CN116845875A - WOA-BP-based short-term photovoltaic output prediction method and device - Google Patents
WOA-BP-based short-term photovoltaic output prediction method and device Download PDFInfo
- Publication number
- CN116845875A CN116845875A CN202310824840.9A CN202310824840A CN116845875A CN 116845875 A CN116845875 A CN 116845875A CN 202310824840 A CN202310824840 A CN 202310824840A CN 116845875 A CN116845875 A CN 116845875A
- Authority
- CN
- China
- Prior art keywords
- data
- day
- output
- prediction
- neural network
- 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 40
- 238000012549 training Methods 0.000 claims description 53
- 238000013528 artificial neural network Methods 0.000 claims description 49
- 239000013598 vector Substances 0.000 claims description 32
- 238000004422 calculation algorithm Methods 0.000 claims description 29
- 241000283153 Cetacea Species 0.000 claims description 23
- 230000005855 radiation Effects 0.000 claims description 22
- 238000005457 optimization Methods 0.000 claims description 17
- 230000006870 function Effects 0.000 claims description 10
- 230000002354 daily effect Effects 0.000 claims description 7
- 238000007781 pre-processing Methods 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 6
- 230000004913 activation Effects 0.000 claims description 4
- 238000003062 neural network model Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 230000002194 synthesizing effect Effects 0.000 claims description 3
- 230000003044 adaptive effect Effects 0.000 claims description 2
- 230000003203 everyday effect Effects 0.000 claims description 2
- 238000012545 processing Methods 0.000 claims description 2
- 230000009467 reduction Effects 0.000 claims description 2
- 238000012360 testing method Methods 0.000 description 5
- 230000006978 adaptation Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000013178 mathematical model Methods 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 238000010248 power generation Methods 0.000 description 3
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 description 1
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- 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
- G06N3/048—Activation functions
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
- H02J2300/26—The renewable source being solar energy of photovoltaic origin involving maximum power point tracking control for photovoltaic sources
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Power Engineering (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Photovoltaic Devices (AREA)
Abstract
The invention discloses a short-term photovoltaic output prediction method and a device based on WOA-BP. The method provided by the invention can improve the accuracy of photovoltaic power prediction in various types of weather, and constructs the photovoltaic output prediction device according to the method, thereby providing a quick and effective new thought for photovoltaic output prediction and better integrating the photovoltaic prediction into practical application.
Description
Technical Field
The invention relates to the technical field of photovoltaic power generation prediction, in particular to a short-term photovoltaic output prediction method and device based on WOA-BP.
Background
Under the background of global energy transformation, the trend of pushing energy to high efficiency, cleanness and diversification is great. Photovoltaic power generation, which is a representative of renewable energy sources, is a necessary trend to receive widespread attention from society, and photovoltaic has many advantages, and the installed capacity has also come to the highest peak in recent years. However, photovoltaic power generation has great volatility and randomness, large-scale photovoltaic access to a power grid brings great challenges to safe and stable operation of a power distribution network, and the problem that photovoltaic is difficult to be absorbed in recent years is also highlighted. The photovoltaic prediction can effectively predict the fluctuation of the photovoltaic output, and has extremely important significance for promoting the efficient fusion of new energy and the traditional power distribution network.
The photovoltaic output has different power characteristics under different weather conditions, the deviation of the prediction result is larger under the condition of larger weather fluctuation by the traditional photovoltaic prediction method, and the prediction precision can be influenced by the random initial weight and the threshold value of the BP neural network. In addition, most of current photovoltaic prediction is theoretical research, and no convenient and feasible hardware device is available, so that the photovoltaic prediction is difficult to apply to actual engineering.
Disclosure of Invention
The invention provides a short-term photovoltaic output prediction method and device based on WOA-BP for improving the accuracy of photovoltaic power prediction results under different weather conditions and improving the practical application capability of a photovoltaic prediction method.
In order to achieve the above effects, the specific technical scheme of the invention is as follows:
the invention firstly provides a short-term photovoltaic output prediction method based on WOA-BP, which comprises the following steps:
step one, importing historical data and forecast day weather forecast data.
Selecting past history data, calculating the pearson correlation coefficient of the photovoltaic power and each meteorological factor data, and determining the input of the BP neural network.
And thirdly, calculating cosine similarity of the weather forecast data and each weather image data in the historical data set, and selecting the historical data with higher cosine similarity value as a training set.
And step four, determining the topological structure of the BP neural network, wherein the topological structure comprises the number of input nodes, the number of hidden layer nodes and the number of output nodes.
And fifthly, searching an optimal initial weight threshold of the BP neural network by using a whale optimization algorithm, assigning the optimal initial weight threshold to the BP neural network, and training to obtain a photovoltaic prediction model.
And step six, substituting weather forecast data of 8-18 points per half hour of the forecast day into the model, and outputting photovoltaic output at the corresponding moment of the forecast day.
The specific method of the first step is as follows: historical data includes data for photovoltaic output power, wind speed, temperature, humidity, total horizontal radiation, and scattered horizontal radiation at the same interval per day. Data of two to three months before the day of prediction are imported, the data are preprocessed, the data at the moment that the photovoltaic output is not generated or the output is very small are removed, and the photovoltaic output and meteorological data with the same time interval of 8 to 18 points per day can be selected.
The specific method of the second step is as follows: and randomly selecting 1 day of data from the historical data in each month, respectively calculating pearson correlation coefficients of the photovoltaic output of each day and various meteorological data, synthesizing calculation results of the monthly correlation coefficients, and selecting meteorological factors with correlation with the photovoltaic output as input of a neural network according to the concept of the pearson correlation coefficients. The invention adopts five factors of wind speed, temperature, humidity, horizontal total radiation and horizontal scattered radiation, and the input number of the neural network is 5.
The specific method of the third step is as follows: the method comprises the steps of selecting a training set by adopting a cosine similarity concept, performing dimension reduction processing on forecast data of various weather factors on a forecast day, sequentially arranging data of five weather factors into a column vector, and performing the same operation on weather data in historical data every day. And calculating weather cosine similarity values of the prediction days and each day in the historical data set at the same time interval, arranging in descending order, and selecting data of the first days of sorting to construct a similarity day set as a training set of the BP neural network.
The specific method of the fourth step is as follows: the input number of the neural network is 5, the output is photovoltaic predicted power, the output number is 1, the number of hidden layers can be tested according to an empirical formula, BP neural network models are respectively built by ten hidden node numbers, training input data are substituted into the models for calculation, the calculated hidden layer node number of the model with the minimum error between the calculated output and the training sample real output is taken as the final hidden node number of the model, and therefore the topological structure of the BP neural network is established.
The specific method of the fifth step is as follows: the training frequency of the BP neural network is 1000, the learning rate is 0.01, the activation functions of the hidden layer and the output layer are respectively 'tan sig' and 'purelin', the initial population number of the whale optimization algorithm is 30, the maximum iteration frequency is 80, the adaptive value function is the training error of the BP neural network, the minimum mean square error is taken as a target, and the whale optimization algorithm is utilized to find the optimal initial weight and threshold. And finally, assigning the optimal initial weight and the threshold value to the neural network for prediction model training.
Another object of the present invention is to provide a WOA-BP based short-term photovoltaic output predicting apparatus for performing the above-described predicting method, the photovoltaic predicting apparatus comprising: the system comprises an input/output module, a data preprocessing module, a prediction input factor selection module, a training set data selection module, a training module and a prediction module, wherein:
the input/output module is mainly responsible for receiving externally imported historical data and forecast day weather data, and outputting photovoltaic output power at each forecast moment after the forecast of the internal model is finished. The data preprocessing module is mainly responsible for eliminating data at the moment that photovoltaic power is not output or output is very small in historical data, screening out forecast daily meteorological data every half hour and storing the forecast daily meteorological data. The prediction input factor selection module is mainly responsible for realizing the second step, embedding a pearson correlation coefficient calculation program in the module, and selecting proper input weather factors. The training set data selection module is mainly responsible for realizing the third step, and a cosine similarity calculation program is embedded in the training set data selection module to determine a similar day training set. The training module is mainly responsible for realizing the fourth step and the fifth step, a BP neural network algorithm and a whale optimization algorithm are embedded in the training module, the topological structure of the BP neural network is firstly determined, and then a WOA-BP prediction model is trained by using a training set of similar days. The prediction module is mainly responsible for realizing the step six, inputting the meteorological factor data stored by the input and output module, and predicting by using a trained model.
The beneficial effects are that:
according to the cosine similarity and the WOA-BP algorithm, the invention provides a more accurate photovoltaic prediction method, and photovoltaic output prediction can be performed under the conditions of stable weather and fluctuation. And a photovoltaic output prediction device is constructed, a photovoltaic prediction method flow is integrated into a hardware device, historical data and predicted solar and weather forecast data are input into the device, the device automatically executes the prediction flow, and finally, the device outputs the photovoltaic output prediction data of 8-18 points per half hour on the prediction day. The method can integrate photovoltaic prediction into practical application more conveniently and rapidly, and better promote safe and stable operation of the photovoltaic-containing power distribution network.
Drawings
FIG. 1 is a flowchart of a short-term photovoltaic output prediction method and apparatus based on WOA-BP;
FIG. 2 is a schematic diagram of a predictive device;
FIG. 3 is a topological structure diagram of a BP neural network photovoltaic prediction model;
FIG. 4 is a flowchart of a whale optimization BP neural network algorithm;
FIG. 5 is an evolutionary curve of the WOA algorithm under weather stationary and fluctuating conditions, where (a) in FIG. 5 shows an evolutionary curve of the WOA algorithm under weather stationary conditions, and (b) in FIG. 5 shows an evolutionary curve of the WOA algorithm under weather fluctuating conditions;
fig. 6 is a graph showing a comparison of a WOA-BP algorithm photovoltaic output prediction curve and a true value under the condition of weather stability and fluctuation, and fig. 6 (a) shows a comparison of a WOA-BP algorithm photovoltaic output prediction curve and a true value under the condition of weather stability, and (b) shows a comparison of a WOA-BP algorithm photovoltaic output prediction curve and a true value under the condition of weather fluctuation.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.
As shown in fig. 1, provided herein is a WOA-BP based short-term photovoltaic output prediction method comprising the steps of:
step one, importing historical data and forecast day weather forecast data. Historical data includes data for photovoltaic output power, wind speed, temperature, humidity, total horizontal radiation, and scattered horizontal radiation at the same interval per day. Data of two to three months before the day of prediction are imported, the data are preprocessed, the data at the moment that the photovoltaic output is not generated or the output is very small are removed, and the photovoltaic output and meteorological data with the same time interval of 8 to 18 points per day can be selected.
Selecting past history data, calculating the pearson correlation coefficient of the photovoltaic power and each meteorological factor data, and determining the input of the BP neural network.
And thirdly, calculating cosine similarity of the weather forecast data and each weather image data in the historical data set, and selecting the historical data with higher cosine similarity value as a training set.
And step four, determining the topological structure of the BP neural network, wherein the topological structure comprises the number of input nodes, the number of hidden layer nodes and the number of output nodes.
And fifthly, searching an optimal initial weight threshold of the BP neural network by using a whale optimization algorithm, assigning the optimal initial weight threshold to the BP neural network, and training to obtain a photovoltaic prediction model.
And step six, substituting weather forecast data of 8-18 points per half hour of the forecast day into the model, and outputting photovoltaic output at the corresponding moment of the forecast day.
In the second step, the neural network input characteristic is established by adopting the concept of pearson correlation coefficient, wherein the pearson correlation coefficient is a function for measuring the correlation degree between two vectors, and the formula is as follows:
wherein R is (x,y) For the vector x= (x) 1 ,x 2 ,…,x n ) Sum vector y= (y) 1 ,y 2 ,…,y n ) Correlation coefficient, x i As the i-th element of the vector x,is the average value of each element of the vector x, y i Is the i-th element of vector y, +.>Is the mean value of each element of the vector y. And randomly selecting 1-day data from the historical data in each month, respectively calculating pearson correlation coefficients of each weather feature vector and the photovoltaic output vector, synthesizing pearson correlation coefficient values in each month, and finally determining the input number according to pearson correlation coefficient concepts.
In the third step, the neural network training set can be selected by adopting the concept of cosine similarity, and the specific steps are as follows: let P be 1 、P 2 、P 3 、P 4 、P 5 Respectively predicting column vectors of wind speed, temperature, humidity, horizontal total radiation and horizontal diffuse radiation at the same interval time of day, H i1 、H i2 、H i3 、H i4 、H i5 Respectively, column vectors of wind speed, temperature, humidity, horizontal total radiation and horizontal diffuse radiation at the same interval time of the i-th day history data. Converting weather data of predicted and historical days into row vectors, i.e.
Predicted solar-air image vector P and i-th day history solar-air image vector H i The cosine similarity formula is carried into to calculate, and the cosine similarity day formula is as follows:
wherein P is j Is the j-th element of vector P, H ij Is the vector H i Is the j-th element of (c). The larger the cosine similarity value is, the higher the similarity is, the weather cosine similarity values of the prediction day and each day in the historical data set at the same time interval are calculated, the weather cosine similarity values are arranged in a descending order, and the data of the first days of the ordering are selected to construct a similarity day set which is used as a training set of the BP neural network.
In the fourth step, the number of hidden layers of the neural network is established by adopting a method of an empirical formula, and the empirical formula is as follows:
wherein num is the number of hidden nodes, m is the number of input nodes, n is the number of output nodes, and a is an integer between 1 and 10. Therefore, the number of hidden nodes is ten, the hidden nodes can be established by adopting a trial-and-error method, before the model is established, the BP neural network model is respectively established by using the ten hidden nodes, training input data is substituted into the model for calculation, the hidden layer node number of the model with the minimum error between calculated output and training sample true output is taken as the final hidden node number of the model, and fig. 3 is a topological structure diagram of the BP neural network photovoltaic prediction model.
Searching an optimal initial weight threshold value of the BP neural network by using a whale optimization algorithm, wherein the training frequency of the BP neural network is 1000, the learning rate is 0.01, the activation functions of the hidden layer and the output layer are respectively 'tan sig' and 'purelin', the initial population number of the whale optimization algorithm is 30, the maximum iteration frequency is 80, the adaptation value function is the training error of the BP neural network, the selected similar calendar history data set is divided into a training set and a testing set, the adaptation value function is the mean square error of the testing set, and the specific formula is as follows:
wherein n is the number of data in the selected test set, T i For the true value of the test set,is a predictive value for the test set. And (5) taking the minimum adaptation value function as a target, and storing the initial weight and the threshold value when the error is minimum. The specific flow of the whale optimization algorithm is as follows:
the whale optimization algorithm can be divided into three stages of surrounding the prey, helically updating the position, and searching for the prey.
(1) Surrounding prey
The mathematical model of whale when surrounding a prey is expressed as follows:
D 1 ＝|CX b (t)-X(t)|
X(t+1)＝X b (t)-AD 1
wherein: d (D) 1 For updating step length, t is iteration number, A and C are vector coefficients, X b (t) represents the current best whale position vector, and X (t) represents the current whale position vector.
The calculation formula of the vector coefficients a and C is as follows:
wherein r is 1 、r 2 Is [0,1]]A is a linear convergence factor, T m Is the maximum number of iterations.
(2) Spiral update position
The mathematical model of whale at spiral update position is expressed as follows:
X(t+1)＝X b (t)+|X b (t)-X(t)|·e bl ·cos(2πl)
wherein b is usually a random number with 1, l being [ -1,1 ].
(3) Searching for prey
Whale during the hunting phase, the mathematical model is expressed as follows:
D 2 ＝|CX rand (t)-X(t)|
X(t+1)＝X rand (t)-AD 2
wherein: d (D) 2 To update step size, X rand (t) is a random whale position vector.
WOA selects predation behavior according to probability P, P is a random number of [0,1], and when P is less than 0.5 and |A| < 1, a surrounding prey mechanism is selected for position updating; when P is less than 0.5 and A is more than or equal to 1, selecting a hunting mechanism for position updating; and when P is more than or equal to 0.5, selecting a spiral updating mechanism to update the position. And after each position update, calculating the current adaptation value, and updating the optimal position until the termination condition is reached. Finally, the optimal initial weight and threshold value are assigned to the BP neural network and are subjected to predictive model training, and fig. 4 is a flowchart of the whale optimization BP neural network algorithm.
As shown in fig. 2, a WOA-BP based short-term photovoltaic output prediction apparatus for performing the above-described prediction method, the photovoltaic prediction apparatus comprising: the system comprises an input/output module, a data preprocessing module, a prediction input factor selection module, a training set data selection module, a training module and a prediction module, wherein:
the input/output module is mainly responsible for receiving externally imported historical data and forecast day weather data, and outputting photovoltaic output power at each forecast moment after the forecast of the internal model is finished.
The data preprocessing module is mainly responsible for eliminating data at the moment that photovoltaic power is not output or output is very small in historical data, screening out forecast daily meteorological data every half hour and storing the forecast daily meteorological data.
The prediction input factor selection module is mainly responsible for realizing the second step, a pearson correlation coefficient calculation program is embedded in the module, pearson correlation coefficient values of photovoltaic output and various meteorological factors in historical data are calculated, and decision is made according to the concept of the pearson correlation coefficients, so that proper input meteorological factors are selected;
the training set data selection module is mainly responsible for realizing the third step, a cosine similarity calculation program is embedded in the module, cosine similarity values of predicted days and historical solar-air image vectors are calculated, descending order is carried out, historical data of the first few days in the record are selected as a similar day training set, and the data are stored;
the training module is mainly responsible for realizing the fourth step and the fifth step, a BP neural network algorithm and a whale optimization algorithm are embedded in the training module, the topological structure of the BP neural network is firstly determined, and then a WOA-BP prediction model is trained by using a training set of similar days.
The prediction module is mainly responsible for realizing the step six, inputting the meteorological factor data stored in the input and output module, predicting by using a trained model, and transmitting the prediction result to the input and output module.
And (3) carrying out calculation analysis:
in order to further verify the validity of short-term photovoltaic prediction of the invention, the historical power and meteorological data of a photovoltaic power station in Australia (DKACC) are used for verification, the historical data comprise photovoltaic actual power and meteorological data with a time interval of 5 minutes per day, five meteorological features of wind speed, temperature, humidity, total horizontal radiation and scattered horizontal radiation are selected as neural network inputs according to the concept of a pearson correlation coefficient, the photovoltaic power is output, the number of hidden layers is determined by adopting a trial and error method according to the data error of a training set, the training frequency of a BP neural network is 1000, the learning rate is 0.01, the activation functions of a hidden layer and an output layer are respectively 'tan sig' and 'purelin', the initial population number of a whale optimization algorithm is 30, and the maximum iteration frequency is 80.
The photovoltaic output data of each half hour under the conditions of weather stability and weather fluctuation are respectively predicted, the evolution curve of the WOA algorithm under the conditions of weather stability and fluctuation is shown in FIG. 5, the comparison graph of the photovoltaic output prediction curve of the WOA-BP algorithm under the conditions of weather stability and fluctuation and the true value is shown in FIG. 6, and the error index analysis tables of MAE, MSE and MAPE under the two conditions are shown in Table 1.
TABLE 1 error index analysis Table under weather stability and fluctuation conditions
According to simulation results and error analysis, the WOA-BP short-term photovoltaic prediction method based on cosine similarity days can conduct accurate photovoltaic output prediction under the condition of stable weather and fluctuation, and the feasibility of the method is verified.
Claims (7)
1. A short-term photovoltaic output prediction method based on WOA-BP is characterized by comprising the following steps:
step one, importing historical data and forecast date weather forecast data;
selecting past history data, calculating the pearson correlation coefficient of photovoltaic power and each meteorological factor data, and determining the input of the BP neural network according to the concept of the pearson correlation coefficient;
calculating cosine similarity of weather forecast data and weather image data in a historical data set, and selecting historical data with higher cosine similarity value as a training set;
determining the topological structure of the BP neural network, wherein the topological structure comprises the number of input nodes, the number of hidden layer nodes and the number of output nodes;
searching an optimal initial weight threshold of the BP neural network by using a whale optimization algorithm, assigning the optimal initial weight threshold to the BP neural network, and training to obtain a photovoltaic prediction model;
and step six, substituting weather forecast data of 8-18 points per half hour of the forecast day into the model, and outputting photovoltaic output at the corresponding moment of the forecast day.
2. A short term photovoltaic output prediction method based on WOA-BP as set forth in claim 1, wherein: step one, the historical data comprise photovoltaic output power, wind speed, temperature, humidity, horizontal total radiation and horizontal scattered radiation at the same interval time every day; and importing and preprocessing historical data of two to three months before the day, removing data at the moment that the photovoltaic output is not generated or is very small, and selecting photovoltaic output and meteorological data at the same time interval of 8 to 18 points per day.
3. A short term photovoltaic output prediction method based on WOA-BP as set forth in claim 1, wherein: the specific steps of the second step are as follows: and randomly selecting 1 day of data from the historical data every month, respectively calculating pearson correlation coefficients of photovoltaic output of each day and various meteorological data, synthesizing pearson correlation coefficient values of each month, selecting meteorological factors with correlation with the photovoltaic output as input of a neural network according to the concept of the pearson correlation coefficients, and finally adopting five factors of wind speed, temperature, humidity, horizontal total radiation and horizontal scattered radiation, wherein the input number of the neural network is 5.
4. A short term photovoltaic output prediction method based on WOA-BP as set forth in claim 1, wherein: the specific method of the third step is as follows: the method comprises the steps of selecting a training set by adopting a cosine similarity concept, performing dimension reduction processing on forecast data of various weather factors in a forecast day, and sequentially arranging data of five weather factors including wind speed, temperature, humidity, horizontal total radiation and horizontal scattered radiation into a column vector; the same operation is carried out on the weather data of each day in the historical data, and weather cosine similarity values of the predicted day and each day in the historical data set at the same time interval are calculated, wherein the specific method comprises the following steps: let P be 1 、P 2 、P 3 、P 4 、P 5 Respectively predicting column vectors of wind speed, temperature, humidity, horizontal total radiation and horizontal diffuse radiation at the same interval time of day, H i1 、H i2 、H i3 、H i4 、H i5 Respectively, converting weather data of the forecast day and the history day into row vectors, namely, converting the weather data of the forecast day and the history day into column vectors of wind speed, temperature, humidity, horizontal total radiation and horizontal diffuse radiation under the same interval time of the history data of the i day
Predicted solar-air image vector P and i-th day history solar-air image vector H i The cosine similarity formula is carried into to calculate, and the cosine similarity day formula is as follows:
wherein P is j Is the j-th element of vector P, H ij Is the vector H i Calculating weather cosine similarity values of the predicted day and each day in the historical data set at the same time interval; and finally, descending order is carried out, and data of the first days of ordering are selected to construct a similar day set which is used as a training set of the BP neural network.
5. A short term photovoltaic output prediction method based on WOA-BP as set forth in claim 1, wherein: the specific method of the fourth step is as follows: the input number of the neural network is 5, the output is photovoltaic predicted power, the output number is 1, the number of hidden layers can be tested according to an empirical formula, BP neural network models are respectively built by ten hidden node numbers, training input data are substituted into the models for calculation, the calculated hidden layer node number of the model with the minimum error between the calculated output and the training sample real output is taken as the final hidden node number of the model, and therefore the topological structure of the BP neural network is established.
6. A short term photovoltaic output prediction method based on WOA-BP as set forth in claim 1, wherein: the specific method of the fifth step is as follows: the training frequency of the BP neural network is 1000, the learning rate is 0.01, the activation functions of the hidden layer and the output layer are respectively 'tan sig' and 'purelin', the initial population number of the whale optimization algorithm is 30, the maximum iteration frequency is 80, the adaptive value function is the training error of the BP neural network, the minimum mean square error is used as a target, the whale optimization algorithm is utilized to find the optimal initial weight and the threshold, and finally the optimal initial weight and the threshold are assigned to the neural network for prediction model training.
7. A WOA-BP based short-term photovoltaic output prediction apparatus for performing the WOA-BP based short-term photovoltaic output prediction method according to any one of claims 1 to 6, characterized in that: the prediction device comprises an input and output module, a data preprocessing module, a prediction input factor selection module, a training set data selection module, a training module and a prediction module;
the input/output module is used for receiving externally imported historical data and predicted daily weather data and outputting photovoltaic output power at each predicted moment after the prediction of the internal model is finished;
the data preprocessing module is used for eliminating data at the moment that photovoltaic power is not output or output is very small in historical data, screening out forecast daily meteorological data every half hour and storing the forecast daily meteorological data;
the prediction input factor selection module is used for embedding a pearson correlation coefficient calculation program into the prediction input factor selection module, calculating pearson correlation coefficient values of photovoltaic output and various meteorological factors in historical data, and making a decision according to the concept of the pearson correlation coefficient so as to select proper input meteorological factors;
the training set data selection module is used for embedding a cosine similarity calculation program into the training set data selection module, calculating cosine similarity values of predicted day and historical solar and air image vectors, sorting the cosine similarity values in descending order, selecting historical data of the first days in the records as a similar day training set, and storing the data;
the training module is used for embedding a BP neural network algorithm and a whale optimization algorithm into the module, firstly determining the topological structure of the BP neural network, and then training a WOA-BP prediction model by using a training set of similar days;
the prediction module is used for inputting the predicted solar and air condition factor data stored in the input and output module, predicting by using a trained model, and transmitting a prediction result to the input and output module.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310824840.9A CN116845875A (en) | 2023-07-06 | 2023-07-06 | WOA-BP-based short-term photovoltaic output prediction method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310824840.9A CN116845875A (en) | 2023-07-06 | 2023-07-06 | WOA-BP-based short-term photovoltaic output prediction method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116845875A true CN116845875A (en) | 2023-10-03 |
Family
ID=88166762
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310824840.9A Pending CN116845875A (en) | 2023-07-06 | 2023-07-06 | WOA-BP-based short-term photovoltaic output prediction method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116845875A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117039894A (en) * | 2023-10-09 | 2023-11-10 | 国家电投集团江西电力工程有限公司 | Photovoltaic power short-term prediction method and system based on improved dung beetle optimization algorithm |
-
2023
- 2023-07-06 CN CN202310824840.9A patent/CN116845875A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117039894A (en) * | 2023-10-09 | 2023-11-10 | 国家电投集团江西电力工程有限公司 | Photovoltaic power short-term prediction method and system based on improved dung beetle optimization algorithm |
| CN117039894B (en) * | 2023-10-09 | 2024-04-05 | 国家电投集团江西电力工程有限公司 | Photovoltaic power short-term prediction method and system based on improved dung beetle optimization algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110070226B (en) | Photovoltaic power prediction method and system based on convolutional neural network and meta-learning | |
| CN113326969B (en) | Short-term wind speed prediction method and system based on improved whale optimization algorithm and ELM | |
| CN112052913B (en) | Distributed photovoltaic power station power data virtual acquisition method | |
| CN111242355A (en) | Photovoltaic probability prediction method and system based on Bayesian neural network | |
| CN114792156A (en) | Photovoltaic output power prediction method and system based on curve characteristic index clustering | |
| Wang et al. | Application of DBN for estimating daily solar radiation on horizontal surfaces in Lhasa, China | |
| CN114676923A (en) | Method and device for predicting generated power, computer equipment and storage medium | |
| CN116845875A (en) | WOA-BP-based short-term photovoltaic output prediction method and device | |
| Aliberti et al. | Forecasting Short-term Solar Radiation for Photovoltaic Energy Predictions. | |
| CN115115125A (en) | Photovoltaic power interval probability prediction method based on deep learning fusion model | |
| CN116345555A (en) | CNN-ISCA-LSTM model-based short-term photovoltaic power generation power prediction method | |
| CN116629416A (en) | Photovoltaic power station power prediction method and device | |
| CN116014722A (en) | Sub-solar photovoltaic power generation prediction method and system based on seasonal decomposition and convolution network | |
| CN112836876B (en) | Power distribution network line load prediction method based on deep learning | |
| CN113111592A (en) | Short-term wind power prediction method based on EMD-LSTM | |
| CN117132132A (en) | Photovoltaic power generation power prediction method based on meteorological data | |
| CN112307672A (en) | BP neural network short-term wind power prediction method based on cuckoo algorithm optimization | |
| CN116187540B (en) | Wind power station ultra-short-term power prediction method based on space-time deviation correction | |
| CN116681154A (en) | Photovoltaic power calculation method based on EMD-AO-DELM | |
| CN113449466B (en) | Solar radiation prediction method and system for optimizing RELM based on PCA and chaos GWO | |
| CN115907131A (en) | Method and system for building electric heating load prediction model in northern area | |
| CN115456286A (en) | Short-term photovoltaic power prediction method | |
| CN114897229A (en) | Ultra-short-term prediction method for irradiance of photovoltaic power station based on AP clustering and Markov chain | |
| CN115758901A (en) | EHO-GRU network-based short-term distributed photovoltaic output power prediction method | |
| CN112581311B (en) | Method and system for predicting long-term output fluctuation characteristics of aggregated multiple wind power plants |
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 |