A converter station sound insulation and noise reduction type enclosure structure design method based on artificial intelligence

Abstract

This invention discloses an artificial intelligence-based design method for the sound insulation and noise reduction enclosure structure of converter stations, specifically relating to the field of converter station noise control. It addresses the problems of low-frequency control difficulties and noise leakage from ventilation openings leading to mismatch between the design scheme and station boundary targets in converter station enclosure structure design. The method involves obtaining equivalent continuous sound levels at representative points at the station boundary according to specifications and recording equipment operating status. These levels are then compared with existing predicted sound levels of the same diameter to obtain source term corrections, generating an equivalent sound power parameter set. Combining ventilation and heat dissipation constraints with the station's layout boundary calculation criteria, machine learning is used to comprehensively obtain the governance allocation coefficient. This guides the retrieval of the enclosure unit library to generate candidate enclosure combinations and matching opening noise reduction configurations. Finally, sound field and thermal ventilation checks are performed to output a constructable layout and node details, improving design feasibility and consistency.

Description

Technical Field

[0001] This invention relates to the field of noise control in converter stations, and more specifically, to a design method for a sound insulation and noise reduction enclosure structure of converter stations based on artificial intelligence. Background Technology

[0002] With converter stations increasingly located near urban built-up areas and surrounding residential and office environments expanding, electrical equipment such as converter transformers and their cooling systems within the stations often become major sources of noise. This noise includes low-frequency line spectrum components caused by equipment vibration and other mechanisms, as well as noise exhibiting different frequency band characteristics depending on the operating status of the cooling devices. This makes the external environment more susceptible to the disturbance caused by the continuous "low-frequency roar" and the superposition of broadband noise. Existing technical literature has already pointed out similar problems in substation scenarios and attempted to address them primarily through propagation path mitigation. For example, the patent application "A Full-Band Noise Barrier and Shielding Device for Substations" (Publication No. CN106192785A) explicitly mentions the urbanization of substations and the increase in noise complaints in its background section, explaining that transformer noise is mainly low-frequency line spectrum and exhibits a full-band distribution. Meanwhile, the patent application "A Noise Reduction Method for Substations" (Publication No. CN106285083A) also explains in its background section that in addition to equipment noise, the main transformer may also induce structural resonance, generating low-frequency secondary structural noise, thus creating a superimposed impact on the surrounding environment.

[0003] In the aforementioned scenarios, while existing technologies propose measures such as deploying sound barriers or soundproof enclosures along the propagation path to reduce transmitted noise, they still have unavoidable shortcomings in practical applications: low-frequency noise has stronger penetration and diffraction characteristics, and conventional sound barriers or enclosure structures are prone to ineffectiveness in the low-frequency range. To compensate for this shortcoming, the structure often has to be made thicker and heavier, which brings inconvenience in transportation, installation, and construction organization, and even squeezes the space for layout and maintenance within the station, making it difficult to achieve a balance between engineering conditions and governance objectives. More importantly, low-frequency problems do not only originate from airborne propagation. Equipment vibration may induce structural resonance, forming secondary structural noise. This type of noise can bypass some airborne noise control measures through the structural path, so even with the addition of barriers or enclosures, there may still be significant low-frequency disturbances outside the station. Ultimately, this results in a situation where repeated installation and thickening are needed to achieve a stable solution, and the overall cost and risk increase simultaneously.

[0004] To address the aforementioned problems, a technical solution is provided. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an artificial intelligence-based design method for the sound insulation and noise reduction enclosure structure of converter stations. This method involves obtaining equivalent continuous sound levels at representative points at the station boundary according to specifications and recording equipment operating status. These levels are then compared with existing predicted sound levels of the same diameter to obtain source term corrections, generating an equivalent sound power parameter set. Combining ventilation and heat dissipation constraints with station layout boundary calculations, a discrimination index is calculated. Machine learning is then used to synthesize the governance allocation coefficient, guiding the retrieval of the enclosure unit library to generate candidate enclosure combinations and matching opening silencing configurations. Finally, sound field and thermal ventilation checks are performed to output a constructable layout and node details, thereby improving design feasibility and consistency and addressing the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] S1: In accordance with acoustic measurement specifications, acoustic measurement results are collected at the representative point of the station boundary and the equivalent continuous sound level at the representative point of the station boundary is calculated. At the same time, the operating status of the main equipment is recorded to form on-site input data.

[0008] S2: Compare the on-site input data with the predicted sound level at the station boundary obtained by the existing noise prediction method using the same caliber to obtain the source term correction amount, and generate an equivalent sound power parameter set based on the source term correction amount;

[0009] S3: Integrate the equivalent sound power parameter set with ventilation and layout constraints, calculate the discrimination index for classification and generate comprehensive treatment allocation coefficients, use the treatment allocation coefficients to guide the unit library search, obtain candidate combinations of building envelope structures and output the matching opening silencing configurations.

[0010] S4: Perform sound field simulation and thermal ventilation verification on the candidate combinations of the building envelope along with the opening silencing configuration, screen and determine the final building envelope design results, and output the building envelope layout diagram and node construction list.

[0011] Furthermore, step S1 includes:

[0012] Multiple representative points were selected outside the station boundary. These representative points covered the direction in which the noise contribution of the converter transformer and cooling system was greatest and were evenly distributed around the station boundary. Integrating sound level meters and spectrum analyzers were used to continuously measure and collect the time sequence of frequency band sound pressure level with 1 / 3 octave band resolution and the corresponding A-weighted instantaneous sound pressure level.

[0013] Furthermore, step S1 includes:

[0014] The time series data of A-weighted instantaneous sound pressure level at representative points at each station boundary and the time series data of sound pressure level in each 1 / 3 octave band are processed by sound energy averaging to obtain the A-weighted equivalent continuous sound level and the band equivalent sound pressure level.

[0015] Furthermore, step S2 includes:

[0016] An outdoor sound propagation attenuation algorithm was used in conjunction with the initial sound power level spectrum to calculate the A-weighted predicted equivalent continuous sound level and the predicted sound pressure level of each 1 / 3 octave band representative point. Under the same station boundary representative point and frequency band resolution, the measured sound level and the predicted sound level were compared to obtain the A-weighted sound level difference and the sound level difference of each frequency band.

[0017] Furthermore, step S2 also includes:

[0018] The A-weighted sound level difference and the sound level difference of each frequency band at all representative points of the station boundary are averaged to obtain the A-weighted source term correction and the source term correction of each frequency band. The corrections are then superimposed on the initial sound power level to generate the A-weighted equivalent sound power level and the equivalent sound power level of each frequency band, and integrated into a structured equivalent sound power parameter set with frequency band identification.

[0019] Furthermore, step S3 includes:

[0020] The equivalent sound power parameter set, ventilation and heat dissipation constraints, and site layout boundary are preprocessed. The proportion of low-frequency relative sound energy is calculated based on the measured frequency band results. The propagation dominance correction factor is obtained by extracting the propagation attenuation difference using existing noise prediction methods and correcting the proportion of low-frequency relative sound energy to obtain the low-frequency propagation dominance ratio.

[0021] Furthermore, step S3 also includes:

[0022] The effective ventilation area is calculated based on the geometric opening area of ​​the ventilation opening and the effective ventilation opening ratio. The insertion loss is converted into an area attenuation factor to obtain the total equivalent sound leakage area of ​​the opening. The sound transmission coefficient is converted based on the area of ​​the enclosure panel and the sound insulation of the panel to obtain the total sound transmission equivalent area of ​​the enclosure. The ratio of the total equivalent sound leakage area of ​​the opening to the total sound transmission equivalent area of ​​the enclosure is calculated to obtain the opening equivalent sound leakage ratio.

[0023] Furthermore, step S3 also includes:

[0024] The low-frequency propagation dominance ratio and the equivalent sound leakage ratio of the openings are normalized. The governance allocation coefficients are obtained by mapping through a pre-trained probabilistic discriminant model. Based on the range of governance allocation coefficients, the retrieval strategy and the key constraints are selected. The candidate combinations of the building envelope structure and the matching opening silencing configurations are output from the building envelope unit library.

[0025] Furthermore, step S4 includes:

[0026] Load candidate combinations of building envelope and opening noise reduction configurations into the simulation environment, input the equivalent sound power parameter set as the sound source, apply frequency-dependent sound insulation and insertion loss boundary conditions to perform sound field simulation verification, and input the effective ventilation area of ​​the opening and the power of the heat source in the station to perform thermal ventilation verification.

[0027] Furthermore, step S4 also includes:

[0028] Based on the predicted A-weighted equivalent continuous sound level and total ventilation volume screening criteria, unqualified schemes were eliminated, and the comprehensive performance index of the remaining schemes was calculated to determine the final design result of the building envelope.

[0029] The technical effects and advantages of the artificial intelligence-based design method for the sound insulation and noise reduction enclosure structure of converter stations in this invention are as follows:

[0030] This invention aligns the standardized measurement results of representative points at the station boundary with existing predictive standards, calibrating the source inputs most prone to deviation from reality into a parameter basis suitable for engineering calculations. Furthermore, it introduces a decision-making allocation mechanism based on discriminant indicators before scheme generation. This allows the selection and combination of the enclosure structure to move beyond empirical thickening or single material specifications, enabling targeted configuration around the main outward transmission paths. This ensures that noise reduction targets are more stably implemented in a constructable structural scheme while meeting engineering constraints such as station layout and ventilation. The continuous progressive relationship from on-site input to source correction, and from decision guidance to verification and finalization, significantly reduces repeated trial and error and subsequent rectification caused by frequency band mismatches or neglected sound leakage from openings. This makes the final enclosure structure scheme more feasible and consistent, facilitating stable and verifiable noise control effects in scenarios where converter stations operate near sensitive areas. Attached Figure Description

[0031] Figure 1 This is a flowchart illustrating the design method for a sound insulation and noise reduction enclosure structure of a converter station based on artificial intelligence, according to the present invention. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] Example 1: Figure 1 This invention presents a design method for a sound insulation and noise reduction enclosure structure of a converter station based on artificial intelligence, comprising:

[0034] S1: In accordance with acoustic measurement specifications, acoustic measurement results are collected at the representative point at the station boundary and the equivalent continuous sound level at the representative point at the station boundary is calculated. At the same time, the operating status of the main equipment is recorded to form on-site input data.

[0035] S2: Compare the on-site input data with the predicted sound level at the station boundary obtained by the existing noise prediction method using the same caliber to obtain the source term correction, and generate an equivalent sound power parameter set based on the source term correction.

[0036] S3: Integrate the equivalent sound power parameter set with ventilation and layout constraints, calculate the discrimination index for classification, and comprehensively generate the treatment allocation coefficient. Use the treatment allocation coefficient to guide the unit library search, obtain candidate combinations of building envelope structures, and output the matching opening silencing configuration.

[0037] S4: Perform sound field simulation and thermal ventilation verification on the candidate combinations of the building envelope along with the opening silencing configuration, screen and determine the final building envelope design results, and output the building envelope layout diagram and node construction list.

[0038] This invention addresses the sound insulation and noise reduction design of converter station enclosure structures, proposing an intelligent design method that starts with limited traceable field information and concludes with computational verification. The approach involves first obtaining acoustic results representing the actual disturbance level at a representative location at the station boundary according to specifications, and then combining this with the equipment's operating status to form a unified field input. This input is then compared with existing noise prediction standards, and the difference between prediction and reality is transformed into source-end correction parameters that can be used for design, thereby avoiding the selection of enclosure structures based on empirical source strength or static assumptions.

[0039] Building upon this foundation, this invention introduces a discrimination and allocation approach oriented towards engineering failure mechanisms. It calculates and generates decision quantities to indicate the main outward transmission paths, guiding the retrieval and combination generation of the building envelope unit library. This ensures that candidate building envelope schemes and opening soundproofing configurations can simultaneously address sound insulation requirements and ventilation constraints during the generation phase, reducing the risk of ineffective thickening and overlooked sound leakage from openings. Finally, candidate schemes undergo acoustic and thermal ventilation verification and screening, outputting layouts and node construction methods directly applicable to construction. This achieves a seamless connection from on-site input, source-end correction, scheme generation to project implementation, thereby improving the stability and feasibility of achieving the desired outcome on the first attempt.

[0040] With converter stations increasingly located near urban built-up areas and surrounding sensitive areas growing, noise generated by converter transformers and their cooling systems has become a major source of environmental disturbance. This noise contains both low-frequency line spectrum components and broadband characteristics, leading to persistent low-frequency booming and broadband superposition effects outside the station. While existing technologies attempt to mitigate the propagation path through sound barriers or enclosure structures, the strong penetrating power of low frequencies, structural sound transmission, and sound leakage through openings often result in unstable mitigation effects, with repeated thickening failing to meet standards. To achieve targeted enclosure structure design, it is essential to first establish a traceable and strictly corresponding field noise input basis that corresponds to actual operating conditions. Therefore, step S1 focuses on standardized acoustic measurements and data processing at the station boundary, providing standardized field input data for subsequent source term correction and mitigation path identification.

[0041] S101 determines the location of the representative point of the station boundary.

[0042] The noise propagation from converter stations has directional differences, and the main noise contribution is often concentrated in the station boundary section on the side of the converter transformer and cooling system. In order to ensure that the measurement results can fully represent the environmental noise exposure level at the station boundary, it is necessary to first scientifically select the measurement point locations to cover the main propagation paths.

[0043] Based on the measurement requirements of the "Emission Standard for Environmental Noise at the Boundary of Industrial Enterprises" (GB 12348—2008) and "Description, Measurement and Evaluation of Acoustic Environmental Noise Part 2: Determination of Environmental Noise Level" (GB / T 3222.2-2022), and considering the characteristics of the converter station boundary, multiple representative measurement points were selected at appropriate distances from the station boundary, at appropriate heights above the ground, and away from obvious reflective surfaces. The representative measurement points needed to cover the directions contributing significantly to station boundary noise, especially the boundary sections near the converter transformer and cooling system, while avoiding reflection interference caused by local terrain, structures, or green belts. At least four points were selected, distributed as far around the station boundary as possible, with priority given to locations with a high number of historical noise complaints or predicted significant contributions. After determining the locations of each representative measurement point, their geographical coordinates and numbers were recorded to form a list of measurement point identifiers.

[0044] The S102 acquires frequency band-resolved sound pressure level timing data.

[0045] After the location of the representative point at the station boundary is determined, in order to capture the complete spectral information of the low-frequency line spectrum and broadband characteristics of the converter station noise, as well as the dynamic changes under the operating conditions, it is necessary to use high-precision equipment to perform continuous frequency band resolution measurement.

[0046] Using an integrating sound level meter and spectrum analyzer meeting Class I accuracy requirements, continuous measurements were conducted at representative measurement points according to the "Emission Standard for Environmental Noise at the Boundary of Industrial Enterprises" (GB 12348—2008) and relevant industry or national standards for audible noise at converter stations. The measurement period covered the converter transformer and cooling system under full-load stable operation, lasting at least thirty minutes, with a sampling interval of no more than one second. The collected data included time-series sound pressure levels in the first and third octave bands covering a center frequency range of 20 Hz to 20 kHz, as well as the A-weighted instantaneous sound pressure levels at corresponding times. All frequency band sound pressure level data were stored with representative measurement point numbers and time series identifiers.

[0047] Example: During the measurement at the boundary of a ±800 kV converter station, technicians selected one representative measurement point each on the east side near the converter transformer, the west side towards the cooling tower, the south side, and the north side. Using a calibrated Class I integrating sound level meter fixed on a tripod that meets the height requirements of the measurement specifications, they continuously recorded data for a sufficient duration during the full-load operation of the equipment to obtain a complete frequency band sound pressure level time sequence for each measurement point.

[0048] S103 calculates the equivalent continuous sound level at the representative point of the station boundary.

[0049] After the frequency band-resolved sound pressure level time series data is acquired, in order to obtain stable and comparable noise level indicators and eliminate the influence of instantaneous fluctuations, it is necessary to perform sound energy averaging on the time series data and calculate the equivalent continuous sound level of each representative measurement point.

[0050] For each representative measurement point, the A-weighted equivalent continuous sound level is calculated based on the collected A-weighted instantaneous sound pressure level time-series data. The specific calculation process is as follows: First, the A-weighted instantaneous sound pressure level value at each sampling time is divided by 10 and then raised to the power of 10 to obtain the sound energy ratio value at the corresponding time; then, the arithmetic mean of the sound energy ratio values ​​at all sampling times is calculated to obtain the average sound energy ratio value; finally, the average sound energy ratio value is multiplied by 10 and the logarithm to the base 10 is taken to obtain the A-weighted equivalent continuous sound level, which is rounded to 0.1 dB.

[0051] Simultaneously, for each 1 / 3 octave band, the equivalent sound pressure level of that band is calculated. The specific calculation process is as follows: First, the band sound pressure level value at each sampling time of that band is divided by 10 and then raised to the power of 10 to obtain the band sound energy ratio value at the corresponding time; then, the arithmetic mean of the band sound energy ratio values ​​at all sampling times is calculated to obtain the average sound energy ratio value of that band; finally, the average sound energy ratio value of that band is multiplied by 10 and then the logarithm to the base 10 is taken to obtain the equivalent sound pressure level of that band, rounded to 0.1 dB.

[0052] Example: In the above converter station measurement, the technicians performed sound energy averaging on the A-weighted instantaneous sound pressure level collected per second over 30 minutes at the representative measurement point on the east side, and finally obtained an A-weighted equivalent continuous sound level of 52.3 dB at the point. At the same time, the full-band 1 / 3 octave equivalent sound pressure level spectrum was obtained for subsequent comparison with the prediction results.

[0053] S104 synchronously records the operating status of major equipment.

[0054] During acoustic measurement, the noise level of the converter station is closely related to the load of the converter transformer and the operating mode of the cooling system. To ensure that the measurement data strictly corresponds to the actual sound source contribution, the equipment operating parameters need to be recorded synchronously throughout the measurement process.

[0055] Throughout the acoustic measurement process, operating parameters of the converter transformer and cooling system are recorded synchronously, including load rate, number of cooling fan speed settings, and cooling system operating mode. Recording is done in a timestamped log table to ensure strict time alignment with the sound pressure level time-series data. The recorded content forms a set of equipment status parameters, including key variables such as load percentage and fan speed settings.

[0056] Example: During the measurement, the operators recorded the current load rate of the converter transformer as 95%, the cooling system as being in forced air cooling mode, and all 8 cooling fans as being running at high speed through the station monitoring system. These parameters were then written into the log table in correspondence with the timestamps of the sound level meter.

[0057] S105 organizes and generates the field input data package.

[0058] After the aforementioned measurements and records are completed, in order to facilitate direct access to subsequent steps and maintain data traceability, all results need to be integrated into a structured data package.

[0059] The coordinates and numbers of each representative measurement point, A-weighted equivalent continuous sound level, full-band 1 / 3 octave band equivalent sound pressure level spectrum, equipment status parameter set corresponding to the measurement period, measurement time, and meteorological conditions are integrated into a structured field input data package. The data package is stored in a table or database format, and all data items include representative measurement point identifiers and timestamps to ensure completeness and consistency. It is then directly output to subsequent source item correction steps.

[0060] Step S1 completes the standardized measurement and data processing of representative points at the station boundary, forming a structured field input data package with measurement point identifiers and timestamps. This data package includes the A-weighted equivalent continuous sound level, the full-band 1 / 3 octave band equivalent sound pressure level spectrum, and a set of equipment status parameters such as the converter transformer load rate and cooling system operation mode recorded simultaneously at each representative point. These data strictly correspond to the actual noise contribution under the stable full-load operation of the converter station and are directly used for comparison with existing noise prediction methods in step S2, ensuring that source term corrections are based on traceable actual measurements and avoiding misjudgments of mitigation paths due to prediction biases.

[0061] Step S1 has generated a structured field input data package containing the weighted equivalent continuous sound level at representative point A of each station boundary, the full-band 1 / 3 octave equivalent sound pressure level spectrum, and the status parameter set of the synchronous equipment. These data strictly reflect the actual noise contribution of the converter station under full-load operation. However, existing noise prediction methods rely on standard sound source databases and simplified propagation models, which are prone to systematic deviations under the influence of specific site topography, equipment aging, or operational variations. Therefore, step S2 compares the field input data with the prediction results using the same aperture, extracts the source term correction, and generates an equivalent sound power parameter set to calibrate the sound source intensity to a level consistent with the actual measurement, providing accurate sound source input for subsequent calculations of low-frequency propagation dominance and opening equivalent leakage ratio.

[0062] S201 calls upon existing noise prediction methods to calculate the predicted sound level at the station boundary.

[0063] Step S1 has generated a structured field input data packet. In order to separate the influence of sound source intensity deviation and propagation path, it is necessary to rerun the existing noise prediction method under the same equipment operating conditions and representative points at the station boundary to obtain the predicted sound level with the same aperture as the actual measurement.

[0064] Existing noise prediction methods can employ the ISO 9613-2 outdoor sound propagation attenuation algorithm combined with the standard sound power level of the equipment, or they can incorporate engineering prediction practices recommended in relevant specifications for converter station noise control design. Input parameters include the initial sound power level spectrum of the converter transformer and cooling system, the geometric location of equipment within the station, ground absorption characteristics, meteorological correction terms, and the coordinates of representative points at the station boundary. For each representative point at the station boundary, the A-weighted predicted equivalent continuous sound level and the predicted sound pressure level for each octave band are calculated. The calculation process first calculates the free-field sound pressure contribution band-by-band based on the initial sound power level, then superimposes the propagation attenuation effects of ground reflection, air absorption, atmospheric refraction, and terrain shielding, finally obtaining the predicted sound level by superimposing sound energy, maintaining the same frequency band resolution and A-weighting aperture as the measured data in step S1.

[0065] Example: In the design of a ±800 kV converter station, the technicians input the initial sound power level spectrum and the coordinates of the equipment layout in the station provided by the converter transformer manufacturer into the prediction software. After running the calculation, the weighted predicted equivalent continuous sound level of the representative point A at each station boundary and the predicted sound pressure level spectrum of the full-band 1 / 3 octave band were obtained.

[0066] S202 Comparison of measured sound level and predicted sound level with the same aperture.

[0067] After the predicted sound level is calculated, in order to quantify the combined deviation between the sound source intensity and the propagation model, it is necessary to directly compare the measured sound level and the predicted sound level under the same measurement point and frequency band conditions.

[0068] The comparison was conducted at representative points at the same station boundaries, with the same frequency band resolution and under the same equipment operating conditions. First, the overall A-weighted sound level difference was calculated. This was done by subtracting the predicted A-weighted equivalent continuous sound level from the measured A-weighted equivalent continuous sound level at the representative station boundary. Simultaneously, the band-specific sound level difference was calculated for each 1 / 3 octave band. This was done by subtracting the predicted sound pressure level for that band from the measured equivalent sound pressure level for that band.

[0069] S203 calculates the source term correction.

[0070] The sound level differences at multiple representative points at the station boundaries have been obtained. In order to obtain a unified equivalent correction that can reflect the combined impact of sound source database deviation and propagation model deviation at the current station site, it is necessary to perform sound energy averaging on the differences at all representative points.

[0071] The calculation process for the A-weighted overall correction is as follows: First, divide the A-weighted sound level difference at each representative station boundary by 10 and then power it with 10 to obtain the sound energy ratio value for the corresponding representative point; then, calculate the arithmetic mean of the sound energy ratio values ​​for all representative points to obtain the average sound energy ratio value; finally, multiply the average sound energy ratio value by 10 and take the logarithm to the base 10 to obtain the A-weighted source term correction. The calculation process for each 1 / 3 octave band correction is as follows: First, divide the sound level difference for that band at each representative station boundary by 10 and then power it with 10 to obtain the sound energy ratio value for the corresponding representative point; then, calculate the arithmetic mean of the sound energy ratio values ​​for that band at all representative points to obtain the average sound energy ratio value for that band; finally, multiply the average sound energy ratio value for that band by 10 and take the logarithm to the base 10 to obtain the band source term correction. This step outputs the A-weighted source term correction and the source term correction values ​​for each band.

[0072] Example: In the converter station mentioned above, technicians performed sound energy averaging on the A-weighted sound level difference at four representative points at the station boundary. The result showed that the A-weighted source term correction was positive, indicating that the initial sound power level was underestimated.

[0073] S204 generates an equivalent acoustic power parameter set.

[0074] The source term correction is used to characterize the systematic deviation between the initial sound power level and the propagation calculation under the current operating conditions from the measured results. To ensure that the sound source description used in subsequent calculations is equivalent to the on-site noise level in an engineering sense, this correction needs to be superimposed on the initial sound power level to form an equivalent sound power level.

[0075] For the overall A-weighted equivalent sound power level, the calculation process involves adding the A-weighted source term correction to the initial A-weighted sound power level to obtain the A-weighted equivalent sound power level. For each 1 / 3 octave band equivalent sound power level, the band equivalent sound power level is obtained by adding the source term correction to the initial sound power level of that band.

[0076] S205 organizes and outputs an equivalent acoustic power parameter set.

[0077] After the equivalent sound power parameter set is generated, in order to facilitate direct calling in subsequent path identification, all the corrected sound power levels need to be integrated into a structured format.

[0078] The A-weighted equivalent sound power level, the equivalent sound power level of each frequency band, and the corresponding equipment status parameter set are integrated into a structured equivalent sound power parameter set. All data are stored with frequency band identifiers and directly output to the subsequent low-frequency propagation dominance ratio and opening equivalent leakage ratio calculation steps.

[0079] Step S2 completes the comparison of the measured sound level and the predicted sound level on the same aperture and corrects the source terms, forming a structured equivalent sound power parameter set that includes the A-weighted equivalent sound power level and the equivalent sound power level of each frequency band. This parameter set directly reflects the actual sound source intensity of the converter transformer and cooling system under full load operation. The correction process eliminates the systematic bias of the initial prediction model, ensuring that the sound source input is equivalent to and consistent with the measured noise level, and directly outputs it to the subsequent path discrimination step to avoid incorrect allocation of treatment paths due to misjudgment of sound source intensity.

[0080] Step S2 has generated an equivalent sound power parameter set that is strictly equivalent to the measured noise level. This parameter set includes the A-weighted equivalent sound power level and the equivalent sound power level spectrum of each 1 / 3 octave band, directly reflecting the actual sound source intensity of the converter transformer and cooling system. However, the external transmission of noise from the converter station is often dominated by the low-frequency propagation dominance and the sound leakage path of openings. Therefore, step S3 preprocesses the equivalent sound power parameter set with ventilation and heat dissipation constraints and the layout boundary within the station. It first calculates the low-frequency propagation dominance ratio and the equivalent sound leakage ratio of openings, and then maps them to the governance allocation coefficients through the historical engineering sample probability discrimination model. This guides the retrieval of the enclosure unit library and outputs candidate combinations of enclosure structures and opening noise reduction configurations that match the actual failure paths.

[0081] The reason for selecting the low-frequency propagation dominance ratio and the equivalent sound leakage ratio of the opening in step S3 is that the most common and decisive failure mechanism of the enclosure structure in the noise reduction project of the converter station usually focuses on two independent main channels that are most easily underestimated by conventional design paths: First, low-frequency noise is relatively dominant in propagation attenuation and diffraction characteristics, which makes it easy to mismatch the target frequency band when selecting the model based solely on the sound insulation index of the material or the empirical spectrum; Second, the sound leakage channel formed by the ventilation opening and its auxiliary components often outperforms the sound transmission channel of the panel in actual contribution, making it difficult to convert the improvement of the panel index into noise reduction benefits at the station boundary. The two parameters mentioned above are used to quantify the two types of main channels using dimensionless indices with unified dimensions. Their calculations rely on two essential but minimal input data: station boundary frequency band measurement and opening structure information. This allows for the engineering classification and constraint focus determination of "low-frequency dominance" and "opening dominance" before candidate combinations are generated. This enables the intelligent structure selection model to select different enclosure unit retrieval strategies and opening noise reduction configuration strategies based on the classification results. This avoids ineffective thickening, avoids investing governance resources in non-dominant channels, and reduces the risk of repeated schemes and rectifications caused by frequency band mismatch or neglect of opening noise leakage.

[0082] S301 Preprocessing Equivalent Sound Power Parameter Set and Constraints.

[0083] Step S2 has output a set of structured equivalent acoustic power parameters. In order to ensure that the path discrimination is strictly compatible with the engineering conditions of the converter station, it is necessary to first integrate the ventilation and heat dissipation constraints and the layout boundary within the station to form a unified preprocessing dataset.

[0084] Ventilation and heat dissipation constraints include the total heat dissipation requirements of the converter transformer and cooling system, the maximum allowable enclosure height on one side, and the minimum effective ventilation area threshold for openings, determined by in-station heat balance calculations or design specifications. In-station layout boundaries include the outline of the area that the enclosure structure can occupy, the width of maintenance access routes, and the location of equipment foundations, extracted from the in-station CAD layout drawings. The equivalent sound power parameter set is correlated with the ventilation and heat dissipation constraints and the in-station layout boundaries to form a preprocessed dataset. All data is stored with equipment status identifiers and is directly used for subsequent propagation simulations and area calculations.

[0085] S302 calculates the relative proportion of low-frequency acoustic energy.

[0086] The low-frequency components of the converter station noise contribute significantly to the outward-propagating disturbance. To quantify their energy proportion, it is necessary to perform energy conversion and proportion calculation based on the measured frequency band results in step S1.

[0087] The predefined low-frequency band set is the 1 / 3 octave band with a center frequency of 63 Hz and below, including the 31.5 Hz, 40 Hz, 50 Hz, and 63 Hz bands. For each representative point at the station boundary, the equivalent sound pressure level of each 1 / 3 octave band across the entire frequency band is first divided by 10 and then raised to the power of 10 to obtain the relative sound energy of each band. Then, the relative sound energies of all bands are summed to obtain the total relative sound energy of the entire frequency band. Simultaneously, the relative sound energies within the low-frequency band set are summed to obtain the low-frequency relative sound energy. The low-frequency relative sound energy is divided by the total relative sound energy of the entire frequency band to obtain the proportion of low-frequency relative sound energy at that representative point. The arithmetic mean of the proportions of low-frequency relative sound energy at all representative points at the station boundary is calculated to obtain the overall low-frequency relative sound energy proportion, which is dimensionless and ranges from 0 to 1.

[0088] Example: In the noise control design of a ±800 kV converter station, technicians extracted the relative sound energy of the frequency band below 63 Hz from the measured frequency spectrum. After calculation, the proportion of the overall low-frequency relative sound energy was found to be approximately 0.45.

[0089] S303 calculates the propagation dominance correction factor and outputs the low-frequency propagation dominance ratio.

[0090] The contribution of low-frequency relative sound energy has been quantified by actual measurements. To further characterize the advantage of low-frequency noise in the propagation process, it is necessary to introduce a propagation attenuation difference correction based on the equivalent sound power parameter set.

[0091] The predefined mid-frequency band set is a 1 / 3 octave band from a center frequency of 125 Hz to 500 Hz. Using existing noise prediction methods and the equivalent sound power parameter set from step S2, the average propagation attenuation of the low-frequency band set and the average propagation attenuation of the mid-frequency band set are extracted at representative points at the same station boundary, in decibels. The average propagation attenuation at the mid-frequency level is calculated by subtracting the average propagation attenuation at the low-frequency level to obtain the propagation attenuation difference. This difference is divided by 10 and then raised to the power of 10 to obtain the propagation dominance correction factor. The arithmetic mean of the propagation dominance correction factors for all representative points is calculated to obtain the overall propagation dominance correction factor. The overall low-frequency relative sound energy ratio is multiplied by the overall propagation dominance correction factor to obtain the low-frequency propagation dominance ratio, which is dimensionless and has a value greater than 0.

[0092] S304 calculates the equivalent sound leakage area of ​​the opening.

[0093] Ventilation openings are one of the main pathways for noise transmission from building envelopes. To quantify their sound leakage capacity, it is necessary to convert the geometric area and sound attenuation performance into an equivalent sound leakage area.

[0094] The layout boundary within the station determines the location and geometric area of ​​all ventilation openings. The effective ventilation opening ratio is determined by the design parameters of louvers or grilles. The effective ventilation area is obtained by multiplying the geometric opening area by the effective ventilation opening ratio. The area attenuation factor is obtained by dividing the insertion loss of the corresponding louver or sound-absorbing component by 10 and then taking the negative power of 10. The effective ventilation area is then multiplied by the area attenuation factor to obtain the equivalent sound leakage area of ​​a single opening, in square meters. The total equivalent sound leakage area of ​​all ventilation openings is obtained by summing the equivalent sound leakage areas of all openings.

[0095] Example: In the preliminary design of the converter station enclosure, the technicians measured the geometric area of ​​the four ventilation openings and consulted the louver insertion loss parameters. After calculation, the total equivalent sound leakage area of ​​the openings was found to be 12.5 square meters.

[0096] S305 calculates the equivalent sound-transmitting area of ​​the building envelope and outputs the equivalent sound leakage ratio of the openings.

[0097] The total equivalent sound leakage area of ​​the opening has been obtained. In order to achieve a comparison of the opening path and the panel path with the same diameter, it is necessary to convert the sound transmission capacity of the enclosure structure panel.

[0098] The building envelope is initially divided into panels, with the area of ​​each panel determined by the layout boundaries. The sound insulation of each panel is determined by the material and thickness design parameters. The sound transmission coefficient is obtained by dividing the panel's sound insulation by 10 and then taking the negative power of 10. The panel area is multiplied by the sound transmission coefficient to obtain the equivalent sound transmission area of ​​a single panel, measured in square meters. The equivalent sound transmission areas of all panels are summed to obtain the total equivalent sound transmission area of ​​the building envelope. The equivalent sound leakage area of ​​the total openings is divided by the total equivalent sound transmission area of ​​the building envelope to obtain the equivalent sound leakage ratio of the openings, which is dimensionless and takes a value greater than 0.

[0099] S306 comprehensive mapping yields the governance allocation coefficients.

[0100] The low-frequency propagation dominance ratio and the equivalent sound leakage ratio of the opening have been quantified as two major failure modes. In order to generate a unified indication of governance direction, it is necessary to perform comprehensive mapping through a probability discrimination model of historical engineering samples.

[0101] First, the low-frequency propagation dominance ratio and the equivalent leakage ratio of the opening are normalized to the same caliber and mapped to the interval of 0 to 1 to obtain the normalized low-frequency intensity and the normalized opening intensity. A historical engineering sample database collects past noise control cases of converter stations or substations, including measured low-frequency proportions, opening leakage ratios, final control methods, and effect verification data. Based on the probabilistic discriminant model trained on the historical engineering sample database, the normalized low-frequency intensity and the normalized opening intensity are input, and the low-frequency dominance probability and the opening dominance probability are output. Dividing the low-frequency dominance probability by the sum of the low-frequency dominance probability and the opening dominance probability yields the control allocation coefficient, which is dimensionless and ranges from 0 to 1.

[0102] In one embodiment, the historical engineering sample probabilistic discrimination model is constructed using a Bayesian network. First, several cases of converter stations or substations that have completed noise control are collected to form a historical engineering sample library. Each sample includes normalized low-frequency intensity and normalized aperture intensity as input features, as well as a label of the control mode confirmed by on-site acceptance or retesting.

[0103] Based on the initial connectivity relationships provided by domain knowledge, the network structure and parameters are learned using a sample library, and information criteria are introduced to constrain the structural complexity in order to avoid overfitting driven by small sample randomness.

[0104] During training, the samples are divided into training and validation sets. By selecting the feature discretization grouping method and prior parameters through cross-validation, the model maintains stable log-likelihood and classification consistency on different validation subsets, thereby ensuring the repeatability of governance allocation coefficient mapping.

[0105] During inference, the current normalized low-frequency intensity and normalized opening intensity are input, and the posterior probability of each governance mode is output. Based on this, the low-frequency dominance probability and opening dominance probability are obtained to participate in the calculation of the governance allocation coefficient.

[0106] S307 retrieves and outputs candidate combinations based on the governance allocation coefficient.

[0107] The governance allocation coefficient has characterized the relative intensity of failure modes. In order to output a containment scheme that matches the current working conditions, it is necessary to guide the search of the containment unit library based on its numerical range.

[0108] The process of determining the allocation coefficient adopts a preset dual-threshold comparison logic, introducing a first threshold and a second threshold. The first threshold is located in the lower part of the coefficient range, and the second threshold is located in the upper part of the coefficient range, with the first threshold being less than the second threshold.

[0109] The comparison and judgment order is as follows: First, the governance allocation coefficient is compared with the first threshold. If the governance allocation coefficient is not greater than the first threshold, it falls into the low value range, and the priority opening control strategy is selected. During the search, the constraint focus is on matching the insertion loss of the opening silencing module with the opening ratio. If the governance allocation coefficient is greater than the first threshold, it continues to be compared with the second threshold. If the governance allocation coefficient is not greater than the second threshold, it falls into the middle value range, and the balanced strategy is selected. During the search, the panel sound insulation and opening silencing configuration are balanced and constrained. If the governance allocation coefficient is greater than the second threshold, it falls into the high value range, and the priority low-frequency panel control strategy is selected. During the search, the constraint focus is on improving panel thickness, materials, and low-frequency sound insulation. The thresholds are determined by statistically analyzing the classification boundaries of different governance modes in the historical engineering sample database.

[0110] The building envelope unit library contains predefined panel modules with different thicknesses and materials corresponding to different sound insulation levels, as well as opening silencing modules with different insertion losses corresponding to different opening ratios. Based on the treatment allocation coefficient falling within the range, the corresponding search strategy and key constraints are selected, including the minimum ventilation area threshold. Candidate combinations of building envelope structures are matched and output from the building envelope unit library, and a matching opening silencing configuration is generated simultaneously.

[0111] Step S3 completes the item-by-item calculation of the low-frequency propagation dominance ratio and the equivalent sound leakage ratio of the opening, and the comprehensive mapping of the historical engineering sample probability discrimination model to form the governance allocation coefficient. This coefficient directly represents the relative dominance of the low-frequency path and the opening path under the current working conditions. Based on the governance allocation coefficient interval, the building envelope unit library is searched, and candidate combinations of building envelope structures and opening silencing configurations that are compatible with ventilation and heat dissipation constraints and site layout boundaries are output to ensure that the governance focus accurately matches the actual failure mechanism. This is directly output to step S4 for sound field and thermal ventilation verification to avoid the cancellation of noise reduction effects caused by the imbalance between panel control and opening control.

[0112] Step S3 has output candidate combinations of building envelope structures and opening silencing configurations that match the actual failure paths. These schemes are designed for operating conditions where low-frequency propagation is dominant or openings provide the primary sound leakage. However, the candidate schemes need to be verified for actual noise reduction effectiveness and thermal ventilation compatibility. Therefore, step S4 uses sound field simulation and thermal ventilation verification based on the equivalent sound power parameter set to select the optimal scheme that meets the station boundary noise target and heat dissipation constraints, and outputs the building envelope layout diagram and a list of node construction methods to ensure that the remediation scheme achieves stable compliance during project implementation.

[0113] S401 Loading Enclosure Candidate Combinations and Opening Noise Reduction Configurations.

[0114] Step S3 has output the candidate combinations of the building envelope and the matching opening silencing configurations. In order to perform dual verification in the unified model, these schemes need to be loaded into the simulation environment first to form a complete acoustic-thermal coupling model.

[0115] Each candidate combination of building envelope includes panel module geometry, material sound insulation spectrum, and panel placement. Opening silencing configurations include ventilation opening locations, louver or silencer insertion loss spectrum, and effective opening ratio. The equivalent sound power parameter set from step S2, the site layout boundary, and ventilation and heat dissipation constraints are loaded together. The model mesh uses a hybrid structured and unstructured mesh, and all data is stored with scheme identifiers for direct use in subsequent sound field and thermal ventilation calculations.

[0116] S402 performs sound field simulation verification.

[0117] After the candidate combinations of the building envelope and the noise reduction configuration of the openings are loaded, in order to quantify the actual noise reduction effect of each scheme, it is necessary to perform a station boundary sound field simulation based on the equivalent sound power parameter set.

[0118] Engineering calculation methods using geometric acoustics and statistical energy analysis, or acoustic simulation software such as finite element method and boundary element method, are employed. An equivalent sound power parameter set is input as the sound source. Frequency-dependent sound insulation boundary conditions are applied to the enclosure panels, and insertion loss and aperture ratio-related boundary conditions are applied to the ventilation openings. During calculation, the frequency band sound pressure contribution reaching representative points at each station boundary via the path through the enclosure panels and the path through the ventilation openings are calculated separately. The frequency band sound pressure level spectrum and A-weighted equivalent continuous sound level at each representative point are obtained by superimposing the sound energy, thus achieving a consistent verification of the noise reduction effect of the candidate schemes.

[0119] Example: In the noise control design of a ±800 kV converter station, the technicians loaded three candidate combinations of enclosure structures into the acoustic software in sequence. After running the simulation, the predicted A-weighted equivalent continuous sound level of each scheme at the representative point of the station boundary was obtained.

[0120] S403 performs thermal ventilation verification.

[0121] After the sound field simulation verification is completed, in order to confirm that the enclosure scheme has no adverse effect on the heat dissipation of the converter transformer and cooling system, thermal ventilation performance calculation needs to be performed simultaneously.

[0122] Computational fluid dynamics (CFD) methods or a ventilation balance model are employed. Inputs include the effective ventilation area of ​​the openings in the candidate combinations of the building envelope, information on internal heat sources, and environmental boundary conditions. The calculation process is as follows: For each ventilation opening, the average velocity is calculated using the selected ventilation model, taking into account the opening geometry and resistance characteristics. The ventilation volume of a single opening is then obtained by multiplying the effective ventilation area by the average velocity. The total ventilation volume is then summed across all openings, and the internal temperature field and surface temperatures of critical equipment are calculated using heat source information. Verification requires that the total ventilation volume is not less than the heat dissipation threshold, and that the equipment surface temperatures meet the design upper limit.

[0123] S404 comprehensive screening determines the final design results of the building envelope.

[0124] The results of acoustic field simulation and thermal ventilation verification have been obtained. In order to select the optimal scheme from multiple candidate combinations of building envelope, a multi-objective screening criterion needs to be applied.

[0125] The pre-set screening criteria first eliminate schemes where the predicted A-weighted equivalent continuous sound level at any representative point at the station boundary exceeds the environmental standard limit or the total ventilation volume is lower than the heat dissipation requirement threshold. For the remaining schemes, a comprehensive performance index is calculated. The calculation process is as follows: the difference between the target noise value at the station boundary and the average predicted A-weighted equivalent continuous sound level at all representative points is calculated as the noise performance component; simultaneously, the difference between the total ventilation volume and the minimum required ventilation volume is calculated as the ventilation performance component; these are then multiplied by the pre-set first and second weighting coefficients and summed to obtain the comprehensive performance index. The scheme with the highest comprehensive performance index is selected as the final building envelope design result; if multiple schemes have similar indices, priority is given based on structural constructability and ease of maintenance.

[0126] Example: In the noise control design of the converter station mentioned above, the technicians compared the verification results of the three sets of schemes. The second set of schemes had the best overall noise performance and ventilation performance and was determined as the final enclosure structure design result.

[0127] S405 outputs the layout diagram of the building envelope and a list of details.

[0128] Once the final design of the building envelope is determined, standardized drawings and a bill of quantities need to be generated to facilitate project implementation.

[0129] Based on the final 3D model, a general layout drawing of the building envelope is generated, including panel divisions, opening locations, maintenance access markings, elevation views, section views, and detailed drawings of key nodes. The node details list specifies the material specifications, thickness, sound insulation, and connection methods for each panel module, as well as the model of the opening silencing module, insertion loss curve, and installation requirements. All drawings and lists are output in CAD and PDF formats, with version identification and design specifications.

[0130] Step S4 completes the acoustic field simulation and thermal ventilation dual verification of the candidate combinations of the building envelope and the noise reduction configuration of the openings, forming the final building envelope design result that meets the noise control target of the station boundary and the heat dissipation requirements of the equipment. This result ensures that the noise reduction effect and thermal performance can be quantitatively verified through energy superposition acoustic field prediction and ventilation volume balance calculation, and directly outputs the layout drawings and node construction list required for engineering implementation, avoiding insufficient noise reduction or heat dissipation obstruction in actual application, and achieving a precise balance between noise control and engineering conditions.

[0131] Specifically, the above are merely preferred embodiments of this application and are not intended to limit this application.

[0132] Other preset parameters whose value range and acquisition logic are not explicitly specified in this invention can be pre-calibrated through offline simulation testing, or set to fixed values ​​according to on-site operating procedures.

[0133] In the description of this specification, references to terms such as "an embodiment," "example," and "specific example" indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0134] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific embodiments described above. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A design method for a sound insulation and noise reduction enclosure structure of a converter station based on artificial intelligence, characterized in that, Including the following steps: S1: In accordance with acoustic measurement specifications, acoustic measurement results are collected at the representative point at the station boundary and the equivalent continuous sound level at the representative point at the station boundary is calculated. At the same time, the operating status of the main equipment is recorded to form on-site input data. S2: Compare the on-site input data with the predicted sound level at the station boundary obtained by the existing noise prediction method using the same caliber to obtain the source term correction amount, and generate an equivalent sound power parameter set based on the source term correction amount; S3: The equivalent sound power parameter set is preprocessed with ventilation and heat dissipation constraints and the layout boundary within the station. Based on the measured frequency band results, the proportion of low-frequency relative sound energy is calculated. The propagation dominance correction factor is obtained by extracting the propagation attenuation difference using existing noise prediction methods. The proportion of low-frequency relative sound energy is corrected to obtain the low-frequency propagation dominance ratio. The effective ventilation area is calculated based on the geometric opening area of ​​the ventilation opening and the effective ventilation opening ratio. The insertion loss is converted into an area attenuation factor to obtain the total equivalent sound leakage area of ​​the opening. The sound transmission coefficient is converted based on the area of ​​the enclosure panel and the sound insulation of the panel to obtain the total sound transmission equivalent area of ​​the enclosure. The ratio of the total equivalent sound leakage area of ​​the opening to the total sound transmission equivalent area of ​​the enclosure is calculated to obtain the equivalent sound leakage ratio of the opening. The low-frequency propagation dominance ratio and the equivalent sound leakage ratio of the opening are normalized. The governance allocation coefficient is obtained by mapping through a pre-trained probability discrimination model. The retrieval strategy and constraint focus are selected based on the governance allocation coefficient interval. The candidate combination of the enclosure structure and the matching opening silencing configuration are output from the enclosure unit library. S4: Perform sound field simulation and thermal ventilation verification on the candidate combinations of the building envelope along with the opening silencing configuration, screen and determine the final building envelope design results, and output the building envelope layout diagram and node construction list.

2. The design method for a sound insulation and noise reduction enclosure structure of a converter station based on artificial intelligence according to claim 1, characterized in that, Step S1 includes: Multiple representative points were selected outside the station boundary. These representative points covered the direction in which the noise contribution of the converter transformer and cooling system was greatest and were evenly distributed around the station boundary. Integrating sound level meters and spectrum analyzers were used to continuously measure and collect the time sequence of frequency band sound pressure level with 1 / 3 octave band resolution and the corresponding A-weighted instantaneous sound pressure level.

3. The design method for a sound insulation and noise reduction enclosure structure of a converter station based on artificial intelligence according to claim 2, characterized in that, Step S1 includes: The time series data of A-weighted instantaneous sound pressure level at representative points at each station boundary and the time series data of sound pressure level in each 1 / 3 octave band are processed by sound energy averaging to obtain the A-weighted equivalent continuous sound level and the band equivalent sound pressure level.

4. The design method for a sound insulation and noise reduction enclosure structure of a converter station based on artificial intelligence according to claim 3, characterized in that, Step S2 includes: An outdoor sound propagation attenuation algorithm was used in conjunction with the initial sound power level spectrum to calculate the A-weighted predicted equivalent continuous sound level and the predicted sound pressure level of each 1 / 3 octave band representative point. Under the same station boundary representative point and frequency band resolution, the measured sound level and the predicted sound level were compared to obtain the A-weighted sound level difference and the sound level difference of each frequency band.

5. The design method for a sound insulation and noise reduction enclosure structure of a converter station based on artificial intelligence according to claim 4, characterized in that, Step S2 also includes: The A-weighted sound level difference and the sound level difference of each frequency band at all representative points of the station boundary are averaged to obtain the A-weighted source term correction and the source term correction of each frequency band. The corrections are then superimposed on the initial sound power level to generate the A-weighted equivalent sound power level and the equivalent sound power level of each frequency band, and integrated into a structured equivalent sound power parameter set with frequency band identification.

6. The design method for a sound insulation and noise reduction enclosure structure of a converter station based on artificial intelligence according to claim 1, characterized in that, Step S4 includes: Load candidate combinations of building envelope and opening noise reduction configurations into the simulation environment, input the equivalent sound power parameter set as the sound source, apply frequency-dependent sound insulation and insertion loss boundary conditions to perform sound field simulation verification, and input the effective ventilation area of ​​the opening and the power of the heat source in the station to perform thermal ventilation verification.

7. The design method for a sound insulation and noise reduction enclosure structure of a converter station based on artificial intelligence according to claim 6, characterized in that, Step S4 also includes: Based on the predicted A-weighted equivalent continuous sound level and total ventilation volume screening criteria, unqualified schemes were eliminated, and the comprehensive performance index of the remaining schemes was calculated to determine the final design result of the building envelope.

Images