precision evaluation index requirement, specifically:
determining an accuracy evaluation index; wherein, the precision evaluation index is a determination coefficient R2 and a root mean square error RMSE;
inputting the original data sequence in the test data set into the optimized original prediction model, and calculating to obtain a prediction result;
calculating the precision evaluation index according to the prediction result and the real result of the original data sequence in the test data set:
and taking the original prediction model corresponding to the precision evaluation index meeting the preset numerical requirement as the target prediction model meeting the precision evaluation index requirement.
7. The method for predicting ultra-short term photovoltaic power generation according to claim 6, wherein the calculation formula of the determination coefficient R2 in the precision evaluation index is:
the calculation formula of the Root Mean Square Error (RMSE) in the precision evaluation index is as follows:
wherein m is the number of samples of the original data sequence in the test dataset, y i As a result of the fact that,to predict result，/>Is the mean value of the real result.
8. An ultra-short term photovoltaic power generation power prediction device, comprising:
the system comprises an original data sequence acquisition module, a data processing module and a data processing module, wherein the original data sequence acquisition module is used for acquiring data related to a centralized photovoltaic power station and generating a plurality of original data sequences; wherein the data related to the centralized photovoltaic power plant comprises base data, historical operating data, historical meteorological data and weather forecast data of the centralized photovoltaic power plant;
the wavelet decomposition module is used for determining the preset wavelet decomposition layer number, carrying out wavelet decomposition processing on the original data sequence and generating a sample data set;
the model training module is used for training and optimizing a preset original prediction model according to the sample data set to obtain a target prediction model; the target prediction model is of a CNN-LSTM network structure and is used for predicting photovoltaic power generation power of the centralized photovoltaic power station within a preset time length in the future.
9. A prediction device of ultra-short term photovoltaic power generation, characterized by comprising a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, the processor implementing the prediction method of ultra-short term photovoltaic power generation according to any one of claims 1 to 7 when executing the computer program.
10. A computer readable storage medium, characterized in that the computer readable storage medium comprises a stored computer program, wherein the computer program, when run, controls a device in which the computer readable storage medium is located to perform the method of predicting ultra-short term photovoltaic power generation according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310399296.8A CN116470486A (en) | 2023-04-06 | 2023-04-06 | Ultra-short-term photovoltaic power generation power prediction method, device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310399296.8A CN116470486A (en) | 2023-04-06 | 2023-04-06 | Ultra-short-term photovoltaic power generation power prediction method, device and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116470486A true CN116470486A (en) | 2023-07-21 |
Family
ID=87183752
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310399296.8A Pending CN116470486A (en) | 2023-04-06 | 2023-04-06 | Ultra-short-term photovoltaic power generation power prediction method, device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116470486A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116722548A (en) * | 2023-08-09 | 2023-09-08 | 深圳航天科创泛在电气有限公司 | Photovoltaic power generation prediction method based on time sequence model and related equipment |
-
2023
- 2023-04-06 CN CN202310399296.8A patent/CN116470486A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116722548A (en) * | 2023-08-09 | 2023-09-08 | 深圳航天科创泛在电气有限公司 | Photovoltaic power generation prediction method based on time sequence model and related equipment |
| CN116722548B (en) * | 2023-08-09 | 2023-12-29 | 深圳航天科创泛在电气有限公司 | Photovoltaic power generation prediction method based on time sequence model and related equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110070226B (en) | Photovoltaic power prediction method and system based on convolutional neural network and meta-learning | |
| CN109002915B (en) | Photovoltaic power station short-term power prediction method based on Kmeans-GRA-Elman model | |
| Raza et al. | An ensemble framework for day-ahead forecast of PV output power in smart grids | |
| Song et al. | Hourly heat load prediction model based on temporal convolutional neural network | |
| Rodríguez et al. | Forecasting intra-hour solar photovoltaic energy by assembling wavelet based time-frequency analysis with deep learning neural networks | |
| CN110909919A (en) | Photovoltaic power prediction method of depth neural network model with attention mechanism fused | |
| CN110942194A (en) | Wind power prediction error interval evaluation method based on TCN | |
| CN110070228B (en) | BP neural network wind speed prediction method for neuron branch evolution | |
| Han et al. | A short-term wind speed interval prediction method based on WRF simulation and multivariate line regression for deep learning algorithms | |
| Aliberti et al. | Forecasting Short-term Solar Radiation for Photovoltaic Energy Predictions. | |
| CN113269314A (en) | New energy power generation scene data migration method based on generation countermeasure network | |
| CN116470486A (en) | Ultra-short-term photovoltaic power generation power prediction method, device and storage medium | |
| CN116629416A (en) | Photovoltaic power station power prediction method and device | |
| CN113984198B (en) | Shortwave radiation prediction method and system based on convolutional neural network | |
| Kim et al. | Solar radiation forecasting based on the hybrid CNN-CatBoost model | |
| Wu et al. | Two-stage decomposition and temporal fusion transformers for interpretable wind speed forecasting | |
| CN115544895B (en) | Photovoltaic power station annual output guarantee rate model optimization method | |
| CN116151464A (en) | Photovoltaic power generation power prediction method, system and storable medium | |
| CN115526376A (en) | Multi-feature fusion generation countermeasure network ultra-short-term wind power prediction method | |
| Ma et al. | Multi-features fusion for short-term photovoltaic power prediction | |
| CN113962432A (en) | Wind power prediction method and system integrating three-dimensional convolution and light-weight convolution threshold unit | |
| CN114282336A (en) | New energy power station output scene generation method and system | |
| CN114511127A (en) | Neural network-based regional photovoltaic output characteristic long-term prediction method and system | |
| Wu et al. | Interpretable wind speed forecasting with meteorological feature exploring and two-stage decomposition | |
| Fan et al. | Ultra-Short-Term Wind Power Forecasting Based on Fluctuation Pattern Clustering and Prediction |
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 |