Method for reducing offshore wind power generation noise based on artificial intelligence waveform structure characteristics

Abstract

The invention discloses a method for reducing offshore wind power generation noise based on artificial intelligence waveform structure features, and relates to the technical field of noise reduction, and the method comprises the steps: collecting, processing, analyzing and processing offshore wind power generation noise data; and establishing a noise reduction prediction model. According to the method, a traditional physical and mechanical noise reduction method is changed into a mathematical and physical structure noise reduction method for the first time, the anti-noise waveform structure characteristic active noise reduction theory is introduced into offshore wind power generation noise reduction, the basic principle and algorithm of the method are provided, and the noise reduction credibility and operability are improved from the noise production source. On the basis of the technology for reducing the noise of offshore wind power generation, a mathematical prediction model for calculating the noise through an exponential function is realized, a novel technology for reducing the noise of a quantitative structure in which the longitudinal direction and the transverse direction can be quantitatively calibrated at the same time and the graph and the table can be compared is displayed, and the technical means for reducing the noise of offshore wind power generation is broadened.

Description

Technical Field

[0001] The present invention relates to the technical field of noise reduction, and in particular to a method for reducing offshore wind power generation noise based on artificial intelligence waveform structure characteristics. Background Art

[0002] Although wind power generation has many advantages and wind power does bring a lot of convenience to mankind, the vibration and noise pollution of wind turbines will produce corresponding low-frequency vibrations and harmful noise, which will have impacts and harms on human health and organisms near wind farms, such as: noise, electromagnetic radiation, ecological vegetation, birds, and marine life. Especially noise, the construction of wind power equipment faces difficult-to-solve environmental problems, which seriously restricts the development of wind power.

[0003] Therefore, how to effectively solve the noise problem and reduce noise is a major issue in improving environmental quality. Summary of the invention

[0004] In order to solve the above technical problems, the present invention discloses a method for reducing offshore wind power generation noise based on artificial intelligence waveform structure characteristics. This method transforms the traditional physical mechanical noise reduction method into a mathematical and physical structural noise reduction method, and introduces the anti-noise waveform structure characteristic active noise reduction theory into offshore wind power generation noise reduction, thereby improving the credibility and operability of noise reduction from the source of production noise.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] The method for reducing offshore wind power generation noise based on artificial intelligence waveform structure characteristics comprises the following steps:

[0007] Collect, process and analyze offshore wind power generation noise data and process it;

[0008] Build a noise reduction prediction model.

[0009] Among them, noise data include noise amplitude data, wind speed, wind direction, temperature, air pressure, blade pitch, heading, and wind tower parameters; artificial intelligence deep learning such as convolutional neural network algorithms are used to achieve high-precision recognition and analysis of waveform structure characteristics such as noise data amplitude, frequency, phase, inflection point, slope, and angle in the signal-to-noise ratio.

[0010] Optionally, the step of establishing a noise reduction prediction model specifically includes:

[0011] Select any sub-seismic amplitude data column, and the grey model is represented by GM(1,n), where 1 represents the first order and n represents the dimension of the variable. Using the GM(1,1) model, the amplitude data column of the original seismic data is:

[0012]

[0013] In the formula, N represents the number of variables;

[0014] Choose any subsequence and record it as:

[0015]

[0016] The data prediction model series is generated by accumulation, and the formula is:

[0017]

[0018] Where, X (1) is a sequence generated by accumulation;

[0019] Construct the accumulation matrix B and constant vector Y N , use the obtained cumulative sequence to establish the model GM(1,1):

[0020]

[0021] In the formula, a is the development coefficient, which reflects the development trend of the original sequence and the cumulative sequence; u is the external input or action, X (1) is an exponential function series;

[0022] Solving the Grey Parameters Using the Least Squares Method Use the least squares method to find a and u, record but:

[0023]

[0024] In the formula,

[0025]

[0026] Gray parameters Substituting into formula (4), we get a and u;

[0027] X (1) Take the derivative, substitute a and u into the time function (4), and take the derivative to obtain:

[0028]

[0029] calculate and Difference and relative error:

[0030]

[0031] In the formula, Refers to calculation and The difference; q (t) refer to and The relative error of the ratio;

[0032] The gray outlier value of the amplitude data is calculated using the model GM(1,1). The model value sequence is predicted using the model, and the predicted value sequence of the original sequence is obtained by cumulative subtraction:

[0033]

[0034] The beneficial effects of the present invention are:

[0035] The advantages of the method of the present invention are: ① in terms of time, all-period structural noise reduction; ② in terms of space, all-round structural noise reduction; ③ in terms of monitoring, dynamic structural noise reduction; ④ in terms of standards, quantitative structural noise reduction is achieved; ⑤ in terms of application, a large amount of manpower, material resources and financial resources are reduced. Specifically:

[0036] First, in terms of theoretical method research: for the first time in the world, it was proposed to transform the traditional physical mechanical noise reduction method into a mathematical and physical structural noise reduction method, and to introduce the active noise reduction theory of anti-noise waveform structure characteristics into offshore wind power generation noise reduction. The basic principles and algorithms of this method were proposed, which improved the credibility and operability of noise reduction from the source of production noise.

[0037] Secondly, in terms of practical application of technology: in the technology of reducing noise from offshore wind power generation, a mathematical prediction model for calculating noise using an exponential function has been realized, demonstrating a new technology for quantitative structural noise reduction that can simultaneously quantify and calibrate longitudinally and transversely and compare graphs and tables with each other, thus broadening the technical means for reducing noise from offshore wind power generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 To study the flow chart of reducing offshore wind power noise based on the noise waveform structure characteristics;

[0039] Figure 2 Schematic diagram of artificial intelligence grey processing of noise vibration wave data volume for realizing the method of reducing offshore wind power generation noise;

[0040] Figure 3 It is the positive and negative phase amplitude diagram of the original data column after the noise amplitude data volume is processed, the horizontal axis is Time, and the vertical axis is CDP (origin);

[0041] Figure 4 It is the positive and negative phase envelope diagram of the amplitude of the original data column after the noise amplitude data volume is processed. The horizontal axis is Time and the vertical axis is CDP(-), CDP(+);

[0042] Figure 5 It is the positive phase diagram of the amplitude of the original data column after the noise amplitude data volume is processed. The horizontal axis is Time and the vertical axis is CDP(-)+CDP(+);

[0043] Figure 6 The data prediction model diagram is generated by accumulating the noise amplitude data after processing, with the horizontal axis CD and the vertical axis Time;

[0044] Figure 7 The results of the method for reducing offshore wind power generation noise based on waveform structure characteristics. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below 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, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the invention claimed for protection, but merely represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0046] Method for achieving structural noise reduction: If we receive a group of sound waves from the outside world, including noise waves and normal sound waves, normal sound waves in this article refer to normal vibration sound waves that people can tolerate; however, based on physical health, it is necessary to reduce or remove the sound wave bands that have an impact on us, and retain the sound wave bands that are harmless to the human body. To eliminate the noise waves that have an impact on us, it is to fit a reverse noise wave through an equation, and the noise wave and the reverse wave cancel each other out, leaving a normal vibration wave. When people's ears receive external noise (including normal vibration waves that people's brains can tolerate), and then through the artificial intelligence gray prediction method of the present invention, these noises are converted into waveform structure reverse vibration waves according to the waveform structure characteristics, and the ears receive this reverse vibration wave and the normal vibration wave at the same time. In this way, the reverse vibration wave cancels out the external noise, and what remains will be a normal vibration wave that people's brains can tolerate. This is the principle of structural noise reduction of the present invention.

[0047] Methods for reducing offshore wind power generation noise based on artificial intelligence waveform structure characteristics, such as Figure 1 As shown, the following steps are included:

[0048] (I) Collect, process and analyze offshore wind power noise data and process:

[0049] Noise data include noise amplitude data, wind speed, wind direction, temperature, air pressure, propeller pitch, heading, and wind tower parameters.

[0050] By utilizing grey system theory and combining it with artificial intelligence deep learning such as convolutional neural network algorithms, we can achieve high-precision recognition and analysis of waveform structural features such as noise data amplitude, frequency, phase, inflection point, slope, angle, etc. in the signal-to-noise ratio, improve the resolution of noise data analysis, ensure that different noise changes can be accurately identified, and realize automated feature extraction, reducing the need for manual operation.

[0051] Basic principles for parameter selection: ① The degree of smoothness of the abnormal characteristics of the vibration noise waveform structure. Try to avoid taking high and low values ​​and peak and valley maximum and minimum values; ② The relationship between the abnormal value of the vibration noise waveform structure and the numerical value of the surrounding structure. In general, the abnormal value of the noise waveform structure has a gradual relationship with the adjacent abnormal value; ③ Comparison of the degree of correlation between the abnormal value of the vibration noise waveform structure and the results of different people in different time periods. ④ Sampling at equal intervals, as shown in the second column of Table 1 below, take a sample every 8 minutes, so that the sampling is representative.

[0052] The specific steps include:

[0053] (1) Collect raw data, such as noise amplitude A, frequency f, and velocity V;

[0054] (2) Prepare a standard Excel table, such as Table 1, with the horizontal direction representing the different data contents collected, such as amplitude value (ORIGIN), peak value (CDP+), trough value (CDP-), etc., and the vertical direction representing the data series (minutes);

[0055] (3) Sampling at equal intervals, as shown in the first column of Table 1, when encountering a maximum value (peak) or a minimum value (trough), take the average of the three adjacent points;

[0056] (4) Perform an accumulation to generate a data column, such as GD in the sixth column of Table 1;

[0057] (5) Output parameter data columns of different noise sources.

[0058] In addition to the external changes in amplitude, phase, frequency, speed, etc., noise must also have changes in internal waveform structural characteristics, which are manifested externally as changes in amplitude, phase, frequency, speed, etc. The microscopic changes in internal waveform structural characteristics such as inflection points, slopes, angles, etc. cannot be recognized by the naked eye, so they must be identified and described using internal waveform structural characteristics. This has been proven feasible and recognizable in theory, production practice, and vibration wave physical models through a large number of practices. See below for details. Figure 2 As shown, this method transforms the discrete data points of the noise data body into a continuous dynamic prediction change process through mathematical topology. Figure 2 (a) is the extracted noise amplitude data series, i.e. the original noise discrete number body; Figure 2 (c) The middle column shows the positive and negative phase amplitude diagrams of the noise; Figure 2 (c) The blue, pink, and red lines in the rightmost two columns are the envelope diagrams of the positive and negative channels of the original noise amplitude; Figure 2 (b) is a schematic diagram of artificial intelligence processing of noise vibration wave data after extracting the noise amplitude data column.

[0059] The processing steps are summarized as follows: ① input the original noise vibration wave data body to be processed; ② generate the amplitude and phase data body of the original noise vibration wave data series; ③ accumulate and generate the data prediction model series; ④ output the processing results.

[0060] (1) Input the original noise vibration wave data to be processed in Table 1

[0061] Table 1 is the original data table of the noise amplitude data column, in which the first column is the serial number; the second column is the time series, the unit is minutes; the third column is the original noise amplitude data, expressed as positive and negative alternating amplitude values; the fourth column is the positive amplitude value column and the negative amplitude value column after manual editing; the fifth column is the all positive amplitude value column after manual editing; the sixth column is the accumulated data column after noise amplitude data processing, the purpose of which is to provide it for prediction of the mathematical model of structural noise reduction.

[0062] Table 1 Noise amplitude data raw data table

[0063]

[0064] (2) Generate the amplitude and phase data of the original noise vibration wave data series

[0065] Figure 3 The positive and negative phase amplitude diagram of the original data column after the noise amplitude data body is processed, corresponding to the third column of Table 1, which is the original noise amplitude data, expressed as positive and negative alternating amplitude values. Figure 4 It is the positive and negative phase envelope diagram of the amplitude of the original data column after the noise amplitude data volume is processed, corresponding to the fourth column of Table 1, which is the positive amplitude value column and the negative amplitude value column after manual editing. Figure 5 This is the positive phase diagram of the amplitude of the original data column after the noise amplitude data volume is processed, corresponding to the fifth column of Table 1, that is, the column of all positive amplitude values ​​after manual editing.

[0066] (3) Accumulate and generate data prediction model series

[0067] Figure 6 The data prediction model diagram is generated by accumulating the noise amplitude data after processing, corresponding to the sixth column of Table 1, which is the data column generated by accumulating the noise amplitude data after processing, and its purpose is to provide it for prediction of the mathematical model of structural noise reduction.

[0068] (II) Establishing a noise reduction prediction model

[0069] It is mainly accomplished by citing the relevant prediction models of grey system theory, that is, each seismic trace of the seismic data body is transformed into an exponential function curve through grey mathematical topology, and then the abnormal value segment of the amplitude structure characteristic of the seismic data body is calculated, and then the error value analysis is performed. Finally, the numerical range of the effectiveness of the structural noise reduction application is pointed out.

[0070] Specifically include:

[0071] Select any sub-seismic amplitude data column, and the grey model is represented by GM(1,n), where 1 represents the first order and n represents the dimension of the variable. Using the GM(1,1) model, the amplitude data column of the original seismic data is:

[0072]

[0073] In the formula, N represents the number of variables;

[0074] Choose any subsequence and record it as:

[0075]

[0076] The data prediction model series is generated by accumulation, and the formula is:

[0077]

[0078] Where, X (1) is a sequence generated by accumulation;

[0079] Construct the accumulation matrix B and constant vector Y N , use the obtained cumulative sequence to establish the model GM(1,1):

[0080]

[0081] In the formula, a is the development coefficient, which reflects the development trend of the original sequence and the cumulative sequence; u is the external input or action, X (1) is an exponential function series; in general, the system action should be exogenous or predetermined, while GM(1,1) is a single-column modeling, which only uses the system's behavior sequence without an external action sequence. The gray action in GM(1,1) is data mined from the background value, which reflects the relationship between data changes, and its exact connotation is gray. The gray action is a concrete manifestation of the extension of connotation. Its existence is the watershed that distinguishes gray modeling from general modeling, and it is also an important symbol to distinguish the gray system view from the gray box view.

[0082] Solving the Grey Parameters Using the Least Squares Method Use the least squares method to find a and u, record but:

[0083]

[0084] In the formula,

[0085]

[0086] Gray parameters Substituting into formula (4), we get a and u;

[0087] X (1) Take the derivative, substitute a and u into the time function (4), and take the derivative to obtain:

[0088]

[0089] calculate and Difference and relative error:

[0090]

[0091] In the formula, Refers to calculation and The difference; q (t) refer to and The relative error of the ratio;

[0092] The gray outlier value of the amplitude data is calculated using the model GM(1,1). The model value sequence is predicted using the model, and the predicted value sequence of the original sequence is obtained by cumulative subtraction:

[0093]

[0094] Figure 7 This is the result of the method of reducing offshore wind power generation noise by waveform structural characteristics. From the superposition of the original noise wave spectrum and the noise-removed wave spectrum through the equation, it can be seen that the two noise wave curves are in good agreement, and the original noise spectrum and the noise-removed spectrum through the equation basically overlap, indicating that the "Structural Noise Reduction Method for Reducing Offshore Wind Power Generation Noise Based on Artificial Intelligence Waveform Structural Characteristics" is very effective and usable. Table 2 is a data table of the results of the study on reducing offshore wind power generation noise by waveform structural characteristics. As can be seen from Table 2, the relative error value is between 2.05-5.77, indicating that the noise removal effect is more than 90%. Table 2 also confirms that Figure 7 This also shows that this method is very effective and usable for denoising.

[0095] Table 2 Data table of waveform structure characteristics research to reduce noise of offshore wind power generation

[0096]

[0097] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by technicians in this technical field within the essential scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A method for reducing offshore wind power generation noise based on artificial intelligence waveform structure characteristics, characterized in that: The steps include: Collect, process and analyze offshore wind power generation noise data and process it; Build a noise reduction prediction model.

2. The method for reducing offshore wind power generation noise based on artificial intelligence waveform structure characteristics according to claim 1 is characterized in that: The steps to establish a noise reduction prediction model include: Select any sub-seismic amplitude data column, and the grey model is represented by GM(1,n), where 1 represents the first order and n represents the dimension of the variable. Using the GM(1,1) model, the amplitude data column of the original seismic data is: In the formula, N represents the number of variables; Choose any subsequence and record it as: The data prediction model series is generated by accumulation, and the formula is: Where, X (1) is a sequence generated by accumulation; Construct the accumulation matrix B and constant vector Y N , use the obtained cumulative sequence to establish the model GM(1,1): In the formula, a is the development coefficient, which reflects the development trend of the original sequence and the cumulative sequence; u is the external input or action, X (1) is an exponential function series; Solving the Grey Parameters Using the Least Squares Method Use the least squares method to find a and u, record but: In the formula, Gray parameters Substituting into formula (4), we get a and u; X (1) Take the derivative, substitute a and u into the time function (4), and take the derivative to obtain: calculate and Difference and relative error: In the formula, Refers to calculation and The difference; q (t) refer to and The relative error of the ratio; The gray outlier value of the amplitude data is calculated using the model GM(1,1). The model value sequence is predicted using the model, and the predicted value sequence of the original sequence is obtained by cumulative subtraction:

Images