CN114492945A - Short-term photovoltaic power prediction method, medium and equipment in electric power market background - Google Patents
Short-term photovoltaic power prediction method, medium and equipment in electric power market background Download PDFInfo
- Publication number
- CN114492945A CN114492945A CN202111679639.3A CN202111679639A CN114492945A CN 114492945 A CN114492945 A CN 114492945A CN 202111679639 A CN202111679639 A CN 202111679639A CN 114492945 A CN114492945 A CN 114492945A
- Authority
- CN
- China
- Prior art keywords
- model
- prediction
- photovoltaic power
- weather
- layer
- 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
- 238000000034 method Methods 0.000 title claims abstract description 67
- 238000010248 power generation Methods 0.000 claims abstract description 38
- 238000012549 training Methods 0.000 claims abstract description 34
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 30
- 230000006870 function Effects 0.000 claims description 24
- 238000005070 sampling Methods 0.000 claims description 17
- 230000005611 electricity Effects 0.000 claims description 12
- 238000004590 computer program Methods 0.000 claims description 11
- 230000010354 integration Effects 0.000 claims description 11
- 230000004913 activation Effects 0.000 claims description 10
- 238000004364 calculation method Methods 0.000 claims description 9
- 238000002790 cross-validation Methods 0.000 claims description 9
- 238000013528 artificial neural network Methods 0.000 claims description 6
- 238000003860 storage Methods 0.000 claims description 6
- 239000011800 void material Substances 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 230000007717 exclusion Effects 0.000 claims description 4
- 230000004927 fusion Effects 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 4
- 238000009825 accumulation Methods 0.000 claims description 3
- 230000008878 coupling Effects 0.000 claims description 3
- 238000010168 coupling process Methods 0.000 claims description 3
- 238000005859 coupling reaction Methods 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000010200 validation analysis Methods 0.000 claims description 3
- 230000007246 mechanism Effects 0.000 abstract description 15
- 230000008901 benefit Effects 0.000 abstract description 6
- 238000004519 manufacturing process Methods 0.000 abstract description 2
- 238000004891 communication Methods 0.000 description 9
- 230000008569 process Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 6
- 230000009471 action Effects 0.000 description 3
- 230000007547 defect Effects 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 230000001364 causal effect Effects 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 241000251468 Actinopterygii Species 0.000 description 1
- 241001251371 Betula chinensis Species 0.000 description 1
- 241000486761 Chilobrachys Species 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 230000001808 coupling effect Effects 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 238000007667 floating Methods 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 230000007773 growth pattern Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 230000000704 physical effect Effects 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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
- G06N3/045—Combinations of networks
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/067—Enterprise or organisation modelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- 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)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- Tourism & Hospitality (AREA)
- General Health & Medical Sciences (AREA)
- General Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Game Theory and Decision Science (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Evolutionary Computation (AREA)
- Development Economics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Artificial Intelligence (AREA)
- Educational Administration (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a short-term photovoltaic power prediction method, medium and equipment in the power market background, wherein the method comprises the following steps of dividing weather into four weather models of sunny days, cloudy days, rain shower and full rain based on NWP values; judging the weather type of the day to be detected, and selecting the latest photovoltaic power historical data and the historical NWP value of the similar day according to the weather type of the day to be detected; inputting the classified data set into each base learner in a layer of prediction model of the Stacking integrated model; obtaining the prediction result of each base learner, constructing a new training set and inputting the new training set into a two-layer prediction model; and obtaining a prediction result of the two-layer element learner, namely the final photovoltaic power generation power prediction value. The method can effectively reduce the dependence of prediction on data by adopting a combined mechanism model and a data-driven algorithm, can give consideration to the accuracy, speed and reliability of prediction under the condition of unsatisfactory data quality and quantity, greatly reduces the cost of acquiring data in actual production, and has remarkable economic benefit.
Description
Technical Field
The invention relates to the technical field of photovoltaic power generation, in particular to a short-term photovoltaic power prediction method, medium and equipment in the power market background.
Background
The method and the device have the advantages that the photovoltaic power prediction is of great importance in the dispatching of the power system, and the research of the short-term photovoltaic power prediction method in the power market background is developed according to the requirement.
The photovoltaic power generation prediction method based on data driving is simple in modeling, mature in algorithm and high in calculation speed, but is completely based on numerical calculation, highly depends on data quality and quantity, does not consider the internal mechanism of photovoltaic power generation, and is poor in reliability of a prediction result. The mechanism driving method can well reflect the photovoltaic internal power generation principle and the coupling effect, can analyze the physical essence of photovoltaic power generation, has high reliability and universality, and has the defects of high model complexity, high calculation difficulty, strong model parameter time-varying property and the like. The existing data-driven prediction method widely applied to engineering is highly dependent on data quality and quantity, and the cost is high when comprehensive and qualified data of a new energy power generation system are obtained.
The document [1] provides an improved clear sky power model based on online updating, ultra-short-term prediction is made for photovoltaic power in small fluctuation weather, and prediction accuracy in the small fluctuation weather in a scale of 3-4 h is improved well. Document [2] proposes a photovoltaic output prediction combination model based on a self-adaptive fuzzy time sequence, which adopts a self-adaptive algorithm to process historical power data, then carries out clustering, domain-of-discourse definition, division and fuzzification on the data, finally carries out prediction by combining a fuzzy time sequence method, defuzzification on the result, carries out simulation on data with the use time interval of 15min on an actual photovoltaic experimental system, and respectively achieves the average absolute error (MAE) value and the average absolute percentage error (EMAPE) value of 1.038MW and 13.34%. Document [3] adopts an ExtraTreeseregressor method to evaluate the feature importance according to the correlation between the photovoltaic output and the external environmental factors, after cleaning and extracting feature quantities of data, an SVM based on an artificial fish swarm method is adopted to classify meteorological data, photovoltaic power generation amount of each category is predicted, and compared with a traditional SVM model, the prediction precision is greatly improved.
Patent [4] discloses a photovoltaic power generation prediction system based on a T-S type fuzzy neural network, which realizes the organic combination of a fuzzy inference system and a neural network learning system, introduces meteorological factors and effectively improves the prediction accuracy and reliability. According to the patent [5], the daily historical processing data of the photovoltaic power station is clustered into K clusters, an improved generalized weather map of one weather professional weather corresponding to one or more digital labels is constructed, and the defect of low prediction accuracy of photovoltaic power generation under a non-fine condition is overcome. The patent [6] classifies the weather by adopting a weather scoring mechanism so as to classify the power generation data, and the probability density function estimation of each type of data is solved through a kernel density function so as to provide a distribution rule in the statistical significance of the power generation data. Patent [7] discloses a photovoltaic power generation prediction system, which obtains predicted power generation capacity by detecting real-time weather conditions, the serial number, the longitude and latitude, the placement angle and photovoltaic material data of photovoltaic power generation equipment and applying a mechanism model.
[1] Mariotte, Zhangxian, Chilobrachys, etc. ultrashort-term photovoltaic power prediction method [ J ] based on modified clear sky model power system automation, 2021,45(11): 44-51;
[2] the short-term power prediction of photovoltaic power generation based on an adaptive fuzzy time series method [ J ]. the Nanjing institute of engineering (Nature science edition), 2014,12(1): 6-13;
[3] ABC-SVM and PSO-RF based photovoltaic micro-grid daily generated power combined prediction method research [ J ] solar energy science and report, 2020,41(03):177 + 183;
[4] a photovoltaic power generation prediction system [ P ] Jiangsu based on a T-S type fuzzy neural network: CN103106544A, 2013-05-15;
[5] span beam, rigor, longitudinal million dan, conception dawn, plum blossom, liu jian hua, a photovoltaic power generation prediction method based on K-means clustering improvement generalized weather [ P ]. Jiangsu province: CN106022538B, 2020-04-07;
[6] the method comprises the following steps of A, obtaining a photovoltaic power generation prediction method [ P ] based on a shape parameter confidence interval, namely, Puweichun, Panxiao, Liyu, Zhangyan and Mashahua: CN108256690B, 2021-10-08;
[7] liule, Huangle, forest, a prediction method [ P ] of a photovoltaic power generation prediction system: CN106909985B, 2021-02-09.
Disclosure of Invention
The invention provides a short-term photovoltaic power prediction method, medium and equipment in the context of the electric power market, which can at least solve one of the problems in the background art.
In order to achieve the purpose, the invention adopts the following technical scheme:
a short-term photovoltaic power prediction method in the context of the electricity market comprises the following steps,
dividing the weather into four weather models of sunny days, cloudy days, rain shower and full rain based on the NWP value;
judging the weather type of the day to be detected, and selecting the latest photovoltaic power historical data and the historical NWP value of the similar day according to the weather type of the day to be detected;
inputting the classified data set into each base learner in a layer of prediction model of the Stacking integrated model;
obtaining the prediction result of each base learner, constructing a new training set and inputting the new training set into a two-layer prediction model;
and obtaining a prediction result of the two-layer element learner, namely the final photovoltaic power generation power prediction value.
Further, based on the NWP value, dividing the weather into four weather models of sunny days, cloudy days, showers and full rains; the method specifically comprises the following steps:
selecting the cloud cover C and the rainfall p in NWP as weather parting factors, dividing the cloud cover into a sunny model and a cloudy model according to the average cloud cover in the daytime, and dividing the cloud cover into a rain gust model and a full rain model according to the rainfall duration in the daytime;
the sunny day model is
The model of cloudy day is
The model of rain shower is
The full rain model is
c1A typing threshold, t, for both sunny and cloudy models1A typing threshold value of a gust rain and full rain model; according to the short-term weather forecast national standard and the meteorological principles, c1＝0.7，t1＝4。
Further, the one-layer prediction model of the Stacking integration model comprises a physical characteristic model of photovoltaic power generation, and the physical characteristic model is shown in formula (1):
P＝ηSI[1-0.005(t+25)] (1)
in the formula, P is photovoltaic power generation power, eta is photovoltaic panel conversion efficiency, S is photovoltaic panel effective area, I is irradiance, and t is photovoltaic panel working temperature.
Furthermore, the one-layer prediction model also comprises a TCN network prediction model, the TCN network prediction model is constructed by the following steps,
and (3) a cavity convolution step, wherein each TCN layer contains L convolution layers, and the expansion convolution calculation formula is as follows:
in the formula: void factor d is (1, …, 2)L) K is the convolution kernel size;
residual linking step:
relu represents a linear rectification function used as an activation function for a neural network; DCConv denotes a void convolution layer;
equations (3), (4) represent the activation function of TCN:
in the formula: w(1)、W(2)Is a weight matrix corresponding to the input, b is an offset vector, S(i,j)And (3) representing an activation function of the ith layer of the jth block, wherein a formula (3) represents a result of hole convolution at the time t, and a formula (4) represents a result after residual error linkage is added.
Further, obtaining the prediction result of each base learner, constructing a new training set and inputting the new training set into a two-layer prediction model, and adopting the following algorithm:
the influence of high-dimensional data on prediction is relieved by adopting a LightGBM model based on a histogram algorithm; the LightGBM model is optimized by adopting a Leaf-wise algorithm with depth limitation;
the objective function of the LightGBM model is as follows:
Obj(t)＝L(t)+Ω(t)+c (5)
in the formula: obj (t) is an optimization target, and omega (t) represents a regular function and reflects the complexity of the model; t represents a sampling time; c represents additional parameters, avoiding over-fitting and optimizing tree depth;
l (t) denotes the loss function by describing the actual values y of the N sampling pointsiAnd the predicted value
To reflect the degree of fit of the model; the definition is as follows:
transmitting residual error information of a previous learner by serially coupling the regression trees; final output
Generated by the accumulation of the remaining trees.
Further, the prediction results of each base learner are obtained, a new training set is constructed and input into the two-layer prediction model, and the following algorithm is also adopted:
and a gradient-based unilateral sampling algorithm GOSS is adopted, so that during sampling, the GOSS can completely reserve large gradient samples meeting the conditions as sampled data, and samples with smaller gradients are sampled randomly.
Furthermore, the prediction results of all the base learners are obtained, a new training set is constructed and input into a two-layer prediction model, and the following algorithm is also adopted,
by adopting the method EFB for binding the mutual exclusion characteristics, after GOSS sampling, the dimensionality of the characteristics is reduced by binding the mutual exclusion characteristics so as to prevent dimensionality disasters and improve the calculation efficiency.
Furthermore, a layer of prediction model of the Stacking integration model divides the data set into a plurality of subsets through cross validation, and the evaluation results are fused, so that the variance of the prediction results of the model is reduced, the generalization capability of the model is improved, and the over-fitting phenomenon is avoided;
the K-fold cross validation is to averagely divide the data set into K parts, wherein K-1 part is used as a training set, and the rest 1 part is used as a validation set; and (5) training by using the training sets under the K conditions to obtain the model hyper-parameters.
In another aspect, the present invention also discloses a computer readable storage medium storing a computer program, which when executed by a processor causes the processor to perform the steps of the method as described above.
In yet another aspect, the present invention also discloses a computer device comprising a memory and a processor, the memory storing a computer program, the computer program, when executed by the processor, causing the processor to perform the steps of the method as described above.
According to the technical scheme, in order to overcome the defects of the prior art, the invention provides the photovoltaic power prediction method taking the mechanism model into consideration and adopting the multi-model fusion Stacking integrated learning mode. Data observation and training principle differences of different algorithms are considered, advantages of each model are fully exerted, a plurality of machine learning algorithms are built, a Stacking integrated learning photovoltaic prediction model embedded in a photovoltaic mechanism model is combined, and a base learner of the model comprises a photovoltaic physical model, a LightGBM algorithm and a time convolution network algorithm.
According to the short-term photovoltaic power prediction method in the power market background, the dependence of prediction on data can be effectively reduced by adopting a combined mechanism model and a data-driven algorithm, the prediction precision, speed and reliability can be considered under the condition that the data quality and quantity are not ideal, the cost of obtaining data in actual production is greatly reduced, and the method has remarkable economic benefit. Meanwhile, the accuracy of the short-term prediction of the new energy can be effectively improved, important support is played in the aspects of scheduling plan formulation, electric power market transaction, load management and the like, the safe and economic operation level of the power grid is further improved, and the grid-connected service quality of a power grid company for new energy power generation enterprises and users is also improved.
Drawings
FIG. 1 is a flow chart of a mechanism model and data driven joint photovoltaic power prediction according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a time convolutional network structure;
FIG. 3 is a diagram of a residual linking principle;
FIG. 4 is a schematic diagram of a histogram algorithm;
FIG. 5 is a graph of the Leaf-wise growth pattern;
FIG. 6 is a schematic diagram of a Stacking integration model;
fig. 7 is a schematic diagram of cross-validation of the present embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention.
The short-term photovoltaic power prediction method in the power market background described in this embodiment specifically includes the following steps:
1.1 mechanism model and data-driven joint photovoltaic power generation prediction
1.1.1 NWP-based weather typing
The generated power of the photovoltaic power station is influenced by the solar irradiance received in the day, the received solar irradiance is influenced by the weather type, and the photovoltaic generated power fluctuation in different weathers is different. And selecting the cloud cover C and the rainfall p in the NWP as the weather typing factors. According to average cloud cover in daytime
The method comprises a sunny model and a cloudy model, and is divided into a rain-gust model and a full-rain model according to the rainfall time t in the daytime.
TABLE 1 weather typing model
In Table 1, c1A typing threshold, t, for both sunny and cloudy models1And the classification threshold values of the rain fall and full rain models. According to the short-term weather forecast national standard and the meteorological principles, c1＝0.7，t1＝4。
1.1.2 Stacking multi-model fusion short-term photovoltaic power generation prediction considering mechanism model
Based on the principle introduction, the photovoltaic physical power generation model is embedded into a layer of prediction model of the Stacking integrated model and set as one of the base learners, and the Stacking multi-model fusion photovoltaic power generation prediction model considering the mechanism model is constructed. The implementation process of short-term photovoltaic power generation power prediction is given below, and a flow chart is shown in fig. 1.
The method comprises the following steps:
(1) and based on the NWP value, dividing the weather into four weather models of sunny days, cloudy days, showers and full rains.
(2) And judging the weather type of the day to be detected, and selecting the latest photovoltaic power historical data and the historical NWP value of the similar day according to the weather type of the day to be detected.
(3) And inputting the classified data set into each base learner in a layer of prediction model of the Stacking integration model.
(4) And obtaining the prediction result of each base learner, constructing a new training set and inputting the new training set into the two-layer prediction model.
(5) And obtaining a prediction result of the two-layer element learner, namely the final photovoltaic power generation power prediction value.
1.2 photovoltaic power generation physical model
The physical model of photovoltaic power generation predicts the generated power based on solar radiation and photoelectric conversion characteristics. The physical property model of photovoltaic power generation is shown in formula (1).
P＝ηSI[1-0.005(t+25)] (1)
In the formula, P is photovoltaic power generation power, eta is photovoltaic panel conversion efficiency, S is photovoltaic panel effective area, I is irradiance, and t is photovoltaic panel working temperature.
1.3TCN network prediction model
The main structure of TCN can be divided into causal convolution for sequences and a hole convolution plus residual model for historical data memory. Due to the fact that causal relationship exists between the layers of the convolution layer, more historical data can be memorized, and the method is suitable for data of the photovoltaic power station.
(1) Convolution of holes
The TCN structure is shown in fig. 2, each TCN layer contains L convolutional layers, and the extended convolution formula is:
in the formula: void factor d is (1, …, 2)L) And k is the convolution kernel size.
(2) Residual linking
Fig. 3 is a residual linkage diagram, and Dropout shows that the activation value of a certain neuron stops working with a certain probability during neuron propagation, so as to enhance the generalization of the model. Relu represents a linear rectification function used as an activation function for a neural network; DC Conv denotes a void convolution layer.
Equations (3), (4) represent the activation function of TCN:
in the formula: w(1)、W(2)Is a weight matrix corresponding to the input, b is an offset vector, S(i,j)And (3) representing an activation function of the ith layer of the jth block, wherein a formula (3) represents a result of hole convolution at the time t, and a formula (4) represents a result after residual error linkage is added.
1.4LightGBM
1.4.1 histogram Algorithm
The LightGBM model adopts a histogram-based algorithm to relieve the influence of high-dimensional data on prediction, improves the calculation speed and avoids the phenomenon of overfitting of the prediction model. The basic idea of histograms consists in converting successive floating point eigenvalues into k integers, giving k to a "bucket" (bin), and constructing a histogram of width k. The structure is shown in fig. 4.
1.4.2 Leaf-wise Algorithm with depth constraint
In the optimization process, the LGBM adopts a Leaf-wise algorithm to find a proper Leaf, then splits and circulates according to the Leaf-wise algorithm. As shown in fig. 5. The LGBM adds a depth limit on the Leaf-wise to prevent the over-fitting phenomenon. The target function of the LGBM is as follows:
Obj(t)＝L(t)+Ω(t)+c (5)
in the formula: obj (t) is an optimization goal, and Ω (t) represents a regular function, reflecting the complexity of the model. t represents a sampling time. c represents an additional parameter, avoiding overfitting and optimizing the tree depth.
L (t) denotes the loss function by describing the actual values y of the N sampling pointsiAnd the predicted value
To reflect the degree of fit of the model. The definition is as follows:
residual information of previous learners is transmitted by serially coupling the regression trees. Final output
Generated by the accumulation of the remaining trees.
1.4.3 unilateral gradient sampling algorithm
A gradient-based one-sided sampling algorithm (GOSS) is employed. During sampling, the GOSS can completely reserve the large-gradient samples meeting the conditions as sampled data, and adopts a random sampling mode for the samples with smaller gradients, so that the data with insufficient training can be kept to pay more attention in the next training, and the sample distribution can not be greatly changed.
1.4.4 mutually exclusive feature bundling algorithm
A mutually exclusive feature binding method (EFB) is employed. After the GOSS sampling, the dimensionality of the features is reduced by binding mutually exclusive features to prevent dimensionality disasters and improve the calculation efficiency. Because most of high-dimensional data is sparse data, most of the features in the feature space are mutually exclusive, the mutually exclusive features can be bound together to form a new feature to reduce the feature dimension.
1.5Stacking ensemble learning framework
1.5.1Stacking integration model
The Stacking integration model divides an original data set into a plurality of sub data sets, inputs the sub data sets into each base learner of the first-layer prediction model, predicts and outputs each result to form a new data set, inputs the new data set into the second-layer prediction model for training, predicts and outputs a final result, and the structure is shown in fig. 6. The Stacking integration model generalizes output results output by a plurality of models, learns combined information among features, and effectively improves the overall prediction precision.
1.5.2 Cross-validation
Cross-validation is typically used to evaluate the predictive performance of the model. A data set is divided into a plurality of subsets by a layer of prediction model of the Stacking integration model through cross validation, evaluation results are fused, variance of the prediction results of the model is reduced, generalization capability of the model is improved, and an over-fitting phenomenon is avoided. The K-fold cross validation is to averagely divide the data set into K parts, wherein K-1 part is used as a training set, and the rest 1 part is used as a validation set. And (5) training by using the training sets under the K conditions to obtain the model hyper-parameters. The process is shown in fig. 7, taking 4 folds as an example.
From the above, the mechanism and the data joint driving prediction mode of the invention can combine the domain prior knowledge model and the data-driven deep learning algorithm. An electric power knowledge model is applied, model parameters are optimized, and model adaptability is improved; the efficiency of machine learning and data mining in data driving is improved, and the machine learning generalization risk is reduced on the premise of not increasing the number of training samples. The mechanism model and the data driving model are effectively combined, organic integration of rules and experience is achieved, the advantages of the two models can be better integrated, better comprehensive performance can be achieved by adopting fewer data samples and a simpler mechanism model, good prediction precision and efficiency are guaranteed, and reliability of a prediction result is effectively improved.
In yet another aspect, the present invention also discloses a computer readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the steps of the short term photovoltaic power prediction method as described above in the context of the electricity market.
In yet another aspect, the present invention also discloses a computer device comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to perform the steps of the short term photovoltaic power prediction method as described above in the context of the electricity market.
It is understood that the system provided by the embodiment of the present invention corresponds to the method provided by the embodiment of the present invention, and the explanation, the example and the beneficial effects of the related contents can refer to the corresponding parts in the method.
The embodiment of the application also provides an electronic device, which comprises a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory complete mutual communication through the communication bus,
a memory for storing a computer program;
a processor, configured to implement the method for short-term photovoltaic power prediction in the power market context when executing a program stored in a memory, the method comprising:
dividing the weather into four weather models of sunny days, cloudy days, rain shower and full rain based on the NWP value;
judging the weather type of the day to be detected, and selecting the latest photovoltaic power historical data and the historical NWP value of the similar day according to the weather type of the day to be detected;
inputting the classified data set into each base learner in a layer of prediction model of the Stacking integrated model;
obtaining the prediction result of each base learner, constructing a new training set and inputting the new training set into a two-layer prediction model;
and obtaining a prediction result of the two-layer element learner, namely the final photovoltaic power generation power prediction value.
The communication bus mentioned in the electronic device may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The communication bus may be divided into an address bus, a data bus, a control bus, etc.
The communication interface is used for communication between the electronic equipment and other equipment.
The Memory may include a Random Access Memory (RAM) or a Non-Volatile Memory (NVM), such as at least one disk Memory. Optionally, the memory may also be at least one memory device located remotely from the processor.
The Processor may be a general-purpose Processor, including a Central Processing Unit (CPU), a Network Processor (NP), and the like; the Integrated Circuit may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), or other Programmable logic devices, discrete Gate or transistor logic devices, or discrete hardware components.
In yet another embodiment provided herein, there is also provided a computer program product containing instructions that, when executed on a computer, cause the computer to perform the method for short-term photovoltaic power prediction in any of the above embodiments in the context of an electricity market.
In the above embodiments, the implementation may be wholly or partially realized by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded and executed on a computer, cause the processes or functions described in accordance with the embodiments of the application to occur, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable device. The computer instructions may be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, from one website site, computer, server, or data center to another website site, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device, such as a server, a data center, etc., that incorporates one or more of the available media. The usable medium may be a magnetic medium (e.g., floppy Disk, hard Disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., Solid State Disk (SSD)), among others.
It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
All the embodiments in the present specification are described in a related manner, and the same and similar parts among the embodiments may be referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment.
The above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A short-term photovoltaic power prediction method in the context of the electricity market is characterized by comprising the following steps,
dividing the weather into four weather models of sunny days, cloudy days, rain shower and full rain based on the NWP value;
judging the weather type of the day to be detected, and selecting the latest photovoltaic power historical data and the historical NWP value of the similar day according to the weather type of the day to be detected;
inputting the classified data set into each base learner in a layer of prediction model of the Stacking integrated model;
obtaining the prediction result of each base learner, constructing a new training set and inputting the new training set into a two-layer prediction model;
and obtaining a prediction result of the two-layer element learner, namely the final photovoltaic power generation power prediction value.
2. The method of short-term photovoltaic power prediction in the context of the electricity market of claim 1, characterized by: dividing the weather into four weather models of sunny days, cloudy days, rain shower and full rain based on the NWP value; the method specifically comprises the following steps:
selecting the cloud cover C and the rainfall p in NWP as weather parting factors, dividing the cloud cover into a sunny model and a cloudy model according to the average cloud cover in the daytime, and dividing the cloud cover into a rain gust model and a full rain model according to the rainfall duration in the daytime;
the sunny day model is
The model of cloudy day is
The model of rain shower is
The full rain model is
c1A typing threshold, t, for both sunny and cloudy models1A typing threshold value of a gust rain and full rain model; according to the short-term weather forecast national standard and the meteorological principles, c1＝0.7，t1＝4。
3. The method of short-term photovoltaic power prediction in the context of the electricity market of claim 1, characterized by: the one-layer prediction model of the Stacking integration model comprises a physical characteristic model of photovoltaic power generation, and the physical characteristic model is shown in a formula (1):
P＝ηSI[1-0.005(t+25)] (1)
in the formula, P is photovoltaic power generation power, eta is photovoltaic panel conversion efficiency, S is photovoltaic panel effective area, I is irradiance, and t is photovoltaic panel working temperature.
4. The method of short-term photovoltaic power prediction in the context of the electricity market of claim 3, characterized in that: the one-layer prediction model further comprises a TCN network prediction model, the TCN network prediction model is constructed by the following steps,
and (3) a cavity convolution step, wherein each TCN layer contains L convolution layers, and the expansion convolution calculation formula is as follows:
in the formula: void factor d is (1, …, 2)L) K is the convolution kernel size;
residual linking step:
relu denotes a linear rectification function used as an activation function for a neural network; DC Conv denotes a void convolution layer;
equations (3), (4) represent the activation function of TCN:
in the formula: w(1)、W(2)Is a weight matrix corresponding to the input, b is an offset vector, S(i,j)And (3) representing an activation function of the ith layer of the jth block, wherein a formula (3) represents a result of hole convolution at the time t, and a formula (4) represents a result after residual error linkage is added.
5. The method of short term photovoltaic power prediction in the context of the electricity market as claimed in claim 4,
obtaining the prediction result of each base learner, constructing a new training set, inputting the new training set into a two-layer prediction model, and adopting the following algorithm:
the influence of high-dimensional data on prediction is relieved by adopting a LightGBM model based on a histogram algorithm; the LightGBM model is optimized by adopting a Leaf-wise algorithm with depth limitation;
the objective function of the LightGBM model is as follows:
Obj(t)＝L(t)+Ω(t)+c (5)
in the formula: obj (t) is an optimization target, and omega (t) represents a regular function and reflects the complexity of the model; t represents a sampling time; c represents additional parameters, avoiding over-fitting and optimizing tree depth;
l (t) denotes the loss function by describing the actual values y of the N sampling pointsiAnd the predicted value
To reflect the degree of fit of the model; the definition is as follows:
transmitting residual error information of a previous learner by serially coupling the regression trees; final output
Generated by the accumulation of the remaining trees.
6. The method of short-term photovoltaic power prediction in the context of the electricity market of claim 5, wherein:
obtaining the prediction result of each base learner, constructing a new training set and inputting the new training set into a two-layer prediction model, and adopting the following algorithm:
and a gradient-based unilateral sampling algorithm GOSS is adopted, so that during sampling, the GOSS can completely reserve large gradient samples meeting the conditions as sampled data, and samples with smaller gradients are sampled randomly.
7. The method of short-term photovoltaic power prediction in the context of the electricity market of claim 6, characterized in that:
obtaining the prediction result of each base learner, constructing a new training set and inputting the new training set into a two-layer prediction model, and adopting the following algorithm,
by adopting the method EFB for binding the mutual exclusion characteristics, after GOSS sampling, the dimensionality of the characteristics is reduced by binding the mutual exclusion characteristics so as to prevent dimensionality disasters and improve the calculation efficiency.
8. The method of short-term photovoltaic power prediction in the context of the electricity market of claim 7, wherein:
a layer of prediction model of the Stacking integration model divides a data set into a plurality of subsets through cross validation, and fusion is carried out on an evaluation result, so that the variance of the prediction result of the model is reduced, the generalization capability of the model is improved, and the occurrence of an overfitting phenomenon is avoided;
the K-fold cross validation is to averagely divide the data set into K parts, wherein K-1 part is used as a training set, and the rest 1 part is used as a validation set; and (5) training by using the training sets under the K conditions to obtain the model hyper-parameters.
9. A computer-readable storage medium, storing a computer program which, when executed by a processor, causes the processor to carry out the steps of the method according to any one of claims 1 to 8.
10. A computer device comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to carry out the steps of the method according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111679639.3A CN114492945A (en) | 2021-12-31 | 2021-12-31 | Short-term photovoltaic power prediction method, medium and equipment in electric power market background |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111679639.3A CN114492945A (en) | 2021-12-31 | 2021-12-31 | Short-term photovoltaic power prediction method, medium and equipment in electric power market background |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114492945A true CN114492945A (en) | 2022-05-13 |
Family
ID=81510180
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111679639.3A Pending CN114492945A (en) | 2021-12-31 | 2021-12-31 | Short-term photovoltaic power prediction method, medium and equipment in electric power market background |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114492945A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114971058A (en) * | 2022-06-09 | 2022-08-30 | 哈尔滨工业大学 | Photovoltaic forecasting method based on depth attention network and clear sky radiation prior fusion |
-
2021
- 2021-12-31 CN CN202111679639.3A patent/CN114492945A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114971058A (en) * | 2022-06-09 | 2022-08-30 | 哈尔滨工业大学 | Photovoltaic forecasting method based on depth attention network and clear sky radiation prior fusion |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wu et al. | Evolving RBF neural networks for rainfall prediction using hybrid particle swarm optimization and genetic algorithm | |
| Song et al. | A novel dual path gated recurrent unit model for sea surface salinity prediction | |
| CN109214566A (en) | Short-term wind power prediction method based on shot and long term memory network | |
| CN114792156B (en) | Photovoltaic output power prediction method and system based on curve characteristic index clustering | |
| Liu et al. | Photovoltaic generation power prediction research based on high quality context ontology and gated recurrent neural network | |
| Zhang et al. | Interval prediction of ultra-short-term photovoltaic power based on a hybrid model | |
| CN115860177A (en) | Photovoltaic power generation power prediction method based on combined machine learning model and application thereof | |
| Ding et al. | Serial-parallel dynamic echo state network: A hybrid dynamic model based on a chaotic coyote optimization algorithm for wind speed prediction | |
| CN113361785A (en) | Power distribution network short-term load prediction method and device, terminal and storage medium | |
| Li et al. | A wind power prediction method based on DE-BP neural network | |
| CN115115125A (en) | Photovoltaic power interval probability prediction method based on deep learning fusion model | |
| CN116826737A (en) | Photovoltaic power prediction method, device, storage medium and equipment | |
| Wang et al. | Tracking and predicting technological knowledge interactions between artificial intelligence and wind power: Multimethod patent analysis | |
| Sang et al. | Ensembles of gradient boosting recurrent neural network for time series data prediction | |
| Cao et al. | Short-term forecasting and uncertainty analysis of photovoltaic power based on the FCM-WOA-BILSTM model | |
| Mukilan et al. | Prediction of rooftop photovoltaic solar potential using machine learning | |
| CN114492945A (en) | Short-term photovoltaic power prediction method, medium and equipment in electric power market background | |
| Mohd et al. | Comparative study of rainfall prediction modeling techniques (A case study on Srinagar, J&K, India) | |
| CN115271242A (en) | Training method, prediction method and device of photovoltaic power generation power prediction model | |
| Sun et al. | Ultra-short-term wind power interval prediction based on fluctuating process partitioning and quantile regression forest | |
| Zhang et al. | Ultra-short-term photovoltaic power prediction based on similar day clustering and temporal convolutional network with bidirectional long short-term memory model: A case study using DKASC data | |
| Tao et al. | Recurrent neural networks application to forecasting with two cases: load and pollution | |
| Tamal et al. | Forecasting of solar photovoltaic output energy using lstm machine learning algorithm | |
| Yuan et al. | Short-term wind speed forecasting using STLSSVM hybrid model | |
| Xu et al. | Short-term wind power prediction based on anomalous data cleaning and optimized lstm network |
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 |