CN115983434A - Short-term wind field wind speed prediction method and system based on neural network - Google Patents
Short-term wind field wind speed prediction method and system based on neural network Download PDFInfo
- Publication number
- CN115983434A CN115983434A CN202211560098.7A CN202211560098A CN115983434A CN 115983434 A CN115983434 A CN 115983434A CN 202211560098 A CN202211560098 A CN 202211560098A CN 115983434 A CN115983434 A CN 115983434A
- Authority
- CN
- China
- Prior art keywords
- wind speed
- formula
- neural network
- dragonfly
- value
- 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
Images
Abstract
The embodiment of the invention discloses a short-term wind field wind speed prediction method and a short-term wind field wind speed prediction system based on a neural network, wherein the method comprises the following steps: acquiring a historical wind speed data set of a target area; preprocessing a historical wind speed data set to obtain a historical wind speed fluctuation amount and a historical wind speed trend amount; extracting historical wind speed fluctuation amount; predicting historical wind speed trend quantity by using a pre-constructed LSTM neural network prediction model to obtain predicted wind speed trend quantity; the pre-constructed LSTM neural network prediction model comprises the steps of optimizing the LSTM neural network by using a dragonfly algorithm to obtain an optimal parameter value; and performing data fusion on the predicted wind speed trend amount and the extracted historical wind speed fluctuation amount, and outputting a predicted result of the wind speed. The invention optimizes the LSTM neural network by using the dragonfly algorithm, thereby improving the prediction precision and the calculation efficiency of the neural network.
Description
Technical Field
The invention relates to the technical field of short-term wind field wind speed prediction methods, in particular to a short-term wind field wind speed prediction method and system based on a neural network.
Background
At present, with the advent of micro-grids, smart grids, distributed generation, and the ever-increasing world demand for energy, these enable clean resources such as wind and solar energy to remain stable on the grid. Sustainable development power resources are increased every year, wind energy has already occupied a great position in the economic development of China, and the wind energy is a renewable resource which is environment-friendly and rich in reserves. In the aspects of the power market, the power management, the optimization of the wind farm, the power evacuation and the like, due to the instability of wind power resources, adverse effects are generated on the operation of the power market, the energy management, the optimization of the wind farm, the evacuation of the power and the like, so that the method has a very important significance on the development of the power market. Therefore, accurate and effective wind speed forecasting has important theoretical and practical application values for development and utilization of large-scale wind resources, and the accurate wind speed forecasting is a key problem.
Therefore, many new forecasting methods are proposed for forecasting wind power generation. In a mode, the method has physics, statistics and artificial intelligence. Because physical simulation requires a large amount of environmental data such as weather, terrain and the like, a physical model is usually adopted for long-time wind forecasting; at present, the common statistical means include an ARIMA method, a Kalman filtering method and the like, and the ARIMA method and the Kalman filtering method forecast the historical wind power generation data based on the historical wind power generation data. Although this method is suitable for short-cycle prediction, it requires a smooth timing and cannot handle non-linear sequences; in recent years, due to the rapid development of computer technology, many neural network models based on neural networks are nonlinear models capable of wind speed prediction, and among these models, SVM (support vector machine), BP neural network, and various intelligent algorithms are most commonly and effectively combined. Some researchers study a mode of combining a support vector machine and an ANN (artificial neural network), and find that the prediction effect of SVM-ANN prediction is better than that of the support vector machine and the ANN prediction. The result shows that compared with the traditional autoregressive moving average value, the comprehensive forecasting result by adopting the artificial neural network is much better. Currently, much research is carried out on related technologies of ultra-short term prediction, and compared with short term prediction, the method has the defects of complex change mode, poor regularity and difficulty in prediction. However, in the existing wind field wind speed prediction, the existing wind field wind speed prediction and measurement rate is not ideal, so a set of comprehensive prediction modes must be searched to improve the prediction accuracy rate.
Disclosure of Invention
Aiming at the defects, the embodiment of the invention discloses a short-term wind field wind speed prediction method and system based on a neural network. For random fluctuation influencing wind speed prediction, wind speed trend prediction of wind speed fluctuation amount is established, and finally fusion of the wind speed trend prediction amount and the wind speed fluctuation amount is carried out, so that wind speed prediction is more accurate.
The embodiment of the invention discloses a short-term wind field wind speed prediction method based on a neural network in a first aspect, which comprises the following steps:
acquiring a historical wind speed data set of a target area;
preprocessing the historical wind speed data set, and obtaining a historical wind speed fluctuation quantity and a historical wind speed trend quantity by using a formula 1; wherein, formula 1 is V y ＝V q +V b In the formula, V y For historical wind speed, V q Is historical wind speed trend quantity; v b The historical wind speed fluctuation amount is obtained;
extracting the historical wind speed fluctuation amount;
predicting the historical wind speed trend quantity distributed according to the time sequence by utilizing a pre-constructed LSTM neural network prediction model to obtain predicted wind speed trend quantity; the pre-constructed LSTM neural network prediction model comprises the following steps: optimizing the LSTM neural network by using a dragonfly algorithm to obtain an optimal parameter value, and taking the optimal parameter value as an initial value of an LSTM neural network prediction model; wherein, the first and the second end of the pipe are connected with each other,the optimal parameter values include the number of network layers l h And the number of neurons g h ；
And performing data fusion on the predicted wind speed trend quantity and the extracted historical wind speed fluctuation quantity, and outputting a wind speed prediction result.
As an optional implementation manner, in the first aspect of the embodiment of the present invention, the preprocessing the historical wind speed training set includes:
decomposing the historical wind speed by using an ensemble empirical mode decomposition mode to obtain n historical wind speed components; wherein, the decomposition formula is as follows:
formula 2 is V y ＝V f1 +V f2 +V f3 +…+V fn In the formula, V y As a component of historical wind speed, V f1 Is a first historical wind speed component, V f2 Is the second historical wind speed component, V f3 Is the third calendar Shi Fengsu weight, V fn Is the nth historical wind speed component;
and reconstructing and calculating the n historical wind speed components by using sample entropy to obtain the historical wind speed trend quantity.
As an alternative implementation manner, in the first aspect of the embodiment of the present invention, performing reconstruction calculation on n historical wind speed components by using sample entropy includes:
calculating sample entropy values of the historical wind speed components by using a formula 3 according to n historical wind speed components obtained by decomposition, sequentially combining the historical wind speed components from large to small according to the sample entropy values, performing standard normalization normal analysis by adopting combined test of deviation and peak position, and combining the maximum historical wind speed components which accord with standard normal distribution to be used as the fluctuation quantity of the historical wind speed;
formula 3 is
In the formula: m is v Calculating dimensions for the reconstruction; r is v Is a threshold value;
calculating to obtain the historical wind speed trend quantity by using a formula 4 according to the historical wind speed fluctuation quantity; wherein the content of the first and second substances,
formula 4 is V q ＝V y -V b ，
In the formula, V y For historical wind speed, V q Is historical wind speed trend quantity; v b Is the historical wind speed fluctuation quantity.
As an alternative implementation, in the first aspect of the embodiment of the present invention, the optimization of the LSTM neural network using the dragonfly algorithm includes the following steps:
setting initialization parameters: the parameter of the LSTM neural network is set by the number of network layers l h And the number of neurons g h The dragonfly algorithm parameters are provided with maximum iteration times and population quantity;
setting a data set: setting corresponding data sets according to the same proportion, wherein the data sets mainly comprise a wind speed trend amount training set consisting of a plurality of historical wind speed trend amounts;
initializing the dragonfly algorithm: dragonfly position initialization X, position change step length initialization delta X, and network layer number l h And the number of neurons g h Set as the combination to be optimized for the LSTM neural network for each dragonfly, where the first row of the X matrix stores the number of parameter network layers l h Value of (a), the second row of the X matrix stores a parameter neuron number g h A value of (d);
updating each weight value: updating s, a, c, f, e, omega' and the network layer number l according to the initial value in the initialization parameter setting step h And the number of neurons g h (ii) a Wherein s is a separation weight, a is a queuing weight, c is a set weight, f is a hunting weight, e is an evasion natural enemy weight, and omega' is an inertia weight;
calculating a fitness value: training LSTM neural network according to wind speed trend amount, and calculating the root mean square error E of output value RMSE The reciprocal of (a) as an individual fitness value M; wherein the fitness value M of the individual: m =1/E RMSE ，E RMSE Calculated using equation 5:
In the formula, Y o Is an actual value, Y m Is a predicted value;
updating the positions of the food and the natural enemies;
updating the position of the dragonfly individual: updating the position and the position change step length of the dragonfly individual, calculating the fitness value M of the corresponding dragonfly individual, and comparing the fitness value M with the fitness value stored in the step of calculating the fitness value, so that the prediction effect is best; the specific process is as follows: every time the dragonfly carries out behavior operation, the maximum adaptability value of the current dragonfly is calculated: if the current dragonfly adaptability value is larger than the stored adaptability value, the current dragonfly adaptability value is used for replacing the originally stored optimal adaptability value, the better value is used as the optimal value of the current dragonfly, and the dragonfly parameter corresponding to the current optimal value is stored to be combined with the network layer number l h And the number of neurons g h Otherwise, the original fitness value and the corresponding dragonfly parameter combination network layer number l are still saved h And the number of neurons g h ；
Judging whether the termination condition of the algorithm is met: judging whether the preset maximum dragonfly iteration times are reached or not, and if so, outputting the optimal network layer number l h And the number of neurons g h (ii) a Otherwise, adding 1 to the iteration times, and skipping to execute the step of updating each weight value.
As an alternative implementation, in the first aspect of the embodiment of the present invention, the positions of the food and the natural enemies are updated according to formulas 6 to 10 of the dragonfly algorithm; the mathematical model of the dragonfly algorithm is as follows:
for dragonfly group with total number of groups N, the ith individual X in the group i As shown in formula 6;
In the formula (I), the compound is shown in the specification,
represents the position of the ith dragonfly individual in the d-dimension; when the individual X i When doing hunting behavior, it will do the following:
separating, wherein the mathematical expression is shown as formula 7;
formula 7 is
In the formula, S i Is the separation of the ith dragonfly individual, X is the current position of the individual, X j Is the position of the jth dragonfly adjacent to the individual, and N is the total number of individuals adjacent to dragonfly X;
queuing, wherein the mathematical expression of the queuing is shown as a formula 8;
In the formula, A i Is the position vector of the ith dragonfly individual when participating in the queuing behavior, V j Is the flight speed of the adjacent individual;
a set, the mathematical expression of which is shown in formula 9;
formula 9 is
In the formula, A i The position vector of the ith dragonfly individual in the collective action is obtained;
hunting food, the mathematical expression of which is shown as formula 10;
In the formula, F i Position vector of ith dragonfly individual during hunting behavior, X + Is the position of the prey.
As an alternative implementation manner, in the first aspect of the embodiment of the present invention, the position and the position change step size of the individual dragonfly are updated by using equation 11;
Wherein s is a separation weight, a is a queuing weight, c is an aggregation weight, f is a prey weight, e is an evasion natural enemy weight, and omega' is an inertia weight.
As an optional implementation manner, in the first aspect of the embodiment of the present invention, the hyper-parameters of the LSTM neural network prediction model are set to be adam solver, the initial learning rate is 0.001, the number of iterations is 3000, the learning rate is reduced by multiplying by a reduction factor of 0.01 after 1000 times of training, and the LSTM neural network prediction model is verified by using the root mean square error.
The second aspect of the embodiment of the invention discloses a short-term wind field wind speed prediction system based on a neural network, which comprises:
the acquisition module is used for acquiring a historical wind speed data set of a target area;
the preprocessing module is used for preprocessing the historical wind speed data set and obtaining a historical wind speed fluctuation quantity and a historical wind speed trend quantity by using a formula 1; wherein, formula 1 is V y ＝V q +V b In the formula, V y For historical wind speed, V q Is historical wind speed trend quantity; v b The historical wind speed fluctuation amount is obtained;
the extraction module is used for extracting the historical wind speed fluctuation amount;
the prediction module is used for predicting the historical wind speed trend quantity distributed according to the time series by utilizing a pre-constructed LSTM neural network prediction model to obtain predicted wind speed trend quantity; the pre-constructed LSTM neural network prediction model comprises the following steps: optimizing the LSTM neural network by using a dragonfly algorithm to obtain an optimal parameter value, and taking the optimal parameter value as an initial value of an LSTM neural network prediction model; wherein the optimal parameter value comprises the number of network layers l h And the number of neurons g h ；
And the fusion module is used for outputting a prediction result of the wind speed after data fusion is carried out on the predicted wind speed trend amount and the extracted historical wind speed fluctuation amount.
A third aspect of an embodiment of the present invention discloses an electronic device, including: a memory storing executable program code; a processor coupled with the memory; the processor calls the executable program codes stored in the memory for executing the short-term wind field wind speed prediction method based on the neural network disclosed by the first aspect of the embodiment of the invention.
A fourth aspect of the embodiments of the present invention discloses a computer-readable storage medium storing a computer program, where the computer program causes a computer to execute the short-term wind field wind speed prediction method based on a neural network disclosed in the first aspect of the embodiments of the present invention.
Compared with the prior art, the embodiment of the invention has the following beneficial effects:
according to the short-term wind field wind speed prediction method based on the neural network, the historical wind speed is decomposed into the wind speed fluctuation quantity and the wind speed trend quantity by preprocessing the historical wind speed, the wind speed fluctuation quantity influencing the wind speed prediction effect is extracted, and only the wind speed trend quantity is predicted, so that the prediction precision of the wind speed trend quantity is improved, and the problem that the error accumulation is easy to occur when a plurality of wind speed components are predicted at the same time, and the whole prediction error is overlarge is solved; meanwhile, the LSTM neural network is optimized by the dragonfly algorithm, so that the prediction precision and the calculation efficiency of the neural network are improved. For random fluctuation influencing wind speed prediction, wind speed trend prediction of wind speed fluctuation amount is established, and finally fusion of the wind speed trend prediction amount and the wind speed fluctuation amount is carried out, so that wind speed prediction is more accurate.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a schematic flow chart of a short-term wind field wind speed prediction method based on a neural network according to an embodiment of the present invention;
FIG. 2 is a system block diagram of a short-term wind field wind speed prediction method based on a neural network disclosed by an embodiment of the invention;
FIG. 3 is a schematic flow chart illustrating the pre-processing of the historical wind speed training set according to the embodiment of the present invention;
FIG. 4 is a schematic flow chart of an optimized LSTM neural network using the dragonfly algorithm according to the embodiment of the present invention;
FIG. 5 is a schematic diagram showing the comparison effect of the predicted result of LSTM, the predicted result of EEMD-LSTM, and the actual value of wind speed on the day with the predicted result of EEMD-DA-LSTM disclosed in the embodiment of the present invention;
FIG. 6 is a schematic structural diagram of a short-term wind field wind speed prediction system based on a neural network according to an embodiment of the present invention;
fig. 7 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
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 only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the terms "first", "second", "third", "fourth", and the like in the description and the claims of the present invention are used for distinguishing different objects, and are not used for describing a specific order. The terms "comprises," "comprising," and any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
For short-term wind speed prediction, the wind power generation has great uncertainty due to the inherent property (randomness) of the wind power generation; therefore, uncertainty is large in the process of predicting the short-term wind speed, and based on the uncertainty, the invention discloses a short-term wind field wind speed prediction method and system based on a neural network.
In the invention, EEMD is called EnsembleEmpiricalMododecomposition (ensemble empirical mode decomposition), and provides a noise-aided data analysis method aiming at the defects of an EMD method. The EEMD decomposition principle is that in order to solve the problems of modal aliasing and the like existing in EMD, noise is added into a signal for auxiliary analysis through a noise-assisted signal processing (NADA). In the EMD method, the ability to obtain a reasonable IMF depends on the distribution of the signal extremum points, and if the signal extremum points are not uniformly distributed, modal aliasing occurs. Therefore, white noise is added into a signal to be decomposed, and by utilizing the uniform distribution of a white noise frequency spectrum, when the signal is added on a white noise background which is uniformly distributed in the whole time-frequency space, signals with different time scales can be automatically distributed on a proper reference scale, and due to the characteristic of zero-mean noise, after multiple averaging, the noise is mutually counteracted, and the result of the integrated mean can be used as a final result. In order to suppress mixing between the IMF components, white gaussian noise with a mean value of zero is added to the EMD decomposition to perform an aided analysis, i.e., the EEMD algorithm.
In the invention, the sample entropy (samplencopy, sampEn) is similar to the physical meaning of the approximate entropy, and the complexity is measured by measuring the probability of generating a new mode in a signal, and the larger the probability of generating the new mode is, the larger the complexity of a sequence is. Sample entropy has two advantages over approximate entropy: the calculation of sample entropy is independent of data length; the sample entropy has better consistency, i.e. the sample entropy is affected to the same extent by the variation of the parameters m and r. The lower the value of sample entropy, the higher the sequence self-similarity; the larger the value of the sample entropy, the more complex the sample sequence.
In the invention, the basic principle DA of the dragonfly algorithm is that social behavior activities among dragonfly individuals are simulated, the social behavior activities are the same as those of most group intelligent algorithms, the behavior of the dragonfly individuals follows the principle of survival, the position of dragonfly food is mapped into a solution in the function optimization process, and the position is moved by searching a food source and avoiding natural enemies.
In the present invention, the LSTM is called a LongShortTermMemory, and as the name suggests, it has a neural network with the capability of memorizing long-term and short-term information.
Example one
Referring to fig. 1, fig. 1 is a schematic flow chart of a short-term wind field wind speed prediction method based on a neural network according to an embodiment of the present invention. The execution main body of the method described in the embodiment of the present invention is an execution main body composed of software or/and hardware, and the execution main body can receive related information in a wired or/and wireless manner and can send a certain instruction. Of course, it may also have certain processing and storage functions. The execution body may control a plurality of devices, such as a remote physical server or a cloud server and related software, or may be a local host or a server and related software for performing related operations on a device installed somewhere. In some scenarios, multiple storage devices may also be controlled, which may be co-located with the device or located in a different location. As shown in FIG. 1, the short-term wind field wind speed prediction method based on the neural network comprises the following steps:
s101: acquiring a historical wind speed data set of a target area;
specifically, the data acquisition and signal processing system 100 in fig. 2 is used to acquire original wind speed data, and the original wind speed data is analyzed and collated to obtain a historical wind speed data set; wherein the time length of the historical measured data of the wind speed is at least more than 5 years. Fig. 2 also includes a wind speed prediction and control system 200 and a prediction result display device 300.
S102: preprocessing the historical wind speed data set, and obtaining a historical wind speed fluctuation quantity and a historical wind speed trend quantity by using a formula 1; wherein, formula 1 is V y ＝V q +V b In the formula, V y For historical wind speed, V q Is historical wind speed trend quantity; v b The historical wind speed fluctuation amount is obtained;
s103: extracting the historical wind speed fluctuation amount;
s104: predicting the historical wind speed trend quantity distributed according to the time sequence by utilizing a pre-constructed LSTM neural network prediction model to obtain predicted wind speed trend quantity; wherein, it is constructed in advanceGood LSTM neural network prediction models include: optimizing the LSTM neural network by using a dragonfly algorithm to obtain an optimal parameter value, and taking the optimal parameter value as an initial value of an LSTM neural network prediction model; wherein the optimal parameter value comprises the network layer number l h And the number of neurons g h ；
S105: and performing data fusion on the predicted wind speed trend quantity and the extracted historical wind speed fluctuation quantity, and outputting a predicted result of the wind speed.
On the basis of the scheme, the short-term wind field wind speed forecasting method based on the neural network in the embodiment of the invention decomposes the historical wind speed into the wind speed fluctuation quantity and the wind speed trend quantity by preprocessing the historical wind speed, extracts the wind speed fluctuation quantity influencing the wind speed forecasting effect, and only forecasts the wind speed trend quantity, thereby improving the forecasting precision of the wind speed trend quantity, solving the problem that the error accumulation is easy to happen when a plurality of wind speed components are forecasted at the same time, and further causing the overlarge integral forecasting error; meanwhile, the invention optimizes the LSTM neural network by using the dragonfly algorithm, thereby improving the prediction precision and the calculation efficiency of the neural network. For random fluctuation influencing wind speed prediction, wind speed trend prediction of wind speed fluctuation amount is established, and finally fusion of the wind speed trend prediction amount and the wind speed fluctuation amount is carried out, so that wind speed prediction is more accurate.
Referring to fig. 3, fig. 3 is a schematic flow chart illustrating a process of preprocessing the historical wind speed training set according to an embodiment of the present invention. As an optional embodiment, preprocessing the historical wind speed training set includes:
s201: decomposing the historical wind speed by using a mode of ensemble empirical mode decomposition to obtain n historical wind speed components; wherein, the decomposition formula is as follows:
formula 2 is V y ＝V f1 +V f2 +V f3 +…+V fn In the formula, V y For the historical wind speed component, V f1 Is a first historical wind speed component, V f2 Is the second historical wind speed component, V f3 Is the third calendar Shi Fengsu component, V fn Is the nth historical wind speed component;
s202: and reconstructing and calculating the n historical wind speed components by using sample entropy to obtain the historical wind speed trend quantity.
As an alternative embodiment, the reconstruction calculation of the n historical wind speed components using the sample entropy comprises:
calculating sample entropy values of the historical wind speed components by using a formula 3 according to n historical wind speed components obtained by decomposition, sequentially combining the historical wind speed components from large to small according to the sample entropy values, performing standard normalization normal analysis by adopting combined test of deviation and peak position, and combining the maximum historical wind speed components which accord with standard normal distribution to be used as the fluctuation quantity of the historical wind speed;
formula 3 is
In the formula: m is v Calculating dimensions for the reconstruction; r is v Is a threshold value;
calculating to obtain the historical wind speed trend quantity by using a formula 4 according to the historical wind speed fluctuation quantity; wherein the content of the first and second substances,
formula 4 is V q ＝V y -V b ，
In the formula, V y For historical wind speed, V q Historical wind speed trend quantities; v b Is the historical wind speed fluctuation quantity.
The fluctuation quantity with strong randomness characteristics in the wind speed components and the wind speed numerical difference at two adjacent moments are taken as the wind speed fluctuation characteristics, and the wind speed change rule in a period of time meets the standard normal distribution. The wind speed components are decomposed by using Ensemble Empirical Mode Decomposition (EEMD), historical wind speeds are decomposed into wind speed fluctuation quantities and wind speed trend quantities, wind speed fluctuation quantities affecting the wind speed prediction effect are extracted, only the wind speed trend quantities are predicted, the prediction accuracy of the wind speed trend quantities is improved, and the problem that errors are easy to accumulate when multiple wind speed components are predicted at the same time, and further the overall prediction error is overlarge is solved.
The hyper-parameters of the LSTM neural network mainly comprise a resolver, a learning rate, iteration times, a Dropout factor, LSTM layer numbers and a neuralNumber of neurons, etc., while the number of LSTM layers and the number of neurons are the most influential factors. In order to improve the speed of network training and the accuracy of prediction, as an optional implementation manner, the hyper-parameters of the LSTM neural network prediction model are set to adam solver, the initial learning rate is 0.001, the iteration number is 3000, the learning rate is reduced by multiplying by a reduction factor of 0.01 after 1000 times of training, and the LSTM neural network prediction model is verified by using root mean square error. In addition, the dragonfly algorithm takes all possible factors of group behaviors (separation, alignment, cohesion, food attraction, natural enemy rejection and random walk of positions) into consideration, so that the dragonfly algorithm can quickly converge near an optimal value and has good global optimization capability and stability. Based on the advantages, the dragonfly algorithm is applied to the parameter optimization of the LSTM neural network, and the most suitable network layer number l is searched under the condition that the accuracy rate of the LSTM neural network is ensured to be maximum h And the number of neurons g h 。
Referring to fig. 4, fig. 4 is a schematic flow chart illustrating optimization of the LSTM neural network by the dragonfly algorithm according to the embodiment of the present invention; in order to determine the number of LSTM network layers and the number of neurons, a more optimal LSTM network structure is constructed, and the LSTM neural network is optimized by using a dragonfly algorithm, and the method comprises the following steps:
s301: initializing parameter setting: the parameters of the LSTM neural network are set with the network layer number l h And the number of neurons g h The parameters of the dragonfly algorithm are provided with maximum iteration times and population number;
s302: setting a data set: setting corresponding data sets according to the same proportion, wherein the data sets mainly comprise a wind speed trend training set consisting of a plurality of historical wind speed trend quantities;
s303: initializing the dragonfly algorithm: dragonfly position initialization X, position change step length initialization delta X, and network layer number l h And the number of neurons g h Set as the combination to be optimized for the LSTM neural network for each dragonfly, where the first row of the X matrix stores the number of parameter network layers l h Value of (a), the number of neurons of the second row of the X matrix storing the parameter g h A value of (d);
S304：updating each weight value: updating s, a, c, f, e, omega' and the network layer number l according to the initial value in the initialization parameter setting step h And the number of neurons g h (ii) a Wherein s is a separation weight, a is a queuing weight, c is an aggregation weight, f is a prey weight, e is an evasion natural enemy weight, and omega' is an inertia weight;
s305: calculating a fitness value: training LSTM neural network according to wind speed trend amount, and calculating the root mean square error E of output value RMSE As the fitness value M of the individual; wherein the fitness value M of the individual: m =1/E RMSE ，E RMSE Calculated using equation 5:
In the formula, Y o Is an actual value, Y m Is a predicted value;
s306: updating the positions of the food and the natural enemies;
s307: updating the position of the dragonfly individual: updating the position and the position change step length of the dragonfly individual by using the formula 11, calculating the fitness value M of the corresponding dragonfly individual, and comparing the fitness value M with the fitness value stored in the step of calculating the fitness value, so that the prediction effect is best; the specific process is as follows: every time the dragonfly carries out behavior operation, the maximum adaptability value of the current dragonfly is calculated: if the current dragonfly adaptability value is larger than the stored adaptability value, the current dragonfly adaptability value is used for replacing the originally stored optimal adaptability value, the better value is used as the optimal value of the current dragonfly, and the parameter combination network layer number l of the dragonfly corresponding to the current optimal value is stored h And the number of neurons g h Otherwise, the original fitness value and the corresponding dragonfly parameter combination network layer number l are still saved h And the number of neurons g h ；
S308: judging whether the termination condition of the algorithm is met: judging whether the preset maximum dragonfly iteration times are reached or not, and if so, outputting the optimal network layer number l h And the number of neurons g h (ii) a Otherwise, adding 1 to the iteration times, and skipping to execute the step of updating each weight value.
As an alternative embodiment, the positions of the food and the natural enemies are updated according to formula 6-formula 10 of the dragonfly algorithm; the mathematical model of the dragonfly algorithm is as follows:
for dragonfly group with total number of groups N, the ith individual X in the group i As shown in formula 6;
In the formula, X i d Represents the position of the ith dragonfly individual in the d-th dimension; when the subject X i When doing hunting behavior, it will do the following:
separating, wherein the mathematical expression is shown as formula 7;
formula 7 is
In the formula, S i Is the separation of the ith dragonfly individual, X is the current position of the individual, X is j Is the position of the jth dragonfly adjacent to the individual, and N is the total number of individuals adjacent to dragonfly X;
queuing, wherein the mathematical expression of the queuing is shown as a formula 8;
In the formula, A i Is the position vector of the ith dragonfly individual in the queuing behavior, V j Is the flight speed of the adjacent individual;
a set, the mathematical expression of which is shown in formula 9;
formula 9 is
In the formula, A i The position vector of the ith dragonfly individual during the collective action is obtained;
hunting food, the mathematical expression of which is shown as formula 10;
In the formula, F i Position vector of ith dragonfly individual during hunting behavior, X + Is the location of the prey.
As an alternative implementation manner, in the first aspect of the embodiment of the present invention, the position and the position change step size of the individual dragonfly are updated by using equation 11;
Wherein s is a separation weight, a is a queuing weight, c is an aggregation weight, f is a prey weight, e is an evasion natural enemy weight, and omega' is an inertia weight.
In the embodiment of the invention, the prediction model comprises a 24-dimensional input layer (inputting 24h wind speed trend amount data in a certain day), an LSTM layer, a Dropout layer, a full connection layer and a 24-dimensional output layer (outputting a 24h wind speed trend amount predicted value in the following day). Assuming that the existing trained LSTM neural network prediction model has the same dimension as the input and output data of the model of the method, namely data with the sampling interval of 1h are selected, the trend data of the wind speed is obtained through the first step of data decomposition and reconstruction calculation and is used as the input of the model, the trend output result is obtained through prediction, the final prediction result of the wind speed can be obtained by adding the extracted wind speed fluctuation quantity to the output result of the wind speed trend according to the formula 4, and the final prediction result of the wind speed is displayed on a human-computer interaction interface.
For example, to predict day 100: using the data of the first 99 days as training, inputting the data of 1-100 days for training, and outputting the data of 2-99 days; the test input is true data at day 99 and the output is predicted value at day 100. The predicted wind speed value on day 100 is obtained by the experiment, and the predicted results of LSTM (corresponding to the reference numeral 12 in the figure), EEMD-LSTM (corresponding to the reference numeral 13 in the figure) and EEMD-DA-LSTM (corresponding to the reference numeral 11 in the figure) are compared with the actual wind speed value on the day (corresponding to the reference numeral 14 in the figure), and the results of the comparison are shown in FIG. 5.
For the evaluation of the experimental prediction results, the Mean Absolute Error (MAE) and the Root Mean Square Error (RMSE) of the LSTM neural network prediction model, EEMD-LSTM, and EEMD-DA-LSTM neural network prediction models were calculated, respectively, and the calculation results are shown in Table 1.
TABLE 1
As can be seen from Table 1, the MAE and RMSE of EEMD-DA-LSTM are smaller than those of other two prediction models, the prediction accuracy is higher, and the method has more advantages in the wind speed short-term prediction field.
Example two
Referring to fig. 6, fig. 6 is a schematic structural diagram of a short-term wind field wind speed prediction system based on a neural network according to an embodiment of the present invention. As shown in fig. 6, a short-term wind field wind speed prediction system based on neural network includes:
an obtaining module 41, configured to obtain a historical wind speed data set of a target area;
the preprocessing module 42 is configured to preprocess the historical wind speed data set, and obtain a historical wind speed fluctuation amount and a historical wind speed trend amount by using formula 1; wherein, formula 1 is V y ＝V q +V b In the formula, V y For historical wind speed, V q Is historical wind speed trend quantity; v b The historical wind speed fluctuation amount is obtained;
an extracting module 43, configured to extract the historical wind speed fluctuation amount;
the prediction module 44 is configured to predict the historical wind speed trend quantity distributed according to the time series by using a pre-constructed LSTM neural network prediction model, so as to obtain a predicted wind speed trend quantity; the pre-constructed LSTM neural network prediction model comprises the following steps: optimizing the LSTM neural network by using a dragonfly algorithm to obtain an optimal parameter value, and taking the optimal parameter value as an initial value of an LSTM neural network prediction model; wherein the optimal parameter value comprises the number of network layers l h And the number of neurons g h ；
And the fusion module 45 is used for performing data fusion on the predicted wind speed trend amount and the extracted historical wind speed fluctuation amount and outputting a predicted result of the wind speed.
According to the short-term wind field wind speed prediction system based on the neural network, the historical wind speed is decomposed into the wind speed fluctuation quantity and the wind speed trend quantity by preprocessing the historical wind speed, the wind speed fluctuation quantity influencing the wind speed prediction effect is extracted, and only the wind speed trend quantity is predicted, so that the prediction precision of the wind speed trend quantity is improved, and the problem that the error accumulation is easy to occur when a plurality of wind speed components are predicted at the same time, and the whole prediction error is overlarge is solved; meanwhile, the invention optimizes the LSTM neural network by using the dragonfly algorithm, thereby improving the prediction precision and the calculation efficiency of the neural network. For random fluctuation influencing wind speed prediction, wind speed trend prediction of wind speed fluctuation amount is established, and finally fusion of the wind speed trend prediction amount and the wind speed fluctuation amount is carried out, so that wind speed prediction is more accurate.
EXAMPLE III
Referring to fig. 7, fig. 7 is a schematic structural diagram of an electronic device according to an embodiment of the disclosure. The electronic device may be a computer, a server, or the like, and may also be an intelligent device such as a mobile phone, a tablet computer, a monitoring terminal, or the like, and an image acquisition device having a processing function. The electronic device may include:
a memory 510 storing executable program code;
a processor 520 coupled to the memory 510;
the processor 520 calls the executable program code stored in the memory 510 to perform some or all of the steps of the short-term wind field wind speed prediction method based on neural network in the first embodiment.
The embodiment of the invention discloses a computer-readable storage medium which stores a computer program, wherein the computer program enables a computer to execute part or all of the steps in the short-term wind field wind speed prediction method based on the neural network in the first embodiment.
The embodiment of the invention also discloses a computer program product, wherein when the computer program product runs on a computer, the computer is enabled to execute part or all of the steps in the short-term wind field wind speed prediction method based on the neural network in the first embodiment.
The embodiment of the invention also discloses an application publishing platform, wherein the application publishing platform is used for publishing the computer program product, and when the computer program product runs on a computer, the computer is enabled to execute part or all of the steps in the short-term wind field wind speed prediction method based on the neural network in the first embodiment.
In various embodiments of the present invention, it should be understood that the sequence numbers of the processes do not mean the execution sequence necessarily in order, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be implemented in the form of hardware, and can also be implemented in the form of a software functional unit.
The integrated units, if implemented as software functional units and sold or used as a stand-alone product, may be stored in a computer accessible memory. Based on such understanding, the technical solution of the present invention, which is a part of or contributes to the prior art in essence, or all or part of the technical solution, can be embodied in the form of a software product, which is stored in a memory and includes several requests for causing a computer device (which may be a personal computer, a server, a network device, or the like, and may specifically be a processor in the computer device) to execute part or all of the steps of the method according to the embodiments of the present invention.
In the embodiments provided herein, it should be understood that "B corresponding to a" means that B is associated with a from which B can be determined. It should also be understood, however, that determining B from a does not mean determining B from a alone, but may also be determined from a and/or other information.
Those of ordinary skill in the art will appreciate that some or all of the steps of the various methods of the embodiments may be implemented by associated hardware instructed by a program, which may be stored in a computer-readable storage medium, including Read-only memory (ROM), random Access Memory (RAM), programmable Read-only memory (PROM), erasable programmable Read-only memory (EPROM), one-time programmable Read-only memory (OTPROM), electrically erasable rewritable Read-only memory (EEPROM), compact disc Read-only memory (CD-ROM) or other optical disc storage, magnetic disc storage, tape storage, or any other computer-readable medium capable of carrying or storing data.
The short-term wind field wind speed prediction method, the short-term wind field wind speed prediction device, the short-term wind field wind speed prediction electronic equipment and the storage medium which are disclosed by the embodiment of the invention are introduced in detail, a specific example is applied in the text to explain the principle and the implementation mode of the invention, and the description of the embodiment is only used for helping to understand the method and the core idea of the invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed, and in summary, the content of the present specification should not be construed as a limitation to the present invention.
Claims (10)
1. A short-term wind field wind speed prediction method based on a neural network is characterized by comprising the following steps:
acquiring a historical wind speed data set of a target area;
preprocessing the historical wind speed data set, and obtaining a historical wind speed fluctuation quantity and a historical wind speed trend quantity by using a formula 1; wherein, formula 1 is V y ＝V q +V b In the formula, V y For historical wind speed, V q Is historical wind speed trend quantity; v b Historical wind speed fluctuation amount;
extracting the historical wind speed fluctuation amount;
predicting the historical wind speed trend quantity distributed according to the time sequence by utilizing a pre-constructed LSTM neural network prediction model to obtain predicted wind speed trend quantity; the pre-constructed LSTM neural network prediction model comprises the following steps: optimizing the LSTM neural network by using a dragonfly algorithm to obtain an optimal parameter value, and taking the optimal parameter value as an initial value of an LSTM neural network prediction model; wherein the optimal parameter value comprises the number of network layers l h And the number of neurons g h ；
And performing data fusion on the predicted wind speed trend quantity and the extracted historical wind speed fluctuation quantity, and outputting a predicted result of the wind speed.
2. The neural network-based short-term wind farm wind speed prediction method of claim 1, wherein preprocessing the historical wind speed training set comprises:
decomposing the historical wind speed by using a mode of ensemble empirical mode decomposition to obtain n historical wind speed components; wherein, the decomposition formula is as follows:
formula 2 is V y ＝V f1 +V f2 +V f3 +…+V fn In the formula, V y As a component of historical wind speed, V f1 Is a first historical wind speed component, V f2 Is the second historical wind speed component, V f3 Is the third calendar Shi Fengsu component, V fn Is the nth historical wind speed component;
and reconstructing and calculating the n historical wind speed components by using sample entropy to obtain the historical wind speed trend quantity.
3. The neural network-based short-term wind farm wind speed prediction method of claim 2, wherein the reconstructing calculations of the n historical wind speed components using sample entropy comprises:
calculating a sample entropy value of each historical wind speed component by using a formula 3 according to n historical wind speed components obtained by decomposition, sequentially combining the historical wind speed components from large to small according to the sample entropy values, performing standard normalization normal analysis by adopting deviation and peak position joint test, and combining the maximum historical wind speed components which accord with the standard normal distribution as the historical wind speed fluctuation quantity;
formula 3 is
In the formula: m is v Calculating dimensions for the reconstruction; r is a radical of hydrogen v Is a threshold value;
calculating to obtain the historical wind speed trend quantity by using a formula 4 according to the historical wind speed fluctuation quantity; wherein, formula 4 is V q ＝V y -V b ，
In the formula, V y For historical wind speed, V q Is historical wind speed trend quantity; v b Is the historical wind speed fluctuation quantity.
4. The method for short-term wind field wind speed prediction based on neural network as claimed in any one of claims 1-3, wherein LSTM neural network is optimized by using dragonfly algorithm, comprising the following steps:
initializing parameter setting: the parameter of the LSTM neural network is set by the number of network layers l h And the number of neurons g h The dragonfly algorithm parameters are provided with maximum iteration times and population quantity;
setting a data set: setting corresponding data sets according to the same proportion, wherein the data sets mainly comprise a wind speed trend amount training set consisting of a plurality of historical wind speed trend amounts;
initializing the dragonfly algorithm: dragonfly position initialization X, position change step length initialization delta X, and network layer number l h And nerveNumber of elements g h Set as the combination to be optimized for the LSTM neural network for each dragonfly, where the first row of the X matrix stores the parameter network layer number l h Value of (a), the second row of the X matrix stores a parameter neuron number g h A value of (d);
updating each weight value: updating s, a, c, f, e, omega' and the network layer number l according to the initial value in the initialization parameter setting step h And the number of neurons g h (ii) a Wherein s is a separation weight, a is a queuing weight, c is an aggregation weight, f is a prey weight, e is an evasion natural enemy weight, and omega' is an inertia weight;
calculating a fitness value: training the LSTM neural network according to the wind speed trend quantity, and calculating the root mean square error E of the output value RMSE As the fitness value M of the individual; wherein the fitness value M of the individual: m =1/E RMSE ，E RMSE Calculated using equation 5:
formula 5 is
In the formula, Y o Is an actual value, Y m Is a predicted value;
updating the positions of the food and the natural enemies;
updating the position of the dragonfly individual: updating the position and the position change step length of the dragonfly individual, calculating the fitness value M of the corresponding dragonfly individual, and comparing the fitness value M with the fitness value saved in the step of calculating the fitness value, wherein the concrete process comprises the following steps: every time the dragonfly carries out behavior operation, the maximum adaptability value of the current dragonfly is calculated: if the current dragonfly adaptability value is larger than the stored adaptability value, the current dragonfly adaptability value is used for replacing the originally stored optimal adaptability value, the better value is used as the optimal value of the current dragonfly, and the parameter combination network layer number l of the dragonfly corresponding to the current optimal value is stored h And the number of neurons g h Otherwise, the original fitness value and the corresponding dragonfly parameter combination network layer number l are still saved h And the number of neurons g h ；
Judging whether the termination condition of the algorithm is met: judgmentIf the interruption reaches the preset dragonfly maximum iteration times, outputting the optimal network layer number l if the interruption reaches the preset dragonfly maximum iteration times h And the number of neurons g h (ii) a Otherwise, adding 1 to the iteration times, and skipping to execute the step of updating each weight value.
5. The neural network-based short-term wind field wind speed prediction method according to claim 4, wherein the positions of food and natural enemies are updated according to formula 6-formula 10 of the dragonfly algorithm; the mathematical model of the dragonfly algorithm is as follows:
for dragonfly group with total number of groups N, the ith individual X in the group i As shown in formula 6;
formula 6 is
In the formula, X i d Represents the position of the ith dragonfly individual in the d-dimension; when the subject X i When doing hunting behavior, it will do the following:
separating, wherein the mathematical expression is shown as formula 7;
formula 7 is
In the formula, S i Is the separation of the ith dragonfly individual, X is the current position of the individual, X is j Is the position of the jth dragonfly adjacent to the individual, and N is the total number of individuals adjacent to dragonfly X;
queuing, wherein the mathematical expression of the queuing is shown as a formula 8;
formula 8 is
In the formula, A i Is the position vector of the ith dragonfly individual when participating in the queuing behavior, V j Is the flight speed of the adjacent individual;
a set, the mathematical expression of which is shown as formula 9;
formula 9Is composed of
/>
In the formula, A i The position vector of the ith dragonfly individual in the collective action is obtained;
hunting food, the mathematical expression of which is shown as formula 10;
formula 10 is F i ＝X + -X
In the formula, F i Position vector, X, for the ith dragonfly individual to do hunting + Is the location of the prey.
6. The neural network-based short-term wind field wind speed prediction method according to claim 4, wherein the position and the position change step size of the dragonfly individual are updated by formula 11;
formula 11 is Δ X t+1 ＝(sS i +aA i +cC i +fF i +eE i )+ω′ΔX t
Wherein s is a separation weight, a is a queuing weight, c is a set weight, f is a hunting weight, e is an escape natural enemy weight, and omega' is an inertia weight.
7. The neural network-based short-term wind field wind speed prediction method of claim 4, wherein the hyper-parameters of the LSTM neural network prediction model are set to adam solver, initial learning rate of 0.001, iteration number of 3000, learning rate is reduced by multiplying by a reduction factor of 0.01 after 1000 times of training, and the LSTM neural network prediction model is verified with root mean square error.
8. A short-term wind field wind speed prediction system based on a neural network is characterized by comprising:
the acquisition module is used for acquiring a historical wind speed data set of a target area;
the preprocessing module is used for preprocessing the historical wind speed data set and obtaining a historical wind speed fluctuation quantity and a historical wind speed trend quantity by using a formula 1; whereinIn the formula 1 is V y ＝V q +V b In the formula, V y For historical wind speed, V q Is historical wind speed trend quantity; v b Historical wind speed fluctuation amount;
the extraction module is used for extracting the historical wind speed fluctuation amount;
the prediction module is used for predicting the historical wind speed trend quantity distributed according to the time series by utilizing a pre-constructed LSTM neural network prediction model to obtain predicted wind speed trend quantity; the pre-constructed LSTM neural network prediction model comprises the following steps: optimizing the LSTM neural network by using a dragonfly algorithm to obtain an optimal parameter value, and taking the optimal parameter value as an initial value of an LSTM neural network prediction model; wherein the optimal parameter value comprises the number of network layers l h And the number of neurons g h ；
And the fusion module is used for outputting a prediction result of the wind speed after data fusion is carried out on the predicted wind speed trend amount and the extracted historical wind speed fluctuation amount.
9. An electronic device, comprising: a memory storing executable program code; a processor coupled with the memory; the processor invokes the executable program code stored in the memory for performing the neural network based short term wind farm wind speed prediction method of any one of claims 1 to 7.
10. A computer-readable storage medium storing a computer program, wherein the computer program causes a computer to perform the neural network-based short-term wind farm wind speed prediction method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211560098.7A CN115983434A (en) | 2022-12-07 | 2022-12-07 | Short-term wind field wind speed prediction method and system based on neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211560098.7A CN115983434A (en) | 2022-12-07 | 2022-12-07 | Short-term wind field wind speed prediction method and system based on neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115983434A true CN115983434A (en) | 2023-04-18 |
Family
ID=85960318
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211560098.7A Pending CN115983434A (en) | 2022-12-07 | 2022-12-07 | Short-term wind field wind speed prediction method and system based on neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115983434A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107703751A (en) * | 2017-10-13 | 2018-02-16 | 河南工程学院 | PID controller optimization method based on dragonfly algorithm |
| CN108519347A (en) * | 2018-04-10 | 2018-09-11 | 中北大学 | A kind of infrared spectrum Wavelength selecting method based on binary system dragonfly algorithm |
| CN108924196A (en) * | 2018-06-20 | 2018-11-30 | 南京邮电大学 | Industry internet green energy resource management system |
| CN109002948A (en) * | 2018-10-29 | 2018-12-14 | 中冶赛迪电气技术有限公司 | The short-term photovoltaic power generation power prediction method of micro-capacitance sensor based on CDA-BP |
| CN112329339A (en) * | 2020-10-27 | 2021-02-05 | 河北工业大学 | Short-term wind speed prediction method for wind power plant |
| CN113051769A (en) * | 2021-04-09 | 2021-06-29 | 中南大学 | Power curve modeling method based on asymmetric loss and hybrid intelligent optimization algorithm |
| CN114595873A (en) * | 2022-02-23 | 2022-06-07 | 安徽理工大学 | Grey-correlation-based DA-LSTM short-term power load prediction method |
-
2022
- 2022-12-07 CN CN202211560098.7A patent/CN115983434A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107703751A (en) * | 2017-10-13 | 2018-02-16 | 河南工程学院 | PID controller optimization method based on dragonfly algorithm |
| CN108519347A (en) * | 2018-04-10 | 2018-09-11 | 中北大学 | A kind of infrared spectrum Wavelength selecting method based on binary system dragonfly algorithm |
| CN108924196A (en) * | 2018-06-20 | 2018-11-30 | 南京邮电大学 | Industry internet green energy resource management system |
| CN109002948A (en) * | 2018-10-29 | 2018-12-14 | 中冶赛迪电气技术有限公司 | The short-term photovoltaic power generation power prediction method of micro-capacitance sensor based on CDA-BP |
| CN112329339A (en) * | 2020-10-27 | 2021-02-05 | 河北工业大学 | Short-term wind speed prediction method for wind power plant |
| CN113051769A (en) * | 2021-04-09 | 2021-06-29 | 中南大学 | Power curve modeling method based on asymmetric loss and hybrid intelligent optimization algorithm |
| CN114595873A (en) * | 2022-02-23 | 2022-06-07 | 安徽理工大学 | Grey-correlation-based DA-LSTM short-term power load prediction method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20140195159A1 (en) | Application of artificial intelligence techniques and statistical ensembling to forecast power output of a wind energy facility | |
| CN112686464A (en) | Short-term wind power prediction method and device | |
| JP2013074695A (en) | Device, method and program for predicting photovoltaic generation | |
| CN110751318A (en) | IPSO-LSTM-based ultra-short-term power load prediction method | |
| JP2009225550A (en) | Power demand predicting method and system, and power generation prediction method | |
| CN109472283B (en) | Dangerous weather prediction method and device based on multiple incremental regression tree model | |
| CN114462718A (en) | CNN-GRU wind power prediction method based on time sliding window | |
| WO2022039675A1 (en) | Method and apparatus for forecasting weather, electronic device and storage medium thereof | |
| CN116595394A (en) | Training method of wind speed correction model, wind speed prediction method, wind speed prediction equipment and medium | |
| CN115221396A (en) | Information recommendation method and device based on artificial intelligence and electronic equipment | |
| CN114493052A (en) | Multi-model fusion self-adaptive new energy power prediction method and system | |
| CN115809795B (en) | Method and device for evaluating bearing capacity of production team based on digitalization | |
| CN116933037A (en) | Photovoltaic output prediction method based on multi-model fusion and related device | |
| CN112307672A (en) | BP neural network short-term wind power prediction method based on cuckoo algorithm optimization | |
| CN113779892B (en) | Method for predicting wind speed and wind direction | |
| CN116760006A (en) | Wind power prediction method, device, equipment and storage medium | |
| CN115983434A (en) | Short-term wind field wind speed prediction method and system based on neural network | |
| CN109840308B (en) | Regional wind power probability forecasting method and system | |
| CN115859765A (en) | Method, device, equipment and storage medium for predicting city expansion | |
| CN112836885B (en) | Combined load prediction method, combined load prediction device, electronic equipment and storage medium | |
| CN115907131A (en) | Method and system for building electric heating load prediction model in northern area | |
| CN112950033B (en) | Reservoir dispatching decision method and system based on reservoir dispatching rule synthesis | |
| CN114897204A (en) | Method and device for predicting short-term wind speed of offshore wind farm | |
| Bantupalli et al. | Wind Speed forecasting using empirical mode decomposition with ANN and ARIMA models | |
| CN115034159A (en) | Power prediction method, device, storage medium and system for offshore wind farm |
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 |