CN110322050B - Wind energy resource data compensation method - Google Patents
Wind energy resource data compensation method Download PDFInfo
- Publication number
- CN110322050B CN110322050B CN201910481772.4A CN201910481772A CN110322050B CN 110322050 B CN110322050 B CN 110322050B CN 201910481772 A CN201910481772 A CN 201910481772A CN 110322050 B CN110322050 B CN 110322050B
- Authority
- CN
- China
- Prior art keywords
- whale
- neural network
- data
- error
- output
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—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
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E40/00—Technologies for an efficient electrical power generation, transmission or distribution
- Y02E40/70—Smart grids as climate change mitigation technology in the energy generation sector
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Abstract
The invention discloses a wind energy resource data compensation method, which optimizes a BP neural network by utilizing a whale optimization algorithm, constructs a data compensation model, and can obtain corresponding output data by inputting meteorological elements so as to compensate missing data during observation. According to the invention, by utilizing the advantages of high convergence speed, high precision and easiness in jumping out of local optimum of the whale optimization algorithm in the aspect of searching optimization, the problem of low prediction precision caused by easiness in falling into local optimum of the BP neural network is solved by a manner of updating the weight threshold of the BP neural network and optimizing the whale. Meanwhile, aiming at the problem of low convergence speed in the BP neural network, the dynamic convergence factor of the whale optimization algorithm is improved, and the reliability and the precision of the algorithm are improved.
Description
Technical Field
The invention relates to a wind energy resource data compensation method.
Background
In the measurement process of wind energy resources, the problem of data loss often occurs, and compared with the ground wind measurement, the wind energy resource data measured by wind measurement modes such as ocean wind measurement, high-altitude wind measurement, complex mountain wind measurement and the like are difficult to supplement in real time. However, the conventional prediction algorithm, such as the BP neural network prediction algorithm, has poor overall reliability, timeliness and accuracy due to randomness of the initial weight, and is difficult to provide accurate data guarantee.
For example, the Chinese invention patent with the patent number 201210460427.0 provides a short-term power prediction method for a wind power plant based on a BP neural network, and the core idea of the method is as follows: the wind speed, the wind direction, the air density and the relative humidity are used as BP neural network input, the power generation output power of the wind power plant is used as BP neural network output, and the BP neural network is used for predicting the power generation output power of the wind power plant in a certain time period. The specific working steps are as follows: a. acquiring historical records of meteorological element data including wind speed, wind direction and air density at the location of a wind power plant and the power generation output power of the wind power plant corresponding to each record; b. correcting the wind speed, the wind direction and the air density into the wind speed, the wind direction and the air density at the hub of the wind turbine generator so as to generate corrected meteorological element data; c. inputting the corrected meteorological element data serving as input data into a BP (back propagation) neural network, and training the BP neural network by using the power generation output power of the wind power plant corresponding to each meteorological element data as the output of the BP neural network; d. acquiring meteorological element data including wind speed, wind direction and air density of the location of the wind power plant in a prediction time period according to the numerical weather forecast, and correcting the wind speed, the wind direction and the air density into the wind speed, the wind direction and the air density of a hub of a wind turbine generator, so as to generate corrected meteorological element data; e. and d, inputting the corrected meteorological element data obtained in the step d into the BP neural network trained in the step c, wherein the data output by the BP neural network is the power generation output power of the wind power plant in the prediction time period. The relative humidity at the hub is input as input data to the BP neural network.
The whole technical scheme still utilizes the BP neural network to predict the power generation output power of the wind power plant, the essence of the BP neural network still belongs to the category of the BP neural network, the BP neural network algorithm is used for predicting the short-term power of the wind power plant, and due to the fact that the method is prone to falling into the local optimal solution, the prediction precision is low, and the prediction result is unreliable. Based on the data, the invention provides a wind energy resource data compensation method, aiming at accurately compensating the missing data of the wind energy resource.
Disclosure of Invention
The invention aims to provide a wind energy resource data compensation method, which solves the problem of low prediction precision caused by the fact that a BP neural network is easy to fall into local optimum by utilizing the advantages that a whale optimization algorithm has high convergence speed and high precision in searching and optimizing and is easy to jump out of the local optimum through a mode that the weight threshold is updated by the BP neural network and the whale optimization is carried out at the same time. Meanwhile, aiming at the problem of low convergence speed in the BP neural network, the advantages of the whale optimization algorithm in the aspect of global optimization capability are utilized, the dynamic convergence factor of the whale optimization algorithm is improved, and the reliability of the algorithm is improved.
Therefore, the invention provides a wind energy resource data compensation method, a data compensation model is constructed through a BP neural network model based on a whale optimization algorithm to prevent the BP neural network from falling into local optimization, and corresponding output data can be obtained according to an input meteorological element model to compensate missing data during observation.
The invention has the advantages that: by utilizing the advantages of high convergence speed, high precision and easiness in jumping out of local optimum of the whale optimization algorithm in the aspect of searching optimization, the problem of low prediction precision caused by easiness in falling into local optimum of a BP neural network is solved by a mode of updating a weight threshold value of the BP neural network and simultaneously carrying out whale optimization. Meanwhile, aiming at the problem of low convergence speed in the BP neural network, the advantages of the whale optimization algorithm in the aspect of global optimization capability are utilized, the dynamic convergence factor of the whale optimization algorithm is improved, and the reliability and the precision of the algorithm are improved.
The present invention will be described in further detail below with reference to the accompanying drawings.
Drawings
FIG. 1 is a flow chart of the WOA-BP algorithm.
FIG. 2 is a graph of WOA-BP prediction error.
FIG. 3 is a graph comparing WOA-BP wind speed training and testing.
Fig. 4 is a diagram of a typical three-layer BP neural network architecture.
Fig. 5 is a flowchart of the BP neural network algorithm.
FIG. 6 is a flow chart of a whale optimization algorithm.
Fig. 7 a shrink wrap mechanism.
FIG. 8 illustrates a spiral position update mechanism.
Detailed Description
The embodiment provides a wind energy resource data compensation method, which has the advantages of strong reliability, accurate prediction result and the like by means of experimental simulation, and solves the problems of local minimization of a BP neural network and low convergence speed of an algorithm, wherein experimental data are as follows.
The method comprises the steps of adopting Chinese country-level ground station hour value data provided by a national weather information center, adopting 13 data such as atmospheric pressure, relative humidity, average temperature and the like every 3 hours from 3 months 16 days to 3 months 18 days in 2019 of Panzhihua city, sichuan as model input, simultaneously obtaining actual wind speed data of the region at the moment as model output, modeling and adopting a front 128-group data training model.
13 data such as atmospheric pressure, relative humidity, average temperature and the like in the 20 th group of data are used as model input, and wind speed data of 10 groups of data after compensation. The data obtained from the experiment are shown in FIGS. 2-3 and Table 1.
TABLE 1WOA-BP neural network part wind speed training table
13 types of data such as atmospheric pressure, relative humidity, average temperature and the like of the last 10 groups of data are input as models, and corresponding wind speed data are output. Experimental data show that the data compensation method has high reliability, low error and good performance, and can provide a reliable data basis for site selection planning of the wind power plant.
The wind energy resource data compensation method is mainly characterized in that a data compensation model is constructed through a BP neural network model based on a whale optimization algorithm to prevent the BP neural network from falling into local optimization, so that corresponding output data can be obtained according to an input meteorological element model to compensate missing data in an observation period.
In order to solve the above problem, the embodiment provides a wind energy resource data compensation method, which is mainly implemented as shown in fig. 1, and the main processes are as follows: initializing parameters, establishing a cellular array, calculating a fitness function (error), judging whether the error is smaller than the optimal solution, if so, taking the error as the current optimal solution, calculating parameters such as A, C, P and the like, if not, continuously calculating the parameters such as A, C, P and the like, and then performing traversal iteration of W1, B1, W2 and B2;
judging whether P is less than 0.5, if so, calculating whether the absolute value of A is greater than 1, if so, performing surrounding search to enlarge the search range and output the current optimal weight threshold, and if the absolute value of A is not greater than 1, searching predation, updating the search to the optimal solution until the current optimal weight threshold is output; then, storing the current optimal solution into a cellular array, recalculating the current optimal solution, and updating W1, B1, W2 and B2 by using a gradient descent method; judging whether the maximum iteration times is met, if so, outputting the current optimal solution and the error, and ending; if not, recalculating the fitness function, and executing the loop in sequence until finishing;
and if the P is not less than 0.5, a bubble network searching mode is spiraled, and local searching is expanded until the current optimal weight threshold value is output.
The specific operation process is as follows:
step 1: data normalization
Adopting a premnx normalization function to normalize the meteorological data such as atmospheric pressure, average temperature, relative humidity and the like so as to facilitate calculation, wherein for example, the atmospheric pressure of an input training sample is 761.7 hectopascal, the average temperature is-2.6 ℃, the relative humidity is 65%, the output wind speed of the training sample is 0.7m/s, and a new group of input data SamIn is generated to be [1, -1, -0.8231]; the wind speed is normalized to generate another new set of output data tnOut of 0.7. Wherein SamIn and tnOut will vary with the number of samples;
step 2: and initializing the neural network parameters according to the normalization result, namely initializing parameters shown in figure 1, and establishing the cellular array.
Adding noise with a numerical value of 0.01 into the normalized output data, aiming at preventing overfitting of the BP neural network, and comprising the following specific steps of:
a. adding the noise and the normalized output data, and marking as SamOut, as shown in formula 1;
SamOut = tnOut + Noise equation 1
b. The Noise is obtained by multiplying a random number by the Noise intensity, and is shown in a formula 2;
noise = NoiseVar Y formula 2
c. The noise intensity NoiseVar is typically 0.01;
d. the random number is obtained by equation 3.
Y = rand () formula 3
Establishing a BP neural network model according to the number of the initialized data in the step 1, inputting the number InDim (3 in the example), the number HiddenUnitNum (8 in the example) of the hidden layer nodes, and outputting the number OutDim (1 in the example) of the layer nodes;
initializing model parameters according to the established BP neural network model, which specifically comprises the following steps:
the learning rate of the BP neural network model is a dynamic learning rate, as shown in formula 4.
lr =1-t/max _ iteration formula 4
The activation function adopted by the BP neural network model is a sigmoid function, and is shown in a formula 5.
And step 3: initializing whale optimization algorithm parameters
The invention aims at the linear convergence factor of a whale optimization algorithm of a data compensation problem
Improvements are made, improved dynamic convergence factors>
As shown in equation 6.
t is the current iteration number, and max _ iteration is the maximum iteration number.
Convergence factor of improved whale optimization algorithm
The variation with the number of iterations is shown in equation 7.
t is the current iteration number, and max _ iteration is the maximum iteration number.
The weight threshold initialization method of the whale optimization algorithm specifically comprises the following steps:
for the calculation of the whale optimization algorithm weight and the threshold, the weight omega between the input layer and the hidden layer is updated by adopting the following formula 1 Threshold B between input layer and hidden layer 1 Weight ω between output layer and hidden layer 2 Threshold B between output layer and hidden layer 2 The initialization is performed using the formulas shown in formulas 8 to 11, respectively.
ω 1 =0.5 × rand (HiddenUnitNum, inDim)) -0.1 equation 8
B 1 =0.5 × rand (HiddenUnitNum, 1) -0.1 formula 9
ω 2 =0.5 × rand (OutDim, hiddenUnitNum) -0.1 equation 10
B 2 =0.5 × rand (OutDim, 1) — 0.1 formula 11
And 4, step 4: calculation of fitness function (error) and current best search agent according to BP neural network model
The fitness function (error function) is shown in equations 11-14.
HiddenOut＝sigmoid(ω 1 *SamIn+repmat(B 1 1, samnum)) formula 12
NetworkOut＝ω 2 *HiddenOut+repmat(B 2 1, samNum) formula 13
Error = SamOut-NetworkOut equation 14
SSE = sumsqr (Error) equation 15
HiddenOut is the hidden layer node output, networkOut is the output layer node output, error is the Error between the network output value and the test data, and SSE is the sum of squares Error.
And 5: updating weight threshold values by utilizing whale optimization algorithm
Recording a group of weights and thresholds in the BP neural network as a search agent, namely a whale;
calculating the minimum fitness value of the whale population by taking the error Sum of Squares (SSE) as a fitness function;
comparing the fitness value of each whale with the individual optimal whale to obtain the current optimal whale position;
and updating the whale population position according to a contraction surrounding mechanism and a spiral position updating mechanism.
Step 6: updating weight threshold by gradient descent method
And updating the weight threshold of the current optimal whale position by using a gradient descent method (neural network back propagation), wherein the updating formula of the weight threshold of the gradient descent method is shown as a formula 16-21.
Δ 2 = Error equation 16
Δ 1 ＝ω 2 T *Δ 2 * Hiddenout (1-Hiddenout) formula 17
In the formula for updating weight threshold, ω 1 Is omega 1 Updated weight, ω 2 ' is omega 2 Updated weight, B 1 ' is B 1 Updated threshold, B 2 ' is B 2 The updated threshold.
And 7: training end judgment
The end conditions of the training are as follows:
the target error reaches a preset value;
the iteration times reach the designated times.
If yes, the model training is forcibly finished. If not, returning to the step 2 of calculating the fitness function and the current optimal search agent, and continuing training
And step 8: outputting training model for data compensation
After the method reaches an end condition, a predictive compensation model is generated.
And inputting data such as any atmospheric pressure, relative humidity, average temperature and the like as the prediction compensation model, so that corresponding wind speed data output can be obtained to compensate missing data in the observation period.
To facilitate understanding of the wind energy resource data compensation method provided by the above embodiments, the present embodiment is described with reference to the following basic knowledge in conjunction with the accompanying drawings:
fig. 4 is a diagram showing a structure of a typical three-layer BP neural network, which is essentially an algorithm for solving the optimized network weights. At this time, the objective function (cost function) of the optimization problem is an error function formed by errors between the network output and the expected output, variables to be optimized are all weights and thresholds in the network, the BP neural network is a correction method for solving the network weights and thresholds which can enable the error function to reach minimum by using a gradient descent method, and a BP neural network algorithm flow chart is shown in fig. 5.
Step 1: initialization weight threshold
The weight of the network connection between every two neurons is initialized to a small random number, and each neuron has a bias value, which is also initialized to a random number. For each input sample X, processing is performed as per step 2.
Step 2: forward propagation input (calculating the input and output of each neuron of the hidden layer and the output layer) provides the input layer of the network according to the training sample X, the output of each neuron is obtained through calculation,
as shown in equation 22.
Note: omega ij Is the network weight from the unit i of the upper layer to the unit j; o is i Is the output of the cell i of the previous layer; a is the bias of the cell, which acts as a threshold to change the activity of the cell.
As can be seen from the above equation, the output of neuron j depends on its total input S j ＝∑ω ij *O i +A i Then by activating the function
The final output is obtained, the activation function is called a sigmoid function, a larger input value can be mapped to a value between the interval 0 and the interval 1, and the function is nonlinear and differentiable, so that the BP neural network algorithm can model a linear inseparable classification problem, and the application range of the BP neural network algorithm is greatly expanded.
And step 3: reverse error propagation
From step 2, the actual output is finally obtained at the output layer, and the error of each output unit can be obtained by comparing with the expected output: (T) j Is the expected output of the output unit) as shown in equation 23.
E j ＝O j (1-O j )(T j -O j ) Equation 23
The resulting error needs to propagate from back to front, and the error of the cell of the previous layer can be calculated from the errors of all the cells of the next layer connected to it, as shown in equation 24.
E j ＝O j (1-O j )∑ k ω jk E k Equation 24
And sequentially obtaining the error of each neuron from the last hidden layer to the first hidden layer.
And 4, step 4: network weight and neuron bias adjustment (weight threshold of modified hidden layer and output layer)
The adjustment of the weights is performed sequentially from the connection weight of the input layer to the first hidden layer, and each connection weight is adjusted according to the following formula, as shown in formula 25.
ω ij ＝ω ij +Δω ij ＝ω ij +(lr)O i E j Equation 25
Note: where lr is the dynamic learning rate, as shown in equation 4.
lr =1-t/max _ iteration formula 4
The adjustment method of the neuron bias is to update each neuron as shown in equation 26.
θ j ＝θ j +Δθ j ＝θ j +(lr)E j Equation 26
And 5: end of judgment
For each sample, if the final output error is smaller than the acceptable range or the iteration number reaches a certain threshold, selecting the next sample, and turning to the step 2 to continue executing again; otherwise, adding 1 to the iteration number, and then turning to the step 2 to continue training by using the current sample.
The BP neural network has good self-learning and self-adaption capability, fault-tolerant capability and nonlinear mapping capability. But again this algorithm has certain drawbacks, among which the main ones include the following two points:
1. local minimization problem: from the mathematical perspective, the traditional BP neural network is an optimization method of local search, and a complex nonlinear problem is solved, the weight of the network is gradually adjusted along the direction of local improvement, so that the algorithm falls into a local extreme value, and the weight converges to a local minimum point, thereby causing the failure of network training. In addition, the BP neural network is very sensitive to initial network weights, and initializing the network with different weights tends to converge on different local minima, which is also the root cause of different results obtained from each training.
2. The convergence speed of the algorithm is low: since the BP neural network algorithm is essentially a gradient descent method, the objective function to be optimized is very complex, and therefore, a 'saw-tooth phenomenon' inevitably occurs, so that the BP algorithm is low in efficiency; because the optimized objective function is complex, certain flat areas inevitably appear under the condition that the output of the neuron is close to 0 or 1, and in the areas, the weight value is changed little, so that the training process is almost stopped; in the BP neural network model, in order to make the network execute the BP algorithm, the step size of each iteration cannot be obtained by using the conventional one-dimensional search method, and the update rule of the step size must be given to the network in advance, which also causes the algorithm to be inefficient. The above results in a phenomenon of slow convergence rate of the BP neural network algorithm.
(2) Whale optimization algorithm
The whale optimization algorithm is a natural heuristic algorithm, simulates the predation behavior of the whale, swims upwards in a spiral posture from the deep part of the sea bottom and spits bubbles, and when the last spitted bubble and the first spitted bubble rise to the water surface at the same time, a bubble net is formed, so that a prey is tightly surrounded like a huge net, and the prey is forced to the center of the net, and the large mouth is opened almost vertically in a bubble ring to swallow the prey collected by the net. The algorithm flow chart is shown in fig. 6.
Specifically, the method comprises the following steps:
calculating the current minimum error by adopting an individual fitness function, taking a weight threshold value (the best whale) of the minimum error as the best search agent of the iteration of the round, and updating the individual position of the search agent by utilizing a whale optimization algorithm, wherein the updating method comprises the following steps:
a. and (3) a prey surrounding stage:
whales with whales can identify the location of the prey and surround it. Since the location of the optimal design in the search space is not known a priori, the WOA algorithm assumes that the current best candidate solution is the target prey or near-optimal. After the best search agent is defined, other search agents will attempt to update their location to the best search agent. This behavior is shown in equations 27, 28.
In the equation (a) for the case,
is the distance of the search agent from the optimal search agent, t is the number of current iterations, and->
And &>
Is a coefficient, X * For the position vector of the best solution obtained at present, X is the position vector, | | is the absolute value. It is worth mentioning here that if there is a better solution, then X * Will be updated in each iteration. Vector->
And &>
The calculation formula of (c) is shown in formulas 29 and 30.
b. A bubble network attack stage:
mathematical simulation is carried out on the behavior of whale-sitting in the air that the bubble net attacks the prey, and two methods are designed:
a shrink wrap mechanism:
Spiral position update mechanism
The spiral position update procedure first calculates whale at (X, Y) and whale at (X) * ,Y * ) The distance between preys. Then a spiral position update equation is established between the whale and prey positions as shown in figure 8, and the spiral motion of the whale with a standing head is simulated as shown in equation 31.
In the process of the movement, the user can exercise,
represents the distance between the ith whale and prey (the current best search agent), b is a constant defining the shape of a logarithmic spiral, and l is [ -1,1]The random number of (1).
The shrink wrap-around mechanism and the spiral position update are performed with a probability of 0.5 during the loop iteration, as shown in equation 32.
After the agent search position is updated through the prey surrounding stage and the bubble network attack stage, the fitness function is recalculated, and the current optimal search agent is updated.
Claims (2)
1. A wind energy resource data compensation method is characterized by comprising the following steps:
optimizing the BP neural network by improving a whale optimization algorithm to obtain an optimized neural network weight threshold; constructing a data compensation model according to the optimized weight threshold of the neural network, and inputting meteorological elements based on the data compensation model to obtain corresponding output data so as to obtain missing data during compensation observation;
the improved whale optimization algorithm specifically comprises the following steps:
linear convergence factor for whale optimization algorithm
The improvement is made, as shown in equation 6,
t is the current iteration number, and max _ iteration is the maximum iteration number;
the improved whale optimization algorithm further comprises the following processes:
calculating the minimum fitness value of the whale population by taking the error Sum of Squares (SSE) as a fitness function;
comparing the fitness value of each whale with the individual optimal whale to obtain the current optimal whale position;
completing the whale population position updating according to a contraction surrounding mechanism and a spiral position updating mechanism;
updating the weight threshold of the current optimal whale position by using a gradient descent method, updating the weight threshold by using the gradient descent method, wherein an updating formula is shown as an equation 16-21,
Δ 2 = Error equation 16
Δ 1 ＝ω 2 T *Δ 2 * HiddenOut formula 17
In the formula for updating weight threshold, ω 1 Is omega 1 Updated weight, ω 2 Is omega 2 Updated weight, B 1 ' is B 1 Updated threshold, B 2 ' is B 2 The updated threshold.
2. The wind energy resource data compensation method of claim 1, wherein the method comprises the following steps: the sum of squared errors SSE, which is used as the fitness function value, is calculated as shown in equations 12-15,
HiddenOut＝sigmoid(ω 1 *SamIn+repmat(B 1 1, samnum)) formula 12
NetworkOut＝ω 2 *HiddenOut+repmat(B 2 1, samNum) formula 13
Error = SamOut-NetworkOut equation 14
SSE = sumsqr (Error) equation 15
The HiddeNout is output of an implied layer node, the NetworkOut is output of an output layer node, the Error is an Error between a network output value and test data, and the SSE is the sum of squares of the errors.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910481772.4A CN110322050B (en) | 2019-06-04 | 2019-06-04 | Wind energy resource data compensation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910481772.4A CN110322050B (en) | 2019-06-04 | 2019-06-04 | Wind energy resource data compensation method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110322050A CN110322050A (en) | 2019-10-11 |
| CN110322050B true CN110322050B (en) | 2023-04-07 |
Family
ID=68119396
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910481772.4A Active CN110322050B (en) | 2019-06-04 | 2019-06-04 | Wind energy resource data compensation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110322050B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111948283A (en) * | 2020-07-07 | 2020-11-17 | 上海交通大学 | System and method for detecting oxidation stability of transformer oil |
| CN113240307B (en) * | 2021-05-24 | 2022-09-27 | 上海电力大学 | Power system economic dispatching method based on improved differential evolution algorithm |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0295424A2 (en) * | 1987-06-08 | 1988-12-21 | International Business Machines Corporation | Method for managing subpage concurrency control and partial transaction rollback in a transaction-oriented system of the write-ahead logging type |
| US8407020B1 (en) * | 2010-09-28 | 2013-03-26 | The United States Of America As Represented By The Secretary Of The Navy | Fast method to search for linear frequency-modulated signals |
| CN107292051A (en) * | 2017-07-07 | 2017-10-24 | 湘潭大学 | A kind of carbide chip chemically-mechanicapolish polishes the Forecasting Methodology of surface roughness |
| CN107911300A (en) * | 2017-10-25 | 2018-04-13 | 西南交通大学 | Multicast routing optimization method based on whale algorithm and its application on Spark platforms |
| CN108021658A (en) * | 2017-12-01 | 2018-05-11 | 湖北工业大学 | A kind of big data intelligent search method and system based on whale optimization algorithm |
| CN108205698A (en) * | 2017-06-15 | 2018-06-26 | 广东工业大学 | A kind of cloud resource load predicting method based on the double string whale optimization algorithms of just remaining chaos |
| CN108830431A (en) * | 2018-08-03 | 2018-11-16 | 广东工业大学 | A kind of Electricity price forecasting solution and relevant apparatus based on whale optimization algorithm |
| CN109101738A (en) * | 2018-08-24 | 2018-12-28 | 河北工业大学 | A kind of IGBT module degree of aging appraisal procedure |
| CN109214449A (en) * | 2018-08-28 | 2019-01-15 | 华北电力大学 | A kind of electric grid investment needing forecasting method |
| CN109345005A (en) * | 2018-09-12 | 2019-02-15 | 中国电力科学研究院有限公司 | A kind of integrated energy system multidimensional optimization method based on improvement whale algorithm |
| CN109765893A (en) * | 2019-01-17 | 2019-05-17 | 重庆邮电大学 | Method for planning path for mobile robot based on whale optimization algorithm |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8015132B2 (en) * | 2008-05-16 | 2011-09-06 | Samsung Electronics Co., Ltd. | System and method for object detection and classification with multiple threshold adaptive boosting |
| US10387773B2 (en) * | 2014-10-27 | 2019-08-20 | Ebay Inc. | Hierarchical deep convolutional neural network for image classification |
-
2019
- 2019-06-04 CN CN201910481772.4A patent/CN110322050B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0295424A2 (en) * | 1987-06-08 | 1988-12-21 | International Business Machines Corporation | Method for managing subpage concurrency control and partial transaction rollback in a transaction-oriented system of the write-ahead logging type |
| US8407020B1 (en) * | 2010-09-28 | 2013-03-26 | The United States Of America As Represented By The Secretary Of The Navy | Fast method to search for linear frequency-modulated signals |
| CN108205698A (en) * | 2017-06-15 | 2018-06-26 | 广东工业大学 | A kind of cloud resource load predicting method based on the double string whale optimization algorithms of just remaining chaos |
| CN107292051A (en) * | 2017-07-07 | 2017-10-24 | 湘潭大学 | A kind of carbide chip chemically-mechanicapolish polishes the Forecasting Methodology of surface roughness |
| CN107911300A (en) * | 2017-10-25 | 2018-04-13 | 西南交通大学 | Multicast routing optimization method based on whale algorithm and its application on Spark platforms |
| CN108021658A (en) * | 2017-12-01 | 2018-05-11 | 湖北工业大学 | A kind of big data intelligent search method and system based on whale optimization algorithm |
| CN108830431A (en) * | 2018-08-03 | 2018-11-16 | 广东工业大学 | A kind of Electricity price forecasting solution and relevant apparatus based on whale optimization algorithm |
| CN109101738A (en) * | 2018-08-24 | 2018-12-28 | 河北工业大学 | A kind of IGBT module degree of aging appraisal procedure |
| CN109214449A (en) * | 2018-08-28 | 2019-01-15 | 华北电力大学 | A kind of electric grid investment needing forecasting method |
| CN109345005A (en) * | 2018-09-12 | 2019-02-15 | 中国电力科学研究院有限公司 | A kind of integrated energy system multidimensional optimization method based on improvement whale algorithm |
| CN109765893A (en) * | 2019-01-17 | 2019-05-17 | 重庆邮电大学 | Method for planning path for mobile robot based on whale optimization algorithm |
Non-Patent Citations (9)
| Title |
|---|
| Credit Card Fraud Detection Based on Whale Algorithm Optimized BP Neural Network;Wang Chunzhi 等;《The 13th International Conference on Computer Science & Education》;20180831;第614-617页 * |
| Optimal renewable resources placement in distribution networks by combined power loss index and whale optimization algorithms;P. Dinakara Prasad Reddy 等;《Journal of Electrical Systems and Information Technology》;20170613;第5卷;第175-191页 * |
| Optimizing connection weights in neural networks using the whale optimization algorithm;Ibrahim Aljarah 等;《Soft Computing》;20161121;第22卷(第1期);第1-15页 * |
| The Whale Optimization Algorithm;Seyedali Mirjalili 等;《Advances in Engineering Software》;20160226;第95卷;第51-67页 * |
| 基于多目标鲸鱼优化的关键质量特性识别方法;李岸达 等;《系统工程》;20190131;第37卷(第1期);第134-142页 * |
| 基于多目标鲸鱼优化算法和Elman神经网络的短期风速预测模型的研究与应用;杜沛;《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》;20180715(第7期);第C042-267页 * |
| 基于多预测模型与非线性组合的神经网络风速预测;王家君 等;《物联网学报》;20211231;第5卷(第4期);第81-89页 * |
| 基于鲸鱼优化算法的汽轮机热耗率模型预测;牛培峰;《化工学报》;20170331;第68卷(第3期);第1049-1057页 * |
| 收敛因子非线性变化的鲸鱼优化算法;龙文 等;《兰州理工大学学报》;20171231;第43卷(第6期);第102-107页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110322050A (en) | 2019-10-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP5888640B2 (en) | Photovoltaic power generation prediction apparatus, solar power generation prediction method, and solar power generation prediction program | |
| CN113326969B (en) | Short-term wind speed prediction method and system based on improved whale optimization algorithm and ELM | |
| CN110084424A (en) | A kind of Methods of electric load forecasting based on LSTM and LGBM | |
| CN108985515B (en) | New energy output prediction method and system based on independent cyclic neural network | |
| CN111461457A (en) | Foundation pit displacement prediction method based on particle swarm optimization BP neural network | |
| CN111144663B (en) | Ultra-short-term wind power prediction method for offshore wind farm considering output fluctuation process | |
| CN110322050B (en) | Wind energy resource data compensation method | |
| CN113468803A (en) | Improved WOA-GRU-based flood flow prediction method and system | |
| CN105574615A (en) | Spatial correlation and genetic algorithm (GA) based wind power forecast method for wavelet-BP neural network | |
| CN114912673A (en) | Water level prediction method based on whale optimization algorithm and long-term and short-term memory network | |
| CN115994629B (en) | GN-RBF-based air humidity prediction method and system | |
| CN112396152A (en) | Flood forecasting method based on CS-LSTM | |
| Syama et al. | A hybrid extreme learning machine model with Lévy flight chaotic whale optimization algorithm for wind speed forecasting | |
| CN105138729B (en) | Based on PSO GRNN wind power plant wind turbine defect air speed value fill methods | |
| CN113722980A (en) | Ocean wave height prediction method, system, computer equipment, storage medium and terminal | |
| El Marghichi et al. | A solar PV model parameters estimation based on an improved manta foraging algorithm with dynamic fitness distance balance | |
| CN116526478B (en) | Short-term wind power prediction method and system based on improved snake group optimization algorithm | |
| CN112307672A (en) | BP neural network short-term wind power prediction method based on cuckoo algorithm optimization | |
| CN115983494A (en) | Short-term wind power prediction method and system for newly-built small-sample wind power plant | |
| CN113610665B (en) | Wind power generation power prediction method based on multi-delay output echo state network | |
| CN114897204A (en) | Method and device for predicting short-term wind speed of offshore wind farm | |
| KR102515608B1 (en) | Photovoltaic forecasting method and system | |
| CN109685242B (en) | Photovoltaic ultra-short term combined prediction method based on Adaboost algorithm | |
| CN113052379A (en) | Tomato water demand prediction method based on BAS-RBF | |
| CN117471575B (en) | Typhoon wave height forecasting method based on BO-LSTM neural network model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |