A supercapacitor energy storage frequency modulation optimization method, medium, and program product

Abstract

The present invention discloses an optimization method, medium, and program product for supercapacitor energy storage and frequency modulation, and belongs to the technical field of wind power generation. The optimization method for supercapacitor energy storage and frequency modulation provided by the present invention, by iteratively constructing a wind speed sequence, and combining sigmoid and tanh activation functions, dynamically adjusts according to the characteristics of the ratio wind speed sequence, can gradually approach the actual wind speed change, improve the accuracy of wind speed prediction, and provide a solid foundation for the coordinated operation of wind farms and power grids. At the same time, according to the wind speed prediction results, the power sent by the wind farm to the power grid can be adjusted in real time, and the power balance can be achieved through the charging and discharging state of the supercapacitor energy storage system. Not only does it avoid the waste of electric energy, but it also improves the utilization efficiency of the energy storage system and extends the service life of the energy storage equipment.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power generation, and in particular to an optimization method, medium and program product for supercapacitor energy storage and frequency modulation. Background Art

[0002] Wind power generation, as a clean, environmentally friendly and sustainable renewable energy technology, has attracted much attention and has developed rapidly in recent years. With the rapid development of technology and the continuous reduction of costs, wind power generation has gradually become an indispensable part of the energy structure of many countries.

[0003] However, the natural randomness of wind speed poses a severe challenge to the grid connection of large-scale wind farms. This randomness causes the output power of wind farms to show volatility and uncertainty, which brings great difficulty to the grid dispatching agency in prediction and arrangement. This not only increases the operating burden of the grid, but also may pose a potential threat to the stability of the system. As one of the core technologies in the field of wind power generation, the accuracy and reliability of wind speed prediction are crucial. Wind farm scheduling is the key to ensuring the coordinated operation between wind farms and power grids. It involves intelligent control and optimization of the output power and operating status of wind farms to balance the output power of wind farms and the actual needs of the power grid, and plays an important role in wind farm scheduling.

[0004] Although wind power forecasting and scheduling technology has made significant progress and has been widely used, it still faces many challenges. In the field of wind speed forecasting in wind farms, traditional forecasting methods mostly rely on decomposition techniques, such as VMD (Variational Mode Decomposition) and EEMD (Ensemble Empirical Mode Decomposition), to preprocess the original wind speed data. The core of these technologies is to decompose the original data into a series of sub-signals with different frequencies and characteristics, so as to more accurately capture the inherent laws and characteristics of wind speed changes. However, these traditional decomposition techniques have exposed some problems in practical applications. EEMD, as an improved EMD (Empirical Mode Decomposition) method, although it overcomes the modal aliasing problem of EMD by adding white noise, is prone to large errors at the endpoints of the data when processing data due to the introduction of white noise and the nonlinear characteristics of the signal itself, the so-called "endpoint effect", which affects the accuracy of the decomposition results and weakens the accuracy of subsequent wind speed forecasts. VMD, as an emerging decomposition method, solves the modal components of the signal through an iterative optimization algorithm. However, in practical applications, VMD also faces the challenge of modal aliasing. Modal aliasing refers to the fact that signal components of different frequencies are mistakenly classified into the same mode during the decomposition process, or signal components of the same frequency are dispersed into multiple modes. This phenomenon will lead to inaccurate decomposition results, thus affecting the reliability of wind speed prediction.

[0005] Therefore, how to effectively decompose wind speed data and improve the accuracy and reliability of wind speed prediction has become a key technical problem that needs to be urgently solved by technical personnel in this field. Summary of the invention

[0006] The purpose of the present invention is to provide a supercapacitor energy storage frequency modulation optimization method, medium, and program product, aiming to fundamentally solve the problem of insufficient wind speed prediction accuracy and reliability in traditional methods.

[0007] The present invention solves the above technical problems through the following technical solutions:

[0008] A method for optimizing supercapacitor energy storage frequency modulation comprises the following steps:

[0009] Step 1: Collect wind speed data and establish an original wind speed sequence; divide the original wind speed sequence into a prediction set wind speed sequence and a verification set wind speed sequence according to a preset ratio;

[0010] The wind speed sequence of the forecast set is X0(t)=[X 01 ,X 02 ,X 03,...,X 0N ], N is the number of subsequences in the wind speed sequence of the prediction set, X 01 ,X 02 ,X 03 ,...,X 0N are the 1st, 2nd, 3rd, ..., Nth subsequences in the wind speed sequence of the prediction set respectively; the wind speed sequence of the prediction set is input into the trained neural network model for prediction to obtain the predicted value, and the benchmark prediction error W is calculated based on the predicted value and the actual value in the wind speed sequence of the verification set;

[0011] Step 2: Calculate the ratio of the latter subsequence to the previous subsequence in the wind speed sequence X0(t) of the forecast set, and obtain the ratio wind speed sequence q(t) = [q1, q2, q3, ..., q N-1 ],q1,q2,q3,...,q N-1 They are the ratios of the second subsequence to the first subsequence, the third subsequence to the second subsequence, ..., the Nth subsequence to the N-1th subsequence in the prediction set wind speed sequence X0(t); the number of subsequences A less than or equal to 1 and the number of subsequences B greater than 1 in the ratio wind speed sequence are counted, B=N-1-A; let L=1, L is the round of constructing the wind speed sequence, and use the prediction set wind speed sequence X0(t) as the first round to construct the wind speed sequence;

[0012] Step 3: Based on the Lth round constructed wind speed sequence and the ratio wind speed sequence q(t), the L+1th round constructed wind speed sequence is obtained by the construction method;

[0013] The construction method specifically comprises the following steps: based on the first subsequence of the wind speed sequence constructed in the Lth round, the first subsequence of the wind speed sequence constructed in the L+1th round is obtained; when A≥B, the sigmoid activation function is used to calculate the remaining subsequence of the wind speed sequence constructed in the L+1th round; when A<B, the tanh activation function is used to calculate the remaining subsequence of the wind speed sequence constructed in the L+1th round, and the first subsequence and the remaining subsequence of the wind speed sequence constructed in the L+1th round are summarized to obtain the wind speed sequence constructed in the L+1th round;

[0014] Step 4: Based on the wind speed sequence X0(t) of the prediction set and all rounds of constructed wind speed sequences obtained in step 3, calculate the residual wind speed sequence of this round. The residual wind speed sequence of this round = the first round of constructed wind speed sequence - all rounds of constructed wind speed sequences obtained in step 3;

[0015] Step 5: Bring the residual wind speed sequence of this round into the trained neural network model for prediction to obtain the prediction error of this round; determine whether the prediction error of this round is ≥ 0.1 times the benchmark prediction error W. If the judgment result is yes, set L = L + 1 and return to step 3; if the judgment result is no, bring all rounds of constructed wind speed sequences obtained in step 3 and the residual wind speed sequence of this round into the trained neural network model for prediction to obtain the final wind speed prediction result;

[0016] Step 6. Calculate the predicted power sent by the wind farm to the grid based on the final wind speed prediction result, obtain the actual power sent by the wind farm to the grid, and calculate the difference between the predicted power and the actual power. When the difference is positive, adjust the super-capacity energy storage system connected to the wind farm exit to the charging state; when the difference is 0, the super-capacity energy storage system connected to the wind farm exit maintains the current state unchanged; when the difference is negative, adjust the super-capacity energy storage system connected to the wind farm exit to the discharging state.

[0017] Furthermore, the preset ratio is 9:1.

[0018] Furthermore, the benchmark prediction error W is specifically:

[0019]

[0020] Among them, y i is the ith predicted value, i is the i-th actual value in the wind speed sequence of the validation set, and na is the number of subsequences in the wind speed sequence of the validation set.

[0021] Furthermore, in the L+1th round, the first subsequence X of the wind speed sequence is constructed a1 Specifically:

[0022] X a1 =rand( , 3 )

[0023] Among them, rand() is a random function, a is the number of constructions, Construct the first subsequence of the wind speed sequence for the Lth round.

[0024] Furthermore, when A≥B, the sigmoid activation function is used to calculate the remaining subsequence of the wind speed sequence constructed in the L+1th round as follows:

[0025] X an =X a（n-1） q n-1 sigmoid( )

[0026] Where n∈{2,3,…,N}, a is the number of constructions, X an is the nth subsequence in the remaining subsequence, X a（n-1） is the n-1th subsequence in the remaining subsequence, q n-1 is the n-1th subsequence in the ratio wind speed sequence, and rand() is a random function.

[0027] Furthermore, when A<B, the tanh activation function is used to calculate the remaining subsequence of the wind speed sequence constructed in the L+1th round as follows:

[0028] X an =X a（n-1） q n-1 tanh( )

[0029] Where n∈{2,3,…,N}, a is the number of constructions, X an is the nth subsequence in the remaining subsequence, X a（n-1） is the n-1th subsequence in the remaining subsequence, q n-1 is the n-1th subsequence in the ratio wind speed sequence, and rand() is a random function.

[0030] Furthermore, after collecting the wind speed data, the following steps are also included: deleting and supplementing abnormal data.

[0031] Furthermore, the neural network model is LSTM (Long Short-Term Memory).

[0032] A computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0033] A computer-readable storage medium stores a computer program, which implements the steps of the above method when executed by a processor.

[0034] Compared with the prior art, the positive and progressive effects of the present invention are:

[0035] The optimization method of supercapacitor energy storage frequency modulation provided by the present invention, through iterative construction of wind speed sequence, combined with sigmoid and tanh activation functions, is dynamically adjusted according to the characteristics of the ratio wind speed sequence, and can gradually approach the actual wind speed change, thereby improving the accuracy of wind speed prediction and providing a solid foundation for the coordinated operation of wind farms and power grids. At the same time, according to the wind speed prediction results, the power sent by the wind farm to the power grid can be adjusted in real time, and the power balance can be achieved through the charging and discharging state of the supercapacitor energy storage system. Not only does it avoid the waste of electric energy, but it also improves the utilization efficiency of the energy storage system and extends the service life of the energy storage equipment.

[0036] Furthermore, the use of LSTM neural network can effectively process the noise and outliers in the input data, thereby improving the robustness of the prediction model, adapting to complex and changeable wind speed changes, and optimizing the model training process.

[0037] Furthermore, by deleting abnormal data, erroneous data points caused by equipment failure, measurement errors, or data transmission errors can be eliminated, thereby improving the overall quality of the wind speed dataset and improving the accuracy and credibility of the prediction results; at the same time, reducing the model's dependence on abnormal data, thereby enhancing the model's generalization ability and enabling it to better adapt to wind speed prediction tasks in different scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The drawings in the specification are used to provide further understanding of the present invention and constitute a part of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0039] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments, which are intended to explain the present invention rather than to limit it.

[0042] The present invention provides a method for optimizing supercapacitor energy storage frequency modulation, comprising the following steps:

[0043] Step 1: Collect wind speed data and establish an original wind speed sequence; divide the original wind speed sequence into a prediction set wind speed sequence and a verification set wind speed sequence according to a preset ratio; the prediction set wind speed sequence is X0(t)=[X 01 ,X 02 ,X 03 ,...,X 0N ], N is the number of subsequences in the wind speed sequence of the prediction set, X 01 ,X 02 ,X 03 ,...,X 0N are the 1st, 2nd, 3rd, ..., Nth subsequences in the wind speed sequence of the prediction set respectively; the wind speed sequence of the prediction set is input into the trained neural network model for prediction to obtain the predicted value, and the benchmark prediction error W is calculated based on the predicted value and the actual value in the wind speed sequence of the verification set;

[0044] Step 2: Calculate the ratio of the latter subsequence to the previous subsequence in the wind speed sequence X0(t) of the forecast set, and obtain the ratio wind speed sequence q(t) = [q1, q2, q3, ..., q N-1 ],q1,q2,q3,...,q N-1 They are the ratios of the second subsequence to the first subsequence, the third subsequence to the second subsequence, ..., the Nth subsequence to the N-1th subsequence in the prediction set wind speed sequence X0(t); the number of subsequences A less than or equal to 1 and the number of subsequences B greater than 1 in the ratio wind speed sequence are counted, B=N-1-A; let L=1, L is the round of constructing the wind speed sequence, and use the prediction set wind speed sequence X0(t) as the first round to construct the wind speed sequence;

[0045] Step 3: Based on the Lth round constructed wind speed sequence and the ratio wind speed sequence q(t), the L+1th round constructed wind speed sequence is obtained by the construction method;

[0046] The construction method specifically comprises the following steps: based on the first subsequence of the wind speed sequence constructed in the Lth round, the first subsequence of the wind speed sequence constructed in the L+1th round is obtained; when A≥B, the sigmoid activation function is used to calculate the remaining subsequence of the wind speed sequence constructed in the L+1th round; when A<B, the tanh activation function is used to calculate the remaining subsequence of the wind speed sequence constructed in the L+1th round, and the first subsequence and the remaining subsequence of the wind speed sequence constructed in the L+1th round are summarized to obtain the wind speed sequence constructed in the L+1th round;

[0047] Step 4: Based on the wind speed sequence X0(t) of the prediction set and all rounds of constructed wind speed sequences obtained in step 3, calculate the residual wind speed sequence of this round. The residual wind speed sequence of this round = the first round of constructed wind speed sequence - all rounds of constructed wind speed sequences obtained in step 3;

[0048] Step 5: Bring the residual wind speed sequence of this round into the trained neural network model for prediction to obtain the prediction error of this round; determine whether the prediction error of this round is ≥ 0.1 times the benchmark prediction error W. If the judgment result is yes, set L = L + 1 and return to step 3; if the judgment result is no, bring all rounds of constructed wind speed sequences obtained in step 3 and the residual wind speed sequence of this round into the trained neural network model for prediction to obtain the final wind speed prediction result;

[0049] Step 6. Calculate the predicted power sent by the wind farm to the grid based on the final wind speed prediction result, obtain the actual power sent by the wind farm to the grid, and calculate the difference between the predicted power and the actual power. When the difference is positive, adjust the super-capacity energy storage system connected to the wind farm exit to the charging state; when the difference is 0, the super-capacity energy storage system connected to the wind farm exit maintains the current state unchanged; when the difference is negative, adjust the super-capacity energy storage system connected to the wind farm exit to the discharging state.

[0050] The optimization method of supercapacitor energy storage frequency modulation provided by the present invention, through iterative construction of wind speed sequence, combined with sigmoid and tanh activation functions, is dynamically adjusted according to the characteristics of the ratio wind speed sequence, and can gradually approach the actual wind speed change, thereby improving the accuracy of wind speed prediction and providing a solid foundation for the coordinated operation of wind farms and power grids. At the same time, according to the wind speed prediction results, the power sent by the wind farm to the power grid can be adjusted in real time, and the power balance can be achieved through the charging and discharging state of the supercapacitor energy storage system. Not only does it avoid the waste of electric energy, but it also improves the utilization efficiency of the energy storage system and extends the service life of the energy storage equipment.

[0051] As a preferred embodiment of the present invention, the preset ratio is 9:1.

[0052] Specifically, the benchmark prediction error W is:

[0053]

[0054] Among them, y i is the ith predicted value, i is the i-th actual value in the wind speed sequence of the validation set, and na is the number of subsequences in the wind speed sequence of the validation set.

[0055] Specifically, the first subsequence X of the wind speed sequence is constructed in the L+1th round a1 Specifically:

[0056] X a1 =rand( , 3 )

[0057] Among them, rand() is a random function, a is the number of constructions, Construct the first subsequence of the wind speed sequence for the Lth round.

[0058] Specifically, when A≥B, the sigmoid activation function is used to calculate the remaining subsequence of the wind speed sequence constructed in the L+1th round as follows:

[0059] X an =X a（n-1） q n-1 sigmoid( )

[0060] Where n∈{2,3,…,N}, a is the number of constructions, X an is the nth subsequence in the remaining subsequence, X a（n-1） is the n-1th subsequence in the remaining subsequence, q n-1 is the n-1th subsequence in the ratio wind speed sequence, and rand() is a random function.

[0061] Specifically, when A<B, the tanh activation function is used to calculate the remaining subsequence of the wind speed sequence constructed in the L+1th round as follows:

[0062] X an =X a（n-1） q n-1 tanh( )

[0063] Where n∈{2,3,…,N}, a is the number of constructions, X an is the nth subsequence in the remaining subsequence, X a（n-1） is the n-1th subsequence in the remaining subsequence, q n-1 is the n-1th subsequence in the ratio wind speed sequence, and rand() is a random function.

[0064] Specifically, after collecting the wind speed data, the following steps are also included: deleting and supplementing abnormal data.

[0065] Specifically, the neural network model is LSTM.

[0066] Embodiment 1

[0067] See also Figure 1 , a supercapacitor energy storage frequency modulation optimization method, comprising the following steps:

[0068] Step 1: Collect wind speed data and establish an original wind speed sequence; divide the original wind speed sequence into a prediction set wind speed sequence and a verification set wind speed sequence according to a preset ratio; the number of subsequences in the prediction set wind speed sequence is 10, and the prediction set wind speed sequence is X0(t)=[X 01 ,X 02 ,X 03 ,X 04 ,X 05 ,X 06 ,X 07 ,X 08 ,X 09 ,X 010 ], where X 01 ,X 02 ,X 03 ,X 04 ,X 05 ,X 06 ,X 07 ,X 08 ,X 09 ,X 010 The 1st to 10th subsequences in the prediction set wind speed sequence are input into the trained neural network model for prediction to obtain the predicted value. Based on the predicted value and the actual value in the validation set wind speed sequence, the benchmark prediction error W is calculated.

[0069] Step 2: Calculate the ratio of the latter subsequence to the former subsequence in the wind speed sequence X0(t) of the forecast set to obtain the ratio wind speed sequence q(t);

[0070] q(t)=[q1,q2,q3,q4,q5,q6,q7,q8,q9]=[ , , , , , , , , ], where q1 is the ratio of the second subsequence to the first subsequence in the prediction set wind speed sequence X0(t), q2 is the ratio of the third subsequence to the second subsequence in the prediction set wind speed sequence X0(t), q3 is the ratio of the fourth subsequence to the third subsequence in the prediction set wind speed sequence X0(t), q4 is the ratio of the fifth subsequence to the fourth subsequence in the prediction set wind speed sequence X0(t), q5 is the ratio of the sixth subsequence to the fifth subsequence in the prediction set wind speed sequence X0(t), q6 is the ratio of the seventh subsequence to the sixth subsequence in the prediction set wind speed sequence X0(t). q7 is the ratio of the 8th subsequence to the 7th subsequence in the prediction set wind speed sequence X0(t), q8 is the ratio of the 9th subsequence to the 8th subsequence in the prediction set wind speed sequence X0(t), and q9 is the ratio of the 10th subsequence to the 9th subsequence in the prediction set wind speed sequence X0(t); the number of subsequences A less than or equal to 1 and the number of subsequences B greater than 1 in the statistical ratio wind speed sequence, B=N-1-A=10-A, let L=1, L is the round of constructing the wind speed sequence, use the prediction set wind speed sequence X0(t) as the first round to construct the wind speed sequence, assuming A≥B;

[0071] Step 3: Based on the first round of constructed wind speed sequence and ratio wind speed sequence q(t), the second round of constructed wind speed sequence is obtained by the construction method;

[0072] The construction method specifically includes the following steps: based on the first subsequence of the wind speed sequence constructed in the first round, the first subsequence of the wind speed sequence constructed in the second round is obtained; since A≥B, the sigmoid activation function is used to calculate the remaining subsequence of the wind speed sequence constructed in the second round; the first subsequence and the remaining subsequence of the wind speed sequence constructed in the second round are summarized to obtain the wind speed sequence constructed in the second round; specifically, based on the first subsequence X of the wind speed sequence of the prediction set 01 , get the first subsequence X of the wind speed sequence constructed in the second round 11 =rand( , 3 ), using the sigmoid activation function, calculate the remaining subsequence X of the wind speed sequence constructed in the second round 1n , X 1n =X 1（n-1） q n-1 sigmoid( ), n∈{2,3,…,N}; specifically: X 12 =X 11 q1 sigmoid( );X 13 =X 12 q2 sigmoid( );……;X 1N =X 1（N-1） q N-1 sigmoid( );

[0073] Sum up the first subsequence X 11 and the remaining subsequences to obtain the complete second round of constructed wind speed sequence X1(t);

[0074] Step 4: Calculate the residual wind speed sequence of this round R1(t)=X0(t)-X1(t);

[0075] Step 5: Bring R1(t) into the trained neural network model for prediction to obtain the first round prediction error W1; determine whether W1 is ≥ 0.1 times the benchmark prediction error W. If the judgment result is yes, set L=L+1 and return to step 3;

[0076] Based on the second round, construct the first subsequence X of the wind speed sequence X1(t) 11 , and obtain the first subsequence X of the wind speed sequence constructed in the third round 21 =rand( , 3 ), using the sigmoid activation function, calculate the remaining subsequences of the wind speed sequence constructed in the third round; X 2n =X 2（n-1） q n-1 sigmoid( ), n∈{2,3,…,N}, specifically, X 22 =X 21 q1 sigmoid( );X 23 =X 22 q2 sigmoid( );……;X 2N =X 2（N-1） q N-1 sigmoid( );

[0077] Sum up the first subsequence X 21 and the remaining subsequences to obtain the complete third round of constructed wind speed sequence X2(t);

[0078] Based on the wind speed sequence X1(t) constructed in the second round and the wind speed sequence X2(t) constructed in the third round, the residual wind speed sequence R2(t)=R1(t)-X2(t)=X0(t)-X1(t)-X2(t) is calculated;

[0079] Bring R2(t) into the trained neural network model for prediction to obtain the prediction error W2 for this round; determine whether W2 is ≥ 0.1 times the benchmark prediction error W. Assuming the judgment result is no, bring all the constructed wind speed sequences X1(t), X2(t) and R2(t) obtained by the construction method in the above process into the trained neural network model for prediction to obtain the final wind speed prediction result;

[0080] Step 6. Calculate the predicted power sent by the wind farm to the grid based on the final wind speed prediction result, obtain the actual power sent by the wind farm to the grid, and calculate the difference between the predicted power and the actual power. When the difference is positive, adjust the super-capacity energy storage system connected to the wind farm exit to the charging state; when the difference is 0, the super-capacity energy storage system connected to the wind farm exit maintains the current state unchanged; when the difference is negative, adjust the super-capacity energy storage system connected to the wind farm exit to the discharging state.

[0081] Embodiment 2

[0082] A method for optimizing supercapacitor energy storage frequency modulation comprises the following steps:

[0083] Step 1: Collect wind speed data and establish an original wind speed sequence; divide the original wind speed sequence into a prediction set wind speed sequence and a verification set wind speed sequence according to a preset ratio; the number of subsequences in the prediction set wind speed sequence is 10, and the prediction set wind speed sequence is X0(t)=[X 01 ,X 02 ,X 03 ,X 04 ,X 05 ,X 06 ,X 07 ,X 08 ,X 09 ,X 010 ], where X 01 ,X 02 ,X 03 ,X 04 ,X 05 ,X 06 ,X 07 ,X 08 ,X 09 ,X 010The 1st to 10th subsequences in the prediction set wind speed sequence are input into the trained neural network model for prediction to obtain the predicted value. Based on the predicted value and the actual value in the validation set wind speed sequence, the benchmark prediction error W is calculated.

[0084] Step 2: Calculate the ratio of the latter subsequence to the former subsequence in the wind speed sequence X0(t) of the forecast set to obtain the ratio wind speed sequence q(t);

[0085] q(t)=[q1,q2,q3,q4,q5,q6,q7,q8,q9]=[ , , , , , , , , ], where q1 is the ratio of the second subsequence to the first subsequence in the prediction set wind speed sequence X0(t), q2 is the ratio of the third subsequence to the second subsequence in the prediction set wind speed sequence X0(t), q3 is the ratio of the fourth subsequence to the third subsequence in the prediction set wind speed sequence X0(t), q4 is the ratio of the fifth subsequence to the fourth subsequence in the prediction set wind speed sequence X0(t), q5 is the ratio of the sixth subsequence to the fifth subsequence in the prediction set wind speed sequence X0(t), q6 is the ratio of the seventh subsequence to the sixth subsequence in the prediction set wind speed sequence X0(t). q7 is the ratio of the 8th subsequence to the 7th subsequence in the prediction set wind speed sequence X0(t), q8 is the ratio of the 9th subsequence to the 8th subsequence in the prediction set wind speed sequence X0(t), and q9 is the ratio of the 10th subsequence to the 9th subsequence in the prediction set wind speed sequence X0(t); the number of subsequences A less than or equal to 1 and the number of subsequences B greater than 1 in the statistical ratio wind speed sequence, B=N-1-A=10-A, let L=1, L is the round of constructing the wind speed sequence, use the prediction set wind speed sequence X0(t) as the first round to construct the wind speed sequence, assuming A<B;

[0086] Step 3: Based on the first round of wind speed sequence and ratio wind speed sequence q(t), the second round of wind speed sequence is obtained by construction method; based on the first subsequence X of the wind speed sequence of the prediction set 01 , get the first subsequence X of the wind speed sequence constructed in the second round 11 =rand( , 3 );

[0087] Since A<B, use the tanh activation function to calculate the remaining subsequences of the wind speed sequence constructed in the second round, X 1n =X1（n-1） q n-1 tanh( ), n∈{2,3,…,N}; specifically: X 12 =X 11 q1 tanh( ), X 13 =X 12 q2 tanh( ), ...X 1N =X 1（N-1） q N-1 tanhd( ), sum up the first subsequence X 11 and the remaining subsequences to obtain the complete second round of constructed wind speed sequence X1(t);

[0088] Step 4: Calculate the residual wind speed sequence of this round R1(t)=X0(t)-X1(t);

[0089] Step 5. Bring R1(t) into the trained neural network model for prediction to obtain the first round prediction error W1; determine whether W1 is ≥ 0.1 times the benchmark prediction error W. If the judgment result is no, bring all rounds of constructed wind speed sequence X1(t) obtained in step 3 and the current round residual wind speed sequence R1(t) into the trained neural network model for prediction to obtain the final wind speed prediction result.

[0090] Step 6. Calculate the predicted power sent by the wind farm to the grid based on the final wind speed prediction result, obtain the actual power sent by the wind farm to the grid, and calculate the difference between the predicted power and the actual power. When the difference is positive, adjust the super-capacity energy storage system connected to the wind farm exit to the charging state; when the difference is 0, the super-capacity energy storage system connected to the wind farm exit maintains the current state unchanged; when the difference is negative, adjust the super-capacity energy storage system connected to the wind farm exit to the discharging state.

[0091] In order to further verify the advantages of this method, the wind speed is predicted by the method of this application and the traditional prediction method (i.e. VMD-prediction algorithm). In the experiment of this application, MAPE (mean absolute percentage error, Mean Absolute Percentage Error) is selected as the evaluation standard of each model. The prediction results (i.e. MAPE value) are as follows:

[0092]

[0093] It can be seen that the wind speed prediction using this method is more accurate and reliable.

[0094] Based on the same inventive concept, an embodiment of the present application provides a computer program product, which includes a computer program stored on a computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer device, the computer device executes the steps of the above-mentioned supercapacitor energy storage frequency modulation optimization method.

[0095] Based on the same inventive concept, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps of the optimization method of supercapacitor energy storage frequency modulation are implemented. Specifically, the computer-readable storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. The volatile memory may include RAM (Random Access Memory) and / or cache memory (cache), etc. The non-volatile memory may include ROM (Read-Only Memory), hard disk, flash memory, optical disk, magnetic disk, etc.

[0096] It should be understood by those skilled in the art that the embodiments of the present invention may be provided as methods or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM (Compact Disc Read-Only Memory), optical storage, etc.) containing computer-usable program codes.

[0097] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer device or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0098] These computer program instructions may also be stored in a computer readable memory capable of directing a computer device or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0099] These computer program instructions can also be loaded onto a computer device or other programmable data processing device so that a series of operation steps are executed on the computer device or other programmable device to produce a process implemented by the computer device, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0100] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0101] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for optimizing supercapacitor energy storage frequency modulation, characterized in that: The following steps are involved: Step 1: Collect wind speed data and establish an original wind speed sequence; divide the original wind speed sequence into a prediction set wind speed sequence and a verification set wind speed sequence according to a preset ratio; The wind speed sequence of the forecast set is X0(t)=[X 01 ,X 02 ,X 03 ,...,X 0N ], N is the number of subsequences in the wind speed sequence of the prediction set, X 01 ,X 02 ,X 03 ,...,X 0N are the 1st, 2nd, 3rd, ..., Nth subsequences in the wind speed sequence of the prediction set respectively; the wind speed sequence of the prediction set is input into the trained neural network model for prediction to obtain the predicted value, and the benchmark prediction error W is calculated based on the predicted value and the actual value in the wind speed sequence of the verification set; Step 2: Calculate the ratio of the latter subsequence to the previous subsequence in the wind speed sequence X0(t) of the forecast set, and obtain the ratio wind speed sequence q(t) = [q1, q2, q3, ..., q N-1 ],q1,q2,q3,...,q N-1 are the ratios of the 2nd subsequence to the 1st subsequence, the 3rd subsequence to the 2nd subsequence, …, the Nth subsequence to the N-1th subsequence in the wind speed sequence X0(t) of the prediction set; Count the number of subsequences A less than or equal to 1 and the number of subsequences B greater than 1 in the ratio wind speed sequence, B=N-1-A; Let L=1, L is the round of constructing the wind speed sequence, and use the predicted set wind speed sequence X0(t) as the first round to construct the wind speed sequence; Step 3: Based on the Lth round constructed wind speed sequence and the ratio wind speed sequence q(t), the L+1th round constructed wind speed sequence is obtained by the construction method; The construction method specifically comprises the following steps: based on the first subsequence of the wind speed sequence constructed in the Lth round, the first subsequence of the wind speed sequence constructed in the L+1th round is obtained; when A≥B, the sigmoid activation function is used to calculate the remaining subsequence of the wind speed sequence constructed in the L+1th round; when A<B, the tanh activation function is used to calculate the remaining subsequence of the wind speed sequence constructed in the L+1th round, and the first subsequence and the remaining subsequence of the wind speed sequence constructed in the L+1th round are summarized to obtain the wind speed sequence constructed in the L+1th round; Step 4: Based on the wind speed sequence X0(t) of the prediction set and all rounds of constructed wind speed sequences obtained in step 3, calculate the residual wind speed sequence of this round. The residual wind speed sequence of this round = the first round of constructed wind speed sequence - all rounds of constructed wind speed sequences obtained in step 3; Step 5: Bring the residual wind speed sequence of this round into the trained neural network model for prediction to obtain the prediction error of this round; determine whether the prediction error of this round is ≥ 0.1 times the benchmark prediction error W. If the judgment result is yes, set L = L + 1 and return to step 3; if the judgment result is no, bring all rounds of constructed wind speed sequences obtained in step 3 and the residual wind speed sequence of this round into the trained neural network model for prediction to obtain the final wind speed prediction result; Step 6. Calculate the predicted power sent by the wind farm to the grid based on the final wind speed prediction result, obtain the actual power sent by the wind farm to the grid, and calculate the difference between the predicted power and the actual power. When the difference is positive, adjust the super-capacity energy storage system connected to the wind farm exit to the charging state; when the difference is 0, the super-capacity energy storage system connected to the wind farm exit maintains the current state unchanged; when the difference is negative, adjust the super-capacity energy storage system connected to the wind farm exit to the discharging state.

2. The optimization method of supercapacitor energy storage frequency modulation according to claim 1, characterized in that: The preset ratio is 9:

1.

3. The optimization method of supercapacitor energy storage frequency modulation according to claim 1, characterized in that: The benchmark prediction error W is specifically: Among them, y i is the ith predicted value, i is the i-th actual value in the wind speed sequence of the validation set, and na is the number of subsequences in the wind speed sequence of the validation set.

4. The method for optimizing supercapacitor energy storage frequency modulation according to claim 1, characterized in that: In the L+1th round, the first subsequence X of the wind speed sequence is constructed a1 Specifically: X a1 =row( ,3 ) Among them, rand() is a random function, a is the number of constructions, Construct the first subsequence of the wind speed sequence for the Lth round.

5. The method for optimizing supercapacitor energy storage frequency modulation according to claim 1, characterized in that: When A≥B, the sigmoid activation function is used to calculate the remaining subsequence of the wind speed sequence constructed in the L+1th round: X an =X a（n-1） q n-1 sigmoid( ) Where n∈{2,3,…,N}, a is the number of constructions, X an is the nth subsequence in the remaining subsequence, X a（n-1） is the n-1th subsequence in the remaining subsequence, q n-1 is the n-1th subsequence in the ratio wind speed sequence, and rand() is a random function.

6. The method for optimizing supercapacitor energy storage frequency modulation according to claim 1, characterized in that: When A<B, the tanh activation function is used to calculate the remaining subsequence of the wind speed sequence constructed in the L+1th round: X an =X a（n-1） q n-1 fishy ) Where n∈{2,3,…,N}, a is the number of constructions, X an is the nth subsequence in the remaining subsequence, X a（n-1） is the n-1th subsequence in the remaining subsequence, q n-1 is the n-1th subsequence in the ratio wind speed sequence, and rand() is a random function.

7. The method for optimizing supercapacitor energy storage frequency modulation according to claim 1, characterized in that: After collecting wind speed data, the following steps are also included: deleting and supplementing abnormal data.

8. The method for optimizing supercapacitor energy storage frequency modulation according to claim 1, characterized in that: The neural network model is LSTM.

9. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images