A pumped storage power station active noise reduction safety protection device and a noise reduction method

Abstract

The present application relates to the technical field of safety protection tools, and is especially a pumped storage power station active noise reduction safety protection device and a noise reduction method, comprising a wearable shell and an active noise reduction mechanism, the inner wall of the wearable shell is provided with sound-absorbing cotton, and an inner lining frame is fixedly arranged on the inner side of the sound-absorbing cotton, the active noise reduction mechanism comprises a sound pickup end, a loudspeaker end and an active noise reduction unit, and the inner walls on the left and right sides of the wearable shell are both provided with jacks. The present application combines passive noise reduction with active noise reduction, first uses the selective noise reduction characteristics of sound-absorbing cotton to reduce high-frequency noise, uses polyester fiber sound-absorbing cotton for sound insulation in the wearable shell to reduce high-frequency noise, and then embeds a low-pass filter in the active noise reduction protection device to filter high-frequency noise again, ensures that the ANC master control module focuses more on the processing of low-frequency noise, and guarantees the low-frequency noise reduction effect.

Description

Technical Field

[0001] This invention relates to the field of safety protection tools, specifically to an active noise reduction safety protection device and noise reduction method for pumped storage power stations. Background Technology

[0002] With the continuous development and construction of pumped storage power stations, hydro-turbine generator units are now trending towards larger sizes. During the operation of the power station, noise pollution from the units, caused by hydraulic, electrical, and mechanical factors, can lead to occupational health hazards. Prolonged exposure to high-noise environments can cause hearing loss, tinnitus, and neurasthenia among power station workers.

[0003] Traditional solutions often involve using in-ear earplugs for noise reduction. However, wearing in-ear protective devices for extended periods can cause discomfort for the wearer. Furthermore, simple soundproofing materials can only reduce some high-frequency noise, and their soundproofing effect is not ideal for the low-frequency noise generated by hydroelectric generator sets.

[0004] Among existing noise reduction technologies, there are applications of active noise reduction by superimposing "anti-phase sound waves." The principle is to use a microphone to collect ambient noise in real time, and then use a circuit to generate an inverse sound wave that cancels out the ambient noise, thus achieving active noise reduction. This noise reduction method has a clear principle and works well for relatively pure low-frequency noise. However, for unpredictable noise, it requires real-time noise information collection via microphone, sound pressure analysis, and calculation of amplitude and phase information to generate the inverse sound wave, necessitating real-time processing. This method places extremely high demands on the precision and performance of the noise reduction system; for 100Hz low-frequency sound waves, a 0.005-second delay results in a 90° phase angle change.

[0005] Pumped storage power station generator units are installed in large underground caverns in mountains. The caverns are enclosed and spacious, and the inner surface of the caverns is made of concrete and cement mortar with low sound absorption coefficients. The noise of the generator units is reflected multiple times in the caverns, and the sound energy decays slowly. The above-mentioned active noise reduction technologies cannot meet the requirements and may even be counterproductive. Therefore, in order to address the above problems, an active noise reduction safety protection device and noise reduction method for pumped storage power stations are proposed. Summary of the Invention

[0006] The purpose of this invention is to provide an active noise reduction safety protection device and method for pumped storage power stations, which can effectively protect the human head while actively reducing external noise, thereby solving the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] An active noise reduction safety protection device for pumped storage power stations includes:

[0009] The wearable shell has sound-absorbing cotton on its inner wall and an inner lining frame fixed inside the sound-absorbing cotton.

[0010] The active noise cancellation mechanism includes a pickup end, a speaker end, and an active noise cancellation unit. The inner walls of both sides of the wearable outer shell have insertion holes, and the inner walls of both sides of the inner liner frame have receiving slots. A slot is located on the rear side of the inner liner frame. The pickup end includes a primary pickup A, a secondary pickup A, a primary pickup B, and a secondary pickup B. The speaker end includes a loudspeaker A and a loudspeaker B. The active noise cancellation unit includes an active noise cancellation controller A and an active noise cancellation controller B housed in the slots. Primary pickup A and primary pickup B are respectively installed in the insertion holes on the left and right sides of the wearable outer shell. Secondary pickup A, loudspeaker A, and secondary pickup B, and loudspeaker B are respectively installed in the receiving slots on the left and right sides of the inner liner frame. Primary pickup A, primary pickup B, and loudspeaker B are respectively connected to active noise cancellation controller A, and secondary pickup A, secondary pickup B, and loudspeaker B are connected to active noise cancellation controller B.

[0011] As a preferred embodiment, both active noise cancellation controller A and active noise cancellation controller B include a primary pickup unit, a secondary pickup unit, an acoustic driver unit, and an ANC main control module. The primary pickup unit of active noise cancellation controller A is connected to primary pickup A, and the secondary pickup unit of active noise cancellation controller A is connected to secondary pickup A. The primary pickup unit of active noise cancellation controller B is connected to primary pickup B, and the secondary pickup unit of active noise cancellation controller B is connected to secondary pickup B. The acoustic driver units of active noise cancellation controller A and active noise cancellation controller B are connected to speaker A and speaker B, respectively. The primary pickup unit, secondary pickup unit, and acoustic driver unit in active noise cancellation controller A and active noise cancellation controller B are connected to the ANC main control module, respectively.

[0012] Active noise reduction methods include the following steps:

[0013] S1. The primary pickup unit picks up ambient noise filtered by sound-absorbing cotton, and then connects it to the ANC main control module after signal amplification, filtering and analog-to-digital conversion.

[0014] S2 and ANC main control modules statistically analyze the amplitude of noise at each frequency point of a certain duration (duration coefficient T), perform modeling and training, and finally generate an inverted audio signal, which is then output to the acoustic drive unit.

[0015] S3, the acoustic drive unit converts the inverted audio signal input from the ANC main control module into a digital-to-analog converter, amplifies the signal, and outputs it to the speaker;

[0016] S4, the secondary pickup unit picks up the residual noise after phase inversion suppression, and then connects it to the ANC main control module after signal amplification, low-pass filtering and analog-to-digital conversion;

[0017] The S5 and ANC main control modules determine whether the residual noise amplitude is higher than the threshold. If it is higher than the threshold, the duration coefficient T is adjusted and the model is re-modeled. Otherwise, the existing inverted audio signal is directly output to continuously suppress noise.

[0018] As a preferred approach, in step S2, modeling and training are performed using dynamic programming based on reinforcement learning to construct the noise sequence. Based on the characteristics of the generator set's "steady-state noise," the entire optimization problem is decomposed into multiple sub-optimization problems. The solutions to these sub-optimization problems are stored and reused, satisfying the dynamic programming conditions. A complete reinforcement learning process is represented using tuples (S, A, P, r, y), where S is the state set, A is the action set, P is the transition probability (corresponding to the noise model), r is the reward function, and y is the discount factor used to calculate the cumulative reward R. The cumulative reward formula is... When 0≤y≤1 and T is a finite value, it is the length coefficient T.

[0019] As a preferred solution, in step S3, specifically, the ANC main control module inverts the noise sequence to obtain inverted audio data, and then outputs the inverted audio data to the speaker via the acoustic drive unit to suppress the original noise.

[0020] As can be seen from the technical solution provided by the present invention above, the beneficial effects of the active noise reduction safety protection device and noise reduction method for pumped storage power stations provided by the present invention are:

[0021] 1. This invention adopts a combination of passive and active noise reduction. First, it utilizes the selective noise reduction properties of sound-absorbing cotton to reduce high-frequency noise. Polyester fiber sound-absorbing cotton is used for sound insulation inside the wearable shell to reduce high-frequency noise. Second, a low-pass filter is embedded in the active noise reduction protection device to perform secondary filtering of high-frequency noise, ensuring that the ANC main control module focuses more on processing low-frequency noise and guaranteeing the low-frequency noise reduction effect.

[0022] 2. This invention is based on the idea of ​​reinforcement learning algorithm, which models the low-frequency noise of the previous time step to suppress the low-frequency noise of the next time step in the opposite direction. Through a secondary noise feedback mechanism, it determines whether to remodel or use the existing model to continue to suppress the noise of the next time step. This algorithm greatly improves the processing performance and ensures the low-frequency noise reduction effect.

[0023] 3. This invention integrates an active noise reduction and protection device into a wearable device to replace existing safety helmets, providing safety protection for workers and making it convenient to use. Attached Figure Description

[0024] Figure 1 : Schematic diagram of active noise cancellation principle;

[0025] Figure 2 External structure diagram of active noise reduction safety protection device;

[0026] Figure 3 Internal structure diagram of active noise cancellation safety protection device;

[0027] Figure 4 Passive noise reduction effect diagram;

[0028] Figure 5 Overall noise reduction effect diagram;

[0029] Figure 6 Schematic diagram of an active noise cancellation controller;

[0030] Figure 7 ANC main control module workflow diagram.

[0031] In the diagram: 1. Wearable shell; 11. Sound-absorbing cotton; 12. Inner frame; 2. Primary pickup A; 21. Secondary pickup A; 22. Primary pickup B; 23. Secondary pickup B; 3. Speaker A; 31. Speaker B; 4. Active noise cancellation controller A; 41. Active noise cancellation controller B; 42. Power manager. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0033] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific embodiments.

[0034] like Figure 1-7 As shown, this embodiment of the invention provides an active noise reduction safety protection device and noise reduction method for pumped storage power stations, including:

[0035] Wearable shell 1, with sound-absorbing cotton 11 provided on the inner wall of wearable shell 1, and an inner liner frame 12 fixedly provided inside the sound-absorbing cotton 11;

[0036] The active noise cancellation mechanism includes a pickup end, a speaker end, and an active noise cancellation unit. The inner walls of the wearable outer shell 1 on both sides have insertion holes, and the inner walls of the inner frame 12 on both sides have receiving grooves. A slot is located on the rear side of the inner frame 12. The pickup end includes a primary pickup A2, a secondary pickup A21, a primary pickup B22, and a secondary pickup B23. The speaker end includes a loudspeaker A3 and a loudspeaker B31. The active noise cancellation unit includes an active noise cancellation controller A4 located within the slot and an active noise cancellation unit. The noise-canceling controller B41, the primary microphones A2 and B22 are respectively installed in the sockets on the left and right sides of the wearable shell 1. The secondary microphones A21, A3, B23, and B31 are respectively installed in the receiving slots on the left and right sides of the inner frame 12. The primary microphones A2, B22, and B31 are respectively connected to the active noise-canceling controller A4, and the secondary microphones A21, B23, and B31 are connected to the active noise-canceling controller B41.

[0037] As a preferred embodiment, both active noise cancellation controller A4 and active noise cancellation controller B41 include a primary pickup unit, a secondary pickup unit, an acoustic driver unit, and an ANC main control module. The primary pickup unit of active noise cancellation controller A4 is connected to the primary pickup A2, and the secondary pickup unit of active noise cancellation controller A4 is connected to the secondary pickup A21. The primary pickup unit of active noise cancellation controller B41 is connected to the primary pickup B22, and the secondary pickup unit of active noise cancellation controller B41 is connected to the secondary pickup B23. The acoustic driver units of active noise cancellation controller A4 and active noise cancellation controller B41 are connected to speaker A3 and speaker B31, respectively. The primary pickup unit, secondary pickup unit, and acoustic driver unit in active noise cancellation controller A4 and active noise cancellation controller B41 are connected to the ANC main control module, respectively.

[0038] Active noise reduction methods include the following steps:

[0039] S1. The primary pickup unit picks up ambient noise filtered by sound-absorbing cotton, and then connects it to the ANC main control module after signal amplification, filtering and analog-to-digital conversion.

[0040] S2 and ANC main control modules statistically analyze the amplitude of noise at each frequency point of a certain duration (duration coefficient T), perform modeling and training, and finally generate an inverted audio signal, which is then output to the acoustic drive unit.

[0041] S3, the acoustic drive unit converts the inverted audio signal input from the ANC main control module into a digital-to-analog converter, amplifies the signal, and outputs it to the speaker;

[0042] S4, the secondary pickup unit picks up the residual noise after phase inversion suppression, and then connects it to the ANC main control module after signal amplification, low-pass filtering and analog-to-digital conversion;

[0043] The S5 and ANC main control modules determine whether the residual noise amplitude is higher than the threshold. If it is higher than the threshold, the duration coefficient T is adjusted and the model is re-modeled. Otherwise, the existing inverted audio signal is directly output to continuously suppress noise.

[0044] As a preferred approach, in step S2, modeling and training are performed using dynamic programming based on reinforcement learning to construct the noise sequence. Based on the characteristics of the generator set's "steady-state noise," the entire optimization problem is decomposed into multiple sub-optimization problems. The solutions to these sub-optimization problems are stored and reused, satisfying the dynamic programming conditions. A complete reinforcement learning process is represented using tuples (S, A, P, r, y), where S is the state set, A is the action set, P is the transition probability (corresponding to the noise model), r is the reward function, and y is the discount factor used to calculate the cumulative reward R. The cumulative reward formula is... When 0≤y≤1 and T is a finite value, it is the length coefficient T.

[0045] As a preferred solution, in step S3, specifically, the ANC main control module inverts the noise sequence to obtain inverted audio data, and then outputs the inverted audio data to the speaker via the acoustic drive unit to suppress the original noise.

[0046] The embodiments of the present invention will now be described in further detail with reference to the accompanying drawings:

[0047] Please see Figure 1-7 The device includes a wearable outer shell 1 and an active noise cancellation mechanism. The inner wall of the wearable outer shell 1 is provided with sound-absorbing cotton 11, and an inner liner frame 12 is fixedly installed inside the sound-absorbing cotton 11. The active noise cancellation mechanism includes a pickup end, a speaker end, and an active noise cancellation unit. Insert holes are provided on the inner walls of both the left and right sides of the wearable outer shell 1, and receiving grooves are provided on the inner walls of both the left and right sides of the inner liner frame 12. A slot is provided on the rear side of the inner liner frame 12. The pickup end includes a primary pickup A2, a secondary pickup A21, a primary pickup B22, and a secondary pickup B23. The speaker end includes a loudspeaker A3 and a loudspeaker B31. The active noise cancellation mechanism... The noise reduction unit includes an active noise reduction controller A4 and an active noise reduction controller B41 installed in the slot. The primary microphones A2 and B22 are installed in the sockets on the left and right sides of the wearable shell 1, respectively. The secondary microphones A21, speaker A3, secondary microphones B23, and speaker B31 are installed in the receiving slots on the left and right sides of the inner frame 12, respectively. The primary microphones A2, B22, and speaker B31 are connected to the active noise reduction controller A4, and the secondary microphones A21, B23, and speaker B31 are connected to the active noise reduction controller B41.

[0048] Specifically, the sound-absorbing cotton 11 is made of polyester fiber, achieving passive noise reduction. As a filling material, the sound-absorbing cotton cushions and protects the wearer's head when the wearable shell 1 is subjected to external impact. Furthermore, the sound-absorbing cotton 11 has selective noise reduction characteristics for high-frequency noise. By using polyester fiber sound-absorbing cotton for sound insulation inside the wearable shell 1, high-frequency noise is significantly reduced, resulting in a passive noise reduction effect. Figure 4 As shown.

[0049] In this embodiment, both the active noise cancellation controller A4 and the active noise cancellation controller B41 include a primary pickup unit, a secondary pickup unit, an acoustic drive unit, and an ANC main control module. The primary pickup unit of the active noise cancellation controller A4 is connected to the primary pickup A2, and the secondary pickup unit of the active noise cancellation controller A4 is connected to the secondary pickup A21. The primary pickup unit of the active noise cancellation controller B41 is connected to the primary pickup B22, and the secondary pickup unit of the active noise cancellation controller B41 is connected to the secondary pickup B23. The acoustic drive units of the active noise cancellation controllers A4 and B41 are respectively connected to the speaker A3 and the speaker B31. The primary pickup unit, the secondary pickup unit, and the acoustic drive unit in the active noise cancellation controllers A4 and B41 are respectively connected to the ANC main control module.

[0050] Furthermore, after the active noise cancellation controller is activated, the primary pickup unit picks up the ambient noise filtered by sound-absorbing cotton through a primary microphone. This noise is mostly low-frequency noise generated by the generator set, and it is an analog signal. The primary pickup unit amplifies the power of the picked-up analog noise signal and uses a low-pass filter to filter out the high-frequency components in the signal. Then, the low-frequency signal after high-frequency filtering is converted from analog to digital and input to the ANC main control module. The noise data generated in this process is "steady-state noise" data, which has a strong regularity in loudness and frequency distribution. The low-pass filter is used to ensure that the ANC main control module focuses more on processing low-frequency noise, ensuring the low-frequency noise reduction effect.

[0051] Furthermore, the ANC main control module receives low-frequency digital signals input from the first-level pickup unit. Based on the reinforcement learning algorithm, it statistically analyzes the energy of noise corresponding to each frequency point in the noise audio for a certain duration (duration coefficient T). This duration is set from 0.5 seconds to 3 seconds and can be dynamically adjusted in subsequent iterative training in combination with sound field characteristics and residual noise. The ANC main control module models and trains the acquired audio signal according to acoustic characteristics, and finally generates an inverted audio signal, which is output to the acoustic drive unit.

[0052] Furthermore, the acoustic drive unit performs digital-to-analog conversion and signal amplification on the inverted audio signal input from the ANC main control module before outputting it to the speaker. This inverted audio signal is mainly used to suppress low-frequency noise in the original environment; this process is called active noise cancellation, and the principle of active noise cancellation is as follows: Figure 1 As shown.

[0053] Furthermore, the secondary pickup unit picks up the suppressed residual noise, which is an analog signal, through a secondary microphone. The secondary pickup unit amplifies the power of the picked-up analog noise signal and filters out the high-frequency components of the signal using a low-pass filter. Then, the high-frequency filtered signal is converted from analog to digital and input into the ANC main control module. This process generates low-frequency audio data, which is provided to the ANC main control module for relearning and evaluation of the active noise cancellation effect.

[0054] This invention provides an active noise reduction safety protection device for pumped storage power stations. Its key features include: firstly, utilizing the selective noise reduction properties of sound-absorbing cotton for high-frequency noise, using polyester fiber sound-absorbing cotton for sound insulation within a wearable outer shell to reduce high-frequency noise; secondly, employing an active noise reduction controller based on a reinforcement learning algorithm to actively reduce the steady-state low-frequency noise generated by the generator set. The overall noise reduction effect is as follows: Figure 5 As shown.

[0055] Active noise reduction methods include the following steps:

[0056] Step 1: The system starts up. The ANC main control module receives low-frequency noise data generated by the first-level pickup unit. This noise remains in a steady state over multiple time periods (duration coefficient T), and its loudness and frequency distribution characteristics remain basically unchanged.

[0057] Step Two: The ANC main control module statistically analyzes the frequency, amplitude, and other data of low-frequency noise data. Based on the duration coefficient T, it models and trains the audio data, constructing a noise sequence using dynamic programming based on reinforcement learning. As the name suggests, dynamic programming implies changes in the noise frequency sequence and amplitude state; programming implies optimization, such as linear optimization, quadratic optimization, or nonlinear optimization. Based on the characteristics of the generator set's "steady-state noise," the entire optimization problem can be decomposed into multiple sub-optimization problems. The solutions to these sub-optimization problems can be stored and reused, fully satisfying the dynamic programming conditions. A complete reinforcement learning process can be represented using tuples (S, A, P, r, y), where S is the state set, A is the action set, P is the transition probability (corresponding to the noise model), r is the reward function, and y is the discount factor used to calculate the cumulative reward R. The cumulative reward formula is... Where 0≤y≤1 and T is a finite value, it refers to the duration coefficient T. This invention reduces a reinforcement learning process to a noisy sequence decision problem, that is, finding a decision sequence that optimizes the noise reduction effect.

[0058] Step 3: The ANC main control module inverts the noise sequence to obtain inverted audio data, which is then output to the speaker via the acoustic driver unit to suppress the original noise. This process is acoustically defined as "active noise reduction." Figure 1 As shown. Sound wave phase is measured in degrees. For a sine wave, 0° represents the starting point of the wave. The first peak is 90°, the waveform becomes negative at 180°, and completes a full cycle at 360°. Therefore, inverted audio data can be obtained by simply inverting each bit of the noise sequence.

[0059] Step 4: The ANC main control module receives the residual noise data generated by the secondary pickup unit, extracts the residual noise amplitude sequence, and determines whether the residual noise amplitude is higher than the threshold. If it is higher than the threshold, the duration coefficient K is adjusted and the model is rebuilt; otherwise, the existing inverted audio signal is directly output to continuously suppress noise. This process is called "receiving feedback." In reinforcement learning, feedback from the decision-making process must be received; this feedback is called a reward. If a positive reward is received, it indicates that the current decision needs to be strengthened (the duration coefficient T should be appropriately shortened, with a minimum value not lower than 0.5 seconds). If a negative feedback is received, it indicates that the current decision needs to be weakened (the duration coefficient T should be appropriately increased, with a maximum value not higher than 3 seconds).

[0060] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An active noise reduction method for an active noise reduction safety protection device in a pumped storage power station, characterized in that: The active noise reduction method is used for active noise reduction of the active noise reduction safety protection device of the pumped storage power station, the active noise reduction safety protection device of the pumped storage power station includes: Wearable shell (1), the inner wall of the wearable shell (1) is provided with sound-absorbing cotton (11), and an inner lining frame (12) is fixedly provided inside the sound-absorbing cotton (11). The active noise cancellation mechanism includes a pickup end, a speaker end, and an active noise cancellation unit. The inner walls on both sides of the wearable shell (1) are provided with insertion holes, and the inner walls on both sides of the inner frame (12) are provided with receiving grooves. The back of the inner frame (12) is provided with a slot. The pickup end includes a primary pickup A (2), a secondary pickup A (21), a primary pickup B (22), and a secondary pickup B (23). The speaker end includes a loudspeaker A (3) and a loudspeaker B (31). The active noise cancellation unit includes an active noise cancellation controller A (4) and an active noise cancellation controller set in the slot. B(41), the first-level pickup A(2) and the first-level pickup B(22) are respectively installed in the sockets on the left and right sides of the wearable shell (1), the second-level pickup A(21), the speaker A(3), the second-level pickup B(23), and the speaker B(31) are respectively installed in the receiving slots on the left and right sides of the inner frame (12), the first-level pickup A(2), the second-level pickup A(21), and the speaker A(3) are respectively connected to the active noise cancellation controller A(4), and the first-level pickup B(22), the second-level pickup B(23), and the speaker B(31) are respectively connected to the active noise cancellation controller B(41); Both the active noise cancellation controller A (4) and the active noise cancellation controller B (41) include a primary pickup unit, a secondary pickup unit, an acoustic drive unit, and an ANC main control module. The primary pickup unit of the active noise cancellation controller A (4) is connected to the primary pickup A (2). The secondary pickup unit of the active noise cancellation controller A (4) is connected to the secondary pickup A (21). The primary pickup unit of the active noise cancellation controller B (41) is connected to the primary pickup B (22). The secondary pickup unit of the active noise cancellation controller B (41) is connected to the secondary pickup B (23). The acoustic drive units of the active noise cancellation controller A (4) and the active noise cancellation controller B (41) are connected to the loudspeaker A (3) and the loudspeaker B (31), respectively. The primary pickup unit, the secondary pickup unit, and the acoustic drive unit in the active noise cancellation controller A (4) and the active noise cancellation controller B (41) are connected to the ANC main control module, respectively. The active noise reduction method includes the following steps: S1. The primary pickup unit picks up ambient noise filtered by sound-absorbing cotton, and then connects it to the ANC main control module after signal amplification, filtering and analog-to-digital conversion. S2, the ANC main control module calculates the amplitude of the noise corresponding to each frequency point of the noise audio with the duration coefficient T, performs modeling and training, and finally generates an inverted audio signal, which is output to the acoustic drive unit. The duration coefficient T is set to 0.5 seconds to 3 seconds. S3, the acoustic drive unit converts the inverted audio signal input from the ANC main control module into a digital-to-analog converter, amplifies the signal, and outputs it to the speaker; S4, the secondary pickup unit picks up the residual noise after phase inversion suppression, and then connects it to the ANC main control module after signal amplification, low-pass filtering and analog-to-digital conversion; The S5 and ANC main control modules determine whether the residual noise amplitude is higher than the threshold. If it is higher than the threshold, the duration coefficient T is adjusted and the model is re-modeled. Otherwise, the existing inverted audio signal is directly output to continuously suppress noise.

2. The active noise reduction method according to claim 1, characterized in that: In step S2, modeling and training are performed using dynamic programming based on reinforcement learning to construct the noise sequence. The entire optimization problem is decomposed into multiple sub-optimization problems. The solutions to these sub-optimization problems are stored and reused, satisfying the dynamic programming conditions. A complete reinforcement learning process is represented by a tuple (S, A, P, r, y), where S is the state set, A is the action set, P is the transition probability (corresponding to the noise model), r is the reward function, and y is the discount factor used to calculate the cumulative reward R. The cumulative reward formula is: , where 0≤y≤1, and when T is a finite value, it is the length coefficient T.

3. The active noise reduction method according to claim 1, characterized in that: In step S3, specifically, the ANC main control module inverts the noise sequence to obtain inverted audio data, and then outputs the inverted audio data to the speaker via the acoustic drive unit to suppress the original noise.

4. An active noise reduction safety protection device for pumped storage power stations, characterized in that: The active noise reduction safety protection device for pumped storage power stations is used to perform the active noise reduction method according to any one of claims 1-3.

Images