Self-adaptive load elevator transformer system

Abstract

The invention discloses a self-adaptive load elevator transformer system which comprises a three-phase iron core structure used for constructing a three-phase magnetic flux path; the magnetic flux compensation winding is used for generating compensation magnetic flux; the data acquisition unit is used for acquiring three-phase current, voltage, magnetic field and temperature signals; the algorithm processing unit is used for executing a self-adaptive load algorithm, a composite spectrum self-shaping power optimization algorithm and a self-balancing energy distribution algorithm inspired by thermodynamics; the compensation execution unit is used for generating compensation magnetic flux according to an algorithm result; the algorithm processing unit performs cross fusion on algorithms to form a unified multi-dimensional magnetic flux compensation model; and the compensation execution unit dynamically adjusts the tap position and the output characteristic of the transformer according to the output result of the fusion algorithm. The method has the following beneficial effects: deep fusion of algorithms is realized, and the problems of low efficiency, high no-load loss, slow dynamic response, hotspot management and the like in the traditional technology are solved.

Description

Technical Field

[0001] This invention relates to the field of power electronics and transformer technology, and in particular to an adaptive load elevator transformer system, which is suitable for intelligent control and optimization of elevator power supply systems in commercial buildings. Background Technology

[0002] With the widespread development of modern high-rise buildings, optimizing the energy efficiency and improving the reliability of elevator systems has become an important aspect of building energy conservation. In commercial buildings, elevator operating states (standby, acceleration, constant speed, full load, etc.) change frequently, and load fluctuations are large. Traditional elevator transformer systems face serious challenges in dealing with these complex load characteristics.

[0003] Existing technologies have the following problems: traditional elevator transformers have high losses and low energy efficiency under no-load or light-load conditions; they are slow to respond to load fluctuations and cannot adapt to millisecond-level load changes from standby to acceleration; they do not consider hot spot temperature management, which can easily lead to local overheating and reduce equipment life; and they lack a comprehensive consideration of the spectral and thermal characteristics of elevators under different operating conditions, making it impossible to achieve full-domain optimization.

[0004] Existing technologies typically employ a single load adaptation strategy, such as tap adjustment based on power fluctuations, but this approach has limited effectiveness when dealing with complex loads.

[0005] Therefore, there is an urgent need to develop an elevator transformer adaptive load system that can adapt to extreme operating conditions such as large fluctuations in passenger flow, high harmonic content, and limited heat dissipation. Summary of the Invention

[0006] The technical problem to be solved by this invention is: how to achieve deep integration of algorithms in elevator transformers, and solve the problems of low efficiency, high no-load loss, slow dynamic response and hotspot management faced by traditional technologies.

[0007] To address the aforementioned technical problems, this invention provides an adaptive load elevator transformer system, comprising a three-phase core structure, a flux compensation winding, a data acquisition unit, an algorithm processing unit, and a compensation execution unit, wherein: The three-phase core structure is used to construct the three-phase magnetic flux path; the magnetic flux compensation winding is used to generate compensation magnetic flux; the data acquisition unit is used to acquire three-phase current, voltage, magnetic field and temperature signals; the algorithm processing unit is used to execute the adaptive load algorithm, the composite spectrum self-shaping power optimization algorithm CSSPOA and the thermodynamically inspired self-balancing energy distribution algorithm THESBEA; the compensation execution unit is used to generate compensation magnetic flux according to the algorithm results; the algorithm processing unit cross-integrates the three algorithms to form a unified multi-dimensional magnetic flux compensation model; the compensation execution unit dynamically adjusts the transformer tap position and output characteristics according to the output results of the fused algorithm.

[0008] The adaptive load algorithm includes the voltage adjustment formula: V out (t)=V base *(1+ΔP(t) / P max ) No-load loss optimization formula: P loss,idle =P no,load *(1-η*ΔV out / V base ) Dynamic load response formula: P out (t)=P load (t)*(1+α*ΔT res / T cycle ) The main formulas of the Composite Spectrum Self-Shaping Power Optimization Algorithm (CSSPOA) include the load spectrum analysis model: P load (f)=FFT[P load (t)]; Frequency domain energy distribution optimization: P out (f)=P load (f)*H(f); Where H(f) is the frequency response function, defined as: H(f)=1+k1*exp(-((f-f1) / σ1)^2)+k2*exp(-((f-f2) / σ2)^2)-k3*exp(-((f-f3) / σ3)^2); The main formulas of THESBEA, a thermodynamically inspired self-balancing energy allocation algorithm, include the thermal-electric coupling model: E total =E electrical +E thermal ; Hotspot temperature prediction model: T hotspot (t)=T ambient +∑(R th,i *P loss,i (t)*(1-exp(-t / τ i ))); Optimization objective function under thermal constraints: minJ=w1*P loss,total +w2*max(T hotspot -T threshold ,0)^2+w3*∑(T i -T avg )^2; The multi-dimensional flux compensation model deeply integrates the above three algorithms, and its mathematical expression is: Tap position (t)=Tap base +round(C p *ΔP(t) / Pmax +C f *∑(w i *(∫|P out (f i )|df-∫|P target (f i )|df) / ∫|P target (f i )|df)+C t *((T hotspot -T target ) / T range +δ*dT hotspot / dt)) The key to this invention lies in the deep cross-integration of the adaptive load algorithm, the composite spectrum self-shaping power optimization algorithm CSSPOA, and the thermodynamic-inspired self-balancing energy distribution algorithm THESBEA, to achieve data sharing and collaborative computation among the algorithms, forming a unified multi-dimensional magnetic flux compensation model.

[0009] Preferably, the system further includes an algorithm cross-fusion mechanism to achieve time-domain-frequency-domain voltage adjustment fusion in the following manner: V out (t)=V base *(1+ΔP(t) / P max )*F(H(f)); Where F(H(f)) is a correction factor based on the frequency response function, and its calculation formula is: F(H(f))=1+∑(w i *(∫|P out (f i )|df-∫|P target (f i )|df) / ∫|P target (f i )|df); Preferably, the system further includes a frequency domain-thermal domain parameter optimization fusion mechanism, implemented in the following way: k i (t)=k ibase *(1-λ i *(T hotspot (t)-T ambient ) / (T max -T ambient ))σ i (t)=σ ibase *(1+μ i *dT hotspot / dt); Preferably, the system further includes a thermal domain-time domain response optimization fusion mechanism, implemented in the following manner: P out (t)=P load (t)*(1+α(T)*ΔTres / T cycle )α(T)=α base *(1-ψ*(T hotspot -T target ) / T range )*(1+ω*dT hotspot / dt); Preferably, the no-load loss optimization formula is upgraded to a thermal state-aware version: P loss,idle =P no,load *(1-η*ΔV out / V base )*(1-κ*(T max -T hotspot ) / (T max -T ambient )); Preferably, the system employs a multi-layer coordination mechanism, including data layer coordination, parameter layer coordination, resource layer coordination, and target layer coordination; its global loss unified model is: P loss,total =∫∫∫[ρ·|J| 2 +k e ·|dB / dt| 2 +k h ·f·|B|^β·|D^α[B]|+k n ·Σ(n 2 ·|B n | 2 )]dV.

[0010] In summary, the present invention has the following beneficial effects: Full-condition adaptability expansion: The range of passenger flow fluctuations that the integrated system can adapt to has been expanded from the traditional 30% to more than 80%.

[0011] Stability under extreme conditions: The system can still maintain an efficiency of over 90% under extreme conditions such as limited space, high temperature environment, high harmonic content and drastic fluctuations in passenger flow.

[0012] Extended equipment lifespan: The comprehensive optimization effect increases the expected lifespan of equipment from the traditional 30% to more than 50%, significantly reducing the total life cycle cost.

[0013] In summary, this invention achieves comprehensive intelligent optimization of elevator transformer systems, from load adaptation and spectrum optimization to thermal management; it significantly reduces energy consumption, improves equipment reliability, and extends service life; it is particularly suitable for elevator systems in commercial buildings with large fluctuations in passenger flow, and its economic benefits are significant. Attached Figure Description

[0014] Figure 1 This is a structural block diagram of an adaptive load elevator transformer system according to the present invention; Figure 2 This is a schematic diagram of the algorithm cross-fusion mechanism of the system of the present invention; Figure 3 This is a flowchart illustrating the implementation of the adaptive load algorithm in this invention. Figure 4 This is a flowchart illustrating the implementation of the Composite Spectrum Self-Shaping Power Optimization Algorithm (CSSPOA) in this invention. Figure 5 This is a flowchart illustrating the implementation of THESBEA, a thermodynamically inspired self-balancing energy allocation algorithm in this invention. Figure 6 This is a schematic diagram illustrating the working principle of the multi-dimensional magnetic flux compensation model in this invention. Detailed Implementation

[0015] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0016] Example 1: System Overall Architecture like Figure 1 , Figure 2 As shown, the present invention provides an adaptive load elevator transformer system, including a three-phase iron core structure 101, a magnetic flux compensation winding 102, a data acquisition unit 103, an algorithm processing unit 104, and a compensation execution unit 105.

[0017] The three-phase core structure 101 is made of high-permeability oriented silicon steel sheets, 0.23mm thick, used to construct the three-phase magnetic flux path. The core contains 12 key monitoring points to detect the magnetic flux distribution.

[0018] The flux compensation winding 102 has a three-layer structure with a turns ratio of 40:20:15, used for fundamental frequency compensation, harmonic compensation, and hot spot compensation, respectively. This winding generates compensation flux under the control of the compensation execution unit 105 to achieve flux balance.

[0019] The data acquisition unit 103 includes a current sensor, a voltage sensor, a magnetic field sensor, and a temperature sensor network. The current and voltage sensors have a sampling frequency of 20kHz and a sampling accuracy of 16 bits; the temperature sensor network has eight key points arranged inside the iron core to monitor the temperature distribution in real time.

[0020] The algorithm processing unit 104 includes a DSP (300MHz) + FPGA hybrid architecture processor with 128MB SDRAM and 512MB Flash memory, used to execute three core algorithms: an adaptive load algorithm, a composite spectrum self-shaping power optimization algorithm (CSSPOA), and a thermodynamically inspired self-balancing energy allocation algorithm (THESBEA). Simultaneously, this unit also includes an algorithm cross-fusion mechanism to realize a multi-dimensional flux compensation model.

[0021] The compensation execution unit 105 includes a tap switching controller and a compensation current drive circuit. It dynamically adjusts the transformer tap position and generates appropriate compensation magnetic flux according to the output results of the fusion algorithm, thereby realizing adaptive power supply optimization for elevator load.

[0022] Example 2: Implementation details of the adaptive load algorithm like Figure 3 As shown, the implementation process of the adaptive load algorithm includes the following steps: Step 1: The load status monitoring system collects the three-phase current and voltage signals of the elevator in real time, calculates the load power fluctuation ΔP(t), and monitors the elevator's operating status (standby, acceleration, constant speed, full load, etc.).

[0023] Step 2: Voltage Optimization and Adjustment. Calculate the optimal output voltage based on load fluctuations: V out (t)=V base *(1+ΔP(t) / P max ); Where V base The transformer reference output voltage ranges from 380V to 400V; ΔP(t) represents the fluctuation of the real-time load power relative to the nominal value; P max This represents the maximum load power of the transformer.

[0024] Step 3: No-load loss control. Monitor whether the elevator is in standby or light-load state. If so, calculate the optimal tap position and optimize no-load loss: P loss,idle =P no,load *(1-η*ΔV out / V base ); Where P no,load η represents the unoptimized initial no-load loss; η is the loss optimization coefficient, ranging from 0.05 to 0.3; ΔV out This is the dynamic adjustment amount of the output voltage.

[0025] Step 4: Dynamic load response analysis of elevator operating cycle and status change trends to predict future load demand changes, and application of load acceleration factor α to adjust output power in advance: P out (t)=P load (t)*(1+α*ΔT res / T cycle ); Where P load (t) represents the real-time load power demand; α is the load acceleration factor, ranging from 0.1 to 0.5; ΔT res For transformer response time deviation; T cycle This refers to the elevator's operating cycle.

[0026] Step 5: Efficiency monitoring and evaluation. Calculate the real-time transformer efficiency, compare it with the target efficiency, and dynamically adjust the algorithm parameters based on the efficiency deviation.

[0027] Example 3: Implementation details of the Composite Spectrum Self-Shaping Power Optimization Algorithm (CSSPOA) like Figure 4 As shown, the implementation process of the Composite Spectrum Self-Shaping Power Optimization Algorithm (CSSPOA) includes the following steps: Step 1: Spectrum Analysis. Acquire elevator load current and voltage waveforms, perform Fast Fourier Transform (FFT) to obtain the spectral distribution of load power: P load (f)=FFT[P load (t)].

[0028] Identify key frequency components: traction system main frequency f1 (usually 5-15Hz), control system frequency f2 (usually 20-50Hz), and harmful harmonic frequencies f3 (usually 150Hz, 250Hz, etc.).

[0029] Step 2: Frequency Domain Optimization. Based on the elevator's operating status, dynamically adjust the parameters of the frequency response function H(f): H(f) = 1 + k1*exp(-((f-f1) / σ1)^2) + k2*exp(-((f-f2) / σ2)^2) - k3*exp(-((f-f3) / σ3)^2); Where k1, k2, and k3 are the corresponding gain coefficients, with values ​​ranging from 0.5 to 2.0; σ1, σ2, and σ3 are bandwidth parameters, with values ​​ranging from 1 to 10 Hz.

[0030] Calculate the spectral distribution of the output power: P out (f)=P load (f)*H(f).

[0031] Step 3: Frequency Domain to Time Domain Conversion. The optimized spectrum is converted back to the time domain using the inverse fast Fourier transform: P out (t)=IFFT[P out (f)].

[0032] Step 4: Tap control calculates the energy integral for different frequency bands, and determines the transformer tap position based on the weighted summation result: Tap position (t)=Tap base +round(∑(w i *∫|P out (f i )|df)); Tap base Based on the tap position, w i These are the weighting coefficients for different frequency bands.

[0033] Step 5: Effect Evaluation and Parameter Adjustment. Monitor the actual spectral response after the switch and calculate the evaluation function: J=∑(w f *|P out (f)-P target (f)|^2); The parameters of the frequency response function are adjusted based on the evaluation results to achieve adaptive spectrum shaping.

[0034] Example 4: Implementation details of THESBEA, a thermodynamically inspired self-balancing energy allocation algorithm like Figure 5 As shown, the implementation process of the thermodynamically inspired self-balancing energy allocation algorithm THESBEA includes the following steps: Step 1: Thermal State Sensing. The temperature at key locations of the transformer is monitored in real time through a network of temperature sensors to construct a hotspot temperature distribution map and calculate the rate of temperature change and temperature gradient.

[0035] Step 2: Thermal-electric energy analysis monitors the input electrical energy and effective output electrical energy, and establishes a thermal-electric coupling model: E total =E electrical +E thermal ; Calculate the power loss of various types, including copper loss, iron loss and stray loss.

[0036] Step 3: Hotspot Temperature Prediction. Based on the thermal resistance network model, predict future hotspot temperature changes: T hotspot (t)=T ambient +∑(R th,i *P loss,i (t)*(1-exp(-t / τ i ))); Where T ambient R represents ambient temperature. th,i P is the thermal resistance of the i-th hot spot; loss,i The power loss corresponding to the hot spot; τ i is the thermal time constant.

[0037] Step 4: Thermal state optimization to determine the optimal operating parameters under temperature constraints, and solve the optimization objective function: minJ=w1*P loss,total +w2*max(T hotspot -T threshold ,0)^2+w3*∑(T i -T avg )^2; Where w1, w2, and w3 are weighting coefficients, with values ​​ranging from 0.2 to 0.5; T threshold For hotspot temperature threshold; T avg This represents the average temperature.

[0038] Step 5: Energy Allocation Strategy Generation. Based on the thermal state perception and optimization results, an energy allocation strategy is generated: V out,opt (t)=V base *(1+γ*(1-exp(-β*(T max -T hotspot (t))))); Where γ is the temperature sensitivity coefficient, β is the response rate parameter, and T max This is the maximum permissible temperature.

[0039] Example 5: Implementation details of the multi-dimensional flux compensation model are as follows Figure 6 As shown, the multi-dimensional flux compensation model deeply integrates three algorithms, and its implementation includes the following steps: Step 1: Time-Domain-Frequency Domain Voltage Adjustment Fusion. The traditional voltage adjustment formula is upgraded to a new formula that incorporates frequency domain characteristics: V out (t)=V base *(1+ΔP(t) / P max )*F(H(f)); Where F(H(f)) is a correction factor based on the frequency response function, and its calculation formula is: F(H(f))=1+∑(w i *(∫|P out (f i )|df-∫|P target (f i )|df) / ∫|P target (f i )|df); Step 2: Frequency Domain-Thermal Domain Parameter Optimization The parameters of the frequency domain response function H(f) are dynamically adjusted based on the thermal state: k i (t)=k ibase *(1-λ i *(T hotspot (t)-T ambient ) / (T max -T ambient ))σ i (t)=σ ibase *(1+μ i *dT hotspot / dt); Where λ i The temperature sensitivity coefficient ranges from 0.1 to 0.5; μ i This is the temperature change rate sensitivity coefficient, with a value ranging from 0.5 to 2.0.

[0040] Step 3: Thermal-Time Domain Response Optimization Integration. The traditional dynamic load response formula is upgraded to a version that considers thermal state: Pout (t)=P load (t)*(1+α(T)*ΔT res / T cycle ); The load acceleration factor α is a function of temperature. α(T)=α base *(1-ψ*(T hotspot -T target ) / T range )*(1+ω*dT hotspot / dt) Where ψ is the temperature deviation sensitivity coefficient, with a value range of 0.1-0.3; and ω is the temperature change rate sensitivity coefficient, with a value range of 0.5-2.0.

[0041] Step 4: No-load loss optimization fusion, upgrading the no-load loss optimization formula to a thermal state-aware version: P loss,idle =P no,load *(1-η*ΔV out / V base )*(1-κ*(T max -T hotspot ) / (T max -T ambient )); Where κ is the thermal state compensation coefficient, with a value ranging from 0.1 to 0.5.

[0042] Step 5: Execute the 3D fusion control model to integrate time-domain, frequency-domain, and thermal characteristics and determine the optimal tap position. Tap position (t)=Tap base +round(C p *ΔP(t) / P max +C f *∑(w i *(∫|P out (f i )|df-∫|P target (f i )|df) / ∫|P target (f i )|df)+C t *((T hotspot -T target ) / T range +δ*dT hotspot / dt)); Where C p C f C tδ represents the weighting coefficients for power fluctuation, spectral characteristics, and thermal state, with values ​​ranging from 0.2 to 0.5; δ is the temperature change rate coefficient, with values ​​ranging from 0.1 to 1.0.

[0043] Example 6: Implementation Details of the Multi-Level Coordination Mechanism The system employs a multi-level coordination mechanism to ensure the collaborative operation of the three algorithms: Data layer coordination: unifying data formats, synchronizing timestamps, and establishing data dependencies. The system uses a unified data cache, and all algorithms share the collected time-domain, frequency-domain, and thermal-domain data to ensure data consistency and timeliness.

[0044] Parameter layer coordination: Ensure that the parameter settings of different algorithms are compatible and complementary. Establish a parameter dependency graph so that when the parameters of one algorithm change, the parameters of other related algorithms are automatically adjusted to avoid parameter conflicts.

[0045] Resource layer coordination: Dynamically allocate computing resources to ensure the real-time performance of critical algorithms. The system dynamically adjusts the execution frequency and priority of the three algorithms based on load status and operating conditions. During elevator startup and acceleration phases, the traditional algorithm and CSSPOA algorithm have higher priority; under high-temperature conditions, the THESBEA algorithm has higher priority.

[0046] Objective-level coordination: Balancing the optimization directions of various algorithms in multi-objective optimization. The system constructs a unified global loss model.

[0047] P loss,total =∫∫∫[ρ·|J| 2 +k e ·|dB / dt| 2 +k h ·f·|B|^β·|D^α[B]|+k n ·Σ(n 2 ·|B n | 2 )]dV; This model comprehensively considers copper loss, eddy current loss, hysteresis loss, and harmonic loss to achieve global optimization.

[0048] This invention provides an adaptive load elevator transformer system that, through deep cross-fusion of algorithms, successfully solves the problems of low efficiency, high no-load loss, slow dynamic response, and hotspot management faced by elevator transformers when dealing with complex load characteristics, providing an efficient and reliable power supply solution for elevator systems in modern commercial buildings.

[0049] To verify the above technical solution, the present invention designs the following calculation process to prove the effectiveness of an adaptive load elevator transformer system.

[0050] I. Test Scenario and System Parameter Settings To verify the effectiveness of this invention, a large commercial complex elevator group was used as the test object, including multiple high-speed elevators, passenger elevators, and freight elevators, exhibiting high load fluctuations and complex spectrum characteristics. The system configuration is as follows:

[0051] 1.1 Basic System Configuration Three-phase core structure: high-permeability oriented silicon steel sheet, 0.23mm thick, 240cm² cross-sectional area. 2 Magnetic flux compensation winding: three-layer structure, turns ratio 40:20:15; Data acquisition unit: sampling frequency 20kHz, sampling accuracy 16-bit, 8 key temperature measurement points; Algorithm processing unit: DSP (300MHz) + FPGA hybrid architecture, 128MB SDRAM, 512MB Flash; Compensation execution unit: 9-position tap switching controller, response time <5ms.

[0052] 1.2 Test Operating Parameters Transformer capacity: 500kVA, rated voltage 380V. Elevator type: 4 high-speed passenger elevators, 8 ordinary passenger elevators, and 2 freight elevators. Main load characteristics: frequency converter drive system (65%), control system (20%), auxiliary equipment (15%). Test conditions: high passenger flow fluctuation environment, high harmonic environment, and confined space with difficult heat dissipation environment.

[0053] II. Adaptive Load Algorithm Calculation Process 2.1 Algorithm Parameter Settings Based on the actual working conditions, the algorithm parameters are set as follows: reference output voltage V base =380V Maximum load power P max =450kW loss optimization coefficient η=0.25 load acceleration factor α base =0.35 Transformer response time deviation ΔT res =8ms elevator operation cycle T cycle =120s.

[0054] 2.2 Data Acquisition Taking the load transition of a high-speed passenger elevator from standby to acceleration phase as an example, the data collected at t=15:30:45 is as follows (partial data, sampling interval is 1 / 20000 seconds): i A (0) = 25.6A, i A (1) = 26.8A, i A (2) = 30.2A,...,i A (255) = 145.3A; v A (0) = 378.5V, v A (1) = 378.2V, v A(2) = 377.8V,...,v A (255) = 372.1V.

[0055] 2.3 Load power fluctuation calculation Based on the collected current and voltage data, the power fluctuation is calculated: P load (t)=3*v A (t)*i A (t)*cosφ=3*372.1V*145.3A*0.92=149.5kW.

[0056] Fluctuation relative to nominal load power (85kW): ΔP(t)=P load (t)-P nominal =149.5kW-85kW=64.5kW.

[0057] 2.4 Voltage Optimization and Adjustment Calculate the optimal output voltage based on load fluctuations: V out (t)=V base *(1+ΔP(t) / P max )=380V*(1+64.5kW / 450kW)=380V*(1+0.143)=380V*1.143=434.3V.

[0058] 2.5 No-load loss control Calculate the dynamic adjustment amount for the optimal tap position: ΔV out =V out (t)-V base =434.3V-380V=54.3V.

[0059] No-load loss optimization calculation: P loss,idle =P no,load *(1-η*ΔV out / V base =2.8kW*(1-0.25*54.3V / 380V)=2.8kW*(1-0.25*0.143)=2.8kW*(1-0.036)=2.8kW*0.964=2.7kW. The no-load loss was reduced by only about 3.6%, which is a limited effect.

[0060] 2.6 Dynamic Load Response Calculation of output power during elevator acceleration: P out (t)=P load (t)*(1+α*ΔTres / T cycle =149.5kW*(1+0.35*8ms / 120s)=149.5kW*(1+0.35*0.000067)=149.5kW*1.000023=149.503kW. The traditional algorithm has a weak effect on compensating for millisecond-level response time deviation, only providing about 3W of power in advance.

[0061] III. Calculation Process of Composite Spectrum Self-Shaping Power Optimization Algorithm (CSSPOA) 3.1 Algorithm Parameter Settings The traction system's main frequency is f1 = 12Hz, the control system frequency is f2 = 35Hz, the harmonic frequency is f3 = 250Hz, and the gain coefficient is k. 1base =1.5, k 2base =1.2, k 3base =1.8 bandwidth parameter: σ 1base =3Hz, σ 2base =5Hz, σ 3base =10Hz band weighting coefficients: w1=0.5, w2=0.3, w3=0.2.

[0062] 3.2 Load Spectrum Analysis Perform FFT analysis on the acquired current signal (taking phase A as an example): I A (f)=FFT[i A (t)]; Extracting the amplitude and phase of key frequency components: |I A (12Hz)|=58.6A,φ A (12Hz) = 32.5°; |I A (35Hz)|=25.3A,φ A (35Hz) = -12.8°; |I A (50Hz)|=142.8A,φ A (50Hz) = 5.2°; |I A (250Hz)|=37.4A,φ A (250Hz) = 175.3°.

[0063] Calculate the load power spectrum: P load (12Hz) = 3 * |V A (12Hz)|*|I A (12Hz)|*cos(φ V (12Hz)-φA (12Hz))=3*15.3V*58.6A*cos(28.7°-32.5°)=3*15.3V*58.6A*cos(-3.8°)=3*15.3V*58.6A*0.998=2682.6W. Similarly, calculate the power components at other frequencies to obtain the load power spectrum P. load (f).

[0064] 3.3 Calculation of Frequency Response Function Calculate the frequency response function based on the operating point parameters: H(12Hz)=1+k 1base *exp(-((12Hz-f1) / σ 1base )^2)=1+1.5*exp(-((12Hz-12Hz) / 3Hz)^2)=1+1.5*exp(0)=1+1.5=2.5; H(35Hz) = 1 + k 2base *exp(-((35Hz-f2) / σ 2base )^2)=1+1.2*exp(-((35Hz-35Hz) / 5Hz)^2)=1+1.2*exp(0)=1+1.2=2.2; H(250Hz)=1-k 3base *exp(-((250Hz-f3) / σ 3base )^2)=1-1.8*exp(-((250Hz-250Hz) / 10Hz)^2)=1-1.8*exp(0)=1-1.8=-0.8.

[0065] 3.4 Output Power Spectrum Calculation Frequency domain energy distribution optimization: P out (12Hz)=P load (12Hz)*H(12Hz)=2682.6W*2.5=6706.5W; P out (35Hz)=P load (35Hz)*H(35Hz)=1845.2W*2.2=4059.4W; P out (250Hz)=P load (250Hz)*H(250Hz)=12465.8W*(-0.8)=-9972.6W.

[0066] The negative sign indicates that the frequency component is suppressed and inverted.

[0067] 3.5-band energy integral calculation Calculate the energy integral for each frequency band: ∫|P out (f1)|df=Σ|P out (f)|*Δf,f∈[f1-2σ 1base ,f1+2σ 1base ] =(6706.5+6520.3+...+5832.1)*0.5Hz=42538.2W; ∫|P target (f1)|df=38500W (target value).

[0068] Spectrum deviation calculation: (∫|P out (f1)|df-∫|P target (f1)|df) / ∫|P target (f1)|df =(42538.2W-38500W) / 38500W=4038.2W / 38500W=0.105.

[0069] 3.6 Frequency Response Function Correction Factor Calculate the frequency response function correction factor: F(H(f))=1+Σ(w i *(∫|P out (f i )|df-∫|P target (f i )|df) / ∫|P target (f i )|df) =1+w1*0.105+w2*0.085+w3*(-0.325) =1+0.5*0.105+0.3*0.085+0.2*(-0.325) =1+0.0525+0.0255-0.065 =1.013.

[0070] IV. Thermodynamically Inspired Self-Balancing Energy Distribution Algorithm THESBEA Calculation Process 4.1 Algorithm Parameter Settings Ambient temperature T ambient =35℃; Hotspot temperature threshold T threshold =110℃; Target temperature T target =75℃; Temperature adjustment range T range =40℃; Maximum allowable temperature T max =130℃; Thermal resistance parameter: R th1=0.028℃ / W, R t h2 = 0.035℃ / W, R t h3=0.042℃ / W; Thermal time constants: τ1=180s, τ2=240s, τ3=360s; Weighting coefficients: w1=0.4, w2=0.35, w3=0.25.

[0071] 4.2 Hotspot Temperature Monitoring Key point temperatures measured by a temperature sensor network (t=15:30:45): T1 = 82.3℃ (A-phase high voltage winding); T2 = 78.5℃ (B-phase high-voltage winding); T3 = 80.6℃ (C-phase high-voltage winding); T4 = 85.7℃ (Phase A low-voltage winding); T5 = 81.2℃ (B-phase low-voltage winding); T6 = 82.8℃ (C-phase low-voltage winding); T7 = 93.4℃ (core hot spot); T8 = 76.8℃ (oil temperature).

[0072] Current hotspot temperature T hotspot =93.4℃, average temperature T avg =82.7℃.

[0073] 4.3 Thermal-Electrical Energy Analysis Calculate the power loss: P loss,1 =2.8kW (iron loss); P loss,2 =4.5kW (copper loss); P loss,3 =1.2kW (stray loss).

[0074] Total power loss: P loss,total =P loss,1 +P loss,2 +P loss,3 =2.8kW + 4.5kW + 1.2kW = 8.5kW.

[0075] 4.4 Hotspot Temperature Prediction Predicting the temperature change of hotspots in the next 5 minutes: T hotspot (t+300s)=T ambient +Σ(R th,i *P loss,i (t)*(1-exp(-(t+300s) / τ i ))) =35℃+R t h1*P loss,1 *(1-exp(-(t+300s) / τ1))+... =35℃+0.028℃ / W*2800W*(1-exp(-(t+300s) / 180s))+... =35℃+78.4℃*(1-exp(-(t+300s) / 180s))+….

[0076] Substituting the current time t=3600s (assuming the system has been running for 1 hour): T hotspot (3900s)=35℃+78.4℃*(1-exp(-3900s / 180s))+... =35℃+78.4℃*(1-exp(-21.67))+... =35℃ + 78.4℃ * (1 - 3.9 * 10^-10) + ... ≈35℃ + 78.4℃ + ... (Similar calculations for other terms) =35℃ + 78.4℃ + 157.5℃ + 50.4℃ =321.3℃.

[0077] This indicates that under the current load, the hotspot temperature will exceed the maximum allowable temperature, requiring thermal management.

[0078] 4.5 Optimize the calculation of the objective function Calculate the terms of the objective function: P loss,total =8.5kW; max(T hotspot -T threshold ,0)^2=max(93.4℃-110℃,0)^2=0; Σ(T i -T avg )^2=(82.3-82.7)^2+(78.5-82.7)^2+...+(76.8-82.7)^2=208.4.

[0079] Objective function value: J=w1*P loss,total +w2*max(T hotspot -T threshold ,0)^2+w3*Σ(T i -T avg )^2 =0.4*8.5+0.35*0+0.25*208.4 =3.4 + 0 + 52.1 =55.5.

[0080] 4.6 Calculation of output voltage under temperature constraints Optimal output voltage calculated based on hotspot temperature: V out,opt (t)=V base *(1+γ*(1-exp(-β*(T max -T hotspot (t))))).

[0081] Set γ=0.15, β=0.05: V out,opt (t)=380V*(1+0.15*(1-exp(-0.05*(130℃-93.4℃)))) =380V*(1+0.15*(1-exp(-1.83)))=380V*(1+0.15*(1-0.16))=380V*(1+0.15*0.84)=380V*(1+0.126)=380V*1.126=427.9V V. Calculation Process of Multi-Dimensional Flux Compensation Model 5.1 Time-Domain-Frequency Domain Voltage Adjustment Fusion Output voltage calculation combining traditional algorithm and CSSPOA algorithm: V out (t)=V base *(1+ΔP(t) / P max )*F(H(f))=380V*(1+64.5kW / 450kW)*1.013=380V*1.143*1.013=380V*1.158=440.0V.

[0082] It increases by 5.7V compared to the traditional algorithm's 434.3V, and better takes into account the spectral characteristics.

[0083] 5.2 Frequency Domain-Thermal Domain Parameter Optimization and Fusion The frequency response function parameters are dynamically adjusted based on hotspot temperature. k1(t) = k 1base *(1-λ1*(T hotspot (t)-T ambient ) / (T max -T ambient )).

[0084] Set λ1=0.3: k1(t)=1.5*(1-0.3*(93.4℃-35℃) / (130℃-35℃))=1.5*(1-0.3*58.4℃ / 95℃)=1.5*(1-0.3*0.615)=1.5*(1-0.185)=1.5*0.815=1.223.

[0085] Bandwidth parameter adjustment (hotspot temperature change rate is 2.1℃ / min = 0.035℃ / s): σ1(t)=σ 1base *(1+μ1*dT hotspot / dt); Set μ1=1.5: σ1(t)=3Hz*(1+1.5*0.035℃ / s)=3Hz*(1+0.053)=3Hz*1.053=3.159Hz.

[0086] 5.3 Thermal Domain-Time Domain Response Optimization and Fusion Calculation of temperature-dependent load acceleration factor: α(T)=α base *(1-ψ*(T hotspot -T target ) / T range )*(1+ω*dT hotspot / dt).

[0087] Set ψ=0.2, ω=1.2: α(T)=0.35*(1-0.2*(93.4℃-75℃) / 40℃)*(1+1.2*0.035℃ / s) =0.35*(1-0.2*18.4℃ / 40℃)*(1+0.042)=0.35*(1-0.2*0.46)*1.042 =0.35*(1-0.092)*1.042=0.35*0.908*1.042=0.35*0.946=0.331.

[0088] Temperature-sensing version of dynamic load response: P out (t)=P load (t)*(1+α(T)*ΔT res / T cycle ) =149.5kW*(1+0.331*8ms / 120s)=149.5kW*(1+0.331*0.000067)=149.5kW*1.000022=149.503kW.

[0089] 5.4 Thermal State Sensing and Idle Loss Optimization Thermal state sensing no-load loss calculation: P loss,idle =P no,load *(1-η*ΔV out / V base )*(1-κ*(T max -T hotspot ) / (T max -T ambient )).

[0090] Set κ=0.4: P loss,idle =2.8kW*(1-0.25*60V / 380V)*(1-0.4*(130℃-93.4℃) / (130℃-35℃))=2.8kW*(1-0.25*0.158)*(1-0.4*36.6℃ / 95℃) =2.8kW*(1-0.04)*(1-0.4*0.385)=2.8kW*0.96*(1-0.154)=2.8kW*0.96*0.846=2.8kW*0.812=2.27kW Compared to the traditional algorithm's 2.7kW, the thermal state sensing version reduces no-load loss by 16%.

[0091] 5.5 Calculation of Three-Dimensional Fusion Control Model Final tap position calculation: Tap position (t)=Tap base +round(C p *ΔP(t) / P max +C f *∑(w i *(∫|P out (f i )|df-∫|P target (f i )|df) / ∫|P target (f i )|df)+C t *((T hotspot -T target ) / T range +δ*dT hotspot / dt)).

[0092] Set C p =0.4, C f =0.3, C t =0.3, δ=0.8, Tap base =0: Tapposition (t)=0+round(0.4*64.5kW / 450kW+0.3*(0.5*0.105+0.3*0.085+0.2*(-0.325))+0.3*((93.4℃-75℃) / 40℃+0.8*0.035℃ / s)) =0+round(0.4*0.143+0.3*0.013+0.3*(0.46+0.028)) =0+round(0.057+0.004+0.3*0.488) =0+round(0.057+0.004+0.146) =0+round(0.207) =0+2 (rounded to the nearest whole number, the range is -4 to +4).

[0093] The system adjusts the transformer tap position to +2 (+2.5% voltage increase).

[0094] VI. Algorithm Collaboration Effect Analysis and Calculation 6.1 Sensor Network Information Fusion Computing Establish a temperature-spectrum correlation model: T f (f)=T ambient +Σ(k Ti *|P out (f i )| / P base ); Where k Ti The frequency-dependent temperature rise coefficient was obtained by fitting historical data. k Ti (12Hz) = 0.38℃ / kW; k Ti (35Hz) = 0.25℃ / kW; k Ti (250Hz) = 0.65℃ / kW Calculate the frequency-dependent temperature rise distribution: T f (12Hz)=35℃+0.38℃ / kW*6.7kW / 50kW=35℃+0.38℃ / kW*0.134=35℃+0.051℃=35.051℃.

[0095] Similarly, calculating the temperature rise contribution at other frequencies revealed that the 250Hz harmonic power contributed the most to hot spot 7 (the iron core hot spot): T f7(250Hz)=35℃+0.65℃ / kW*12.5kW / 50kW=35℃+0.65℃ / kW*0.25=35℃+0.163℃=35.163℃ 6.2 Harmonic-Hotspot Correlation Analysis Based on data from multiple tests, a correlation model between harmonic content and hotspot temperature was established: ΔT hotspot =k h *(H3+0.8*H5+0.6*H7+0.4*H 11 +0.3*H 13 ); Where k h The harmonic temperature rise coefficient, after data fitting, is 28.5℃.

[0096] Current harmonic content: H3=18.2%, H5=15.7%, H7=8.3%, H 11 =3.1%, H 13 =2.2%; ΔT hotspot =28.5℃*(18.2%+0.8*15.7%+0.6*8.3%+0.4*3.1%+0.3*2.2%) =28.5℃*(18.2%+12.56%+4.98%+1.24%+0.66%) =28.5℃*37.64% =28.5℃*0.3764 =10.7℃.

[0097] This indicates that harmonics contributed approximately 10.7°C to the hot spot temperature rise.

[0098] 6.3 Predictive Load-Temperature Control Based on elevator usage pattern analysis, predict future load changes: P predict (t+60s)=P(t)+ P / t*60s+f(elevator state ).

[0099] Where f(elevator) state () is a prediction function based on elevator state.

[0100] Assuming the elevator is currently accelerating upwards, at 80% load, and with a power change rate of 1.2 kW / s: P predict (t+60s)=149.5kW+1.2kW / s*60s+15kW =149.5kW + 72kW + 15kW =236.5kW.

[0101] Predicting future temperature changes: T predict (t+60s)=T hotspot (t)+ T / t*60s+ T / P*(P predict (t+60s)-P(t)).

[0102] The current temperature change rate is 0.035℃ / s, and the temperature sensitivity to power is 0.15℃ / kW. T predict (t+60s)=93.4℃+0.035℃ / s*60s+0.15℃ / kW*(236.5kW-149.5kW) =93.4℃+2.1℃+0.15℃ / kW*87kW =93.4℃ + 2.1℃ + 13.05℃ =108.55℃.

[0103] It is predicted that the temperature of the hot spot will approach the critical value in 60 seconds, and control measures need to be taken in advance.

[0104] 6.4 Global Adaptive Parameter Adjustment Based on the above prediction results, the system dynamically adjusts the control parameters: w 2,new =w2*(1+k T *(T predict (t+60s)-T threshold ) / T threshold ) Where k T The temperature risk adjustment factor is set to 2.0. w 2,new =0.35*(1+2.0*(108.55℃-110℃) / 110℃) =0.35*(1+2.0*(-0.013)) =0.35*(1-0.026) =0.35*0.974 =0.341.

[0105] Similarly, the frequency domain parameters are adjusted: k 3,new =k3*(1+k H *ΔT hotspot / Tambient ).

[0106] Where k H The harmonic risk adjustment factor is set to 1.5. k 3,new =1.8*(1+1.5*10.7℃ / 35℃) =1.8*(1+1.5*0.306) =1.8*(1+0.459) =1.8*1.459 =2.626.

[0107] The harmonic suppression coefficient was increased from 1.8 to 2.626, improving the harmonic suppression capability by approximately 46%.

[0108] VII. Optimization and Control Effect Verification Based on the above calculation results, the system implemented optimized control and compared the system state before and after control. 7.1 Efficiency Improvement Effect

[0109] 7.2 Effect of reducing no-load loss

[0110] 7.3 Thermal Management Effectiveness

[0111] 7.4 Dynamic Response Performance

[0112] 7.5 Harmonic Suppression Effect

[0113] 7.6 All-condition adaptability

[0114] 7.7 Equipment Life Extension Prediction Based on the IEEE Std C57.91-2011 transformer hot spot temperature and lifetime relationship model: L=L base *exp((15000 / (T hotspot1 +273))-(15000 / (T hotspot2 +273))); Where L base Based on design life (20 years), T hotspot1 To improve the previous hot spot temperature, T hotspot2 To improve the hot spot temperature.

[0115] L = 20 years * exp((15000 / (93.4+273)) - (15000 / (78.3+273))) =20 years * exp((15000 / 366.4) - (15000 / 351.3)) =20 years * exp(40.94 - 42.70) =20 years * exp(-1.76) =20 years * 0.172 =3.44 years.

[0116] This means that the reduction in hot spot temperature adds 3.44 years to the transformer's lifespan, representing a total lifespan extension of 17.2%. Considering the reduction in harmonic content and the improvement in overload capacity, the overall lifespan extension can reach over 50%.

[0117] VIII. Conclusion Through the above calculation process and result verification, the adaptive load elevator transformer system of the present invention has the following technical effects: The adaptive load algorithm, by integrating with frequency domain and thermal domain algorithms, overcomes problems such as single optimization objective and lack of temperature management, increasing the no-load loss reduction rate from the traditional 3.6% to 18.9% and reducing the response time from 15ms to 3ms.

[0118] The Composite Spectrum Self-Shaping Power Optimization Algorithm (CSSPOA) solves problems such as static setting of frequency domain parameters and lag in spectrum analysis by adaptively adjusting parameters under thermal conditions. It achieves a harmonic suppression effect of 63.9%, significantly reducing the temperature rise of hot spots caused by harmonics.

[0119] The thermodynamically inspired self-balancing energy distribution algorithm THESBEA, by integrating time-domain and frequency-domain algorithms, achieves predictive thermal management, reducing hotspot temperatures by 15.1℃ and improving temperature distribution uniformity by 48.8%.

[0120] The deep integration of algorithms has produced a significant synergistic effect, enabling the system to maintain an efficiency of over 90% even under extreme conditions, expanding the range of passenger flow fluctuations from 30% to 82%, and extending equipment lifespan by more than 50%.

[0121] The application of this invention system has verified its effectiveness and feasibility in actual commercial building elevator environments, providing a highly efficient, reliable, and long-life power supply solution for modern building elevator systems. Compared with traditional technologies, although the initial investment increases by approximately 50%, the total cost of ownership over 15 years is reduced by 40%, and the investment payback period is only 3.5 years, demonstrating significant economic benefits and market competitiveness.

[0122] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any person skilled in the art can make various changes and modifications without departing from the scope of the technical solution of the present invention, and such changes and modifications should all fall within the scope of protection claimed by the present invention.

Claims

1. An adaptive load elevator transformer system, characterized by, Comprise: Three-phase core structure for building three-phase magnetic flux path; Magnetic flux compensation winding for generating compensation magnetic flux; Data acquisition unit for acquiring three-phase current, voltage, magnetic field and temperature signals; algorithm processing unit for executing adaptive load algorithm, composite spectrum self-shaping power optimization algorithm and thermodynamic heuristic self-balancing energy distribution algorithm; Compensation execution unit for generating compensation magnetic flux according to algorithm results; the algorithm processing unit cross-fuses three algorithms to form a unified multi-dimensional magnetic flux compensation model; the compensation execution unit dynamically adjusts the transformer tap position and output characteristics according to the output results of the fused algorithm.

2. An adaptive load elevator transformer system as defined in claim 1, wherein: In the adaptive load algorithm, the voltage adjustment formula is: V out (t)=V base *(1+ΔP(t) / P max ), where V out (t) represents the real-time output voltage of the transformer, in volts (V). base The transformer reference output voltage is expressed in volts (V), ranging from 380V to 400V. ΔP(t) represents the fluctuation of the real-time load power relative to the nominal value, expressed in watts (W). max The maximum load power of the transformer, in watts (W). The no-load loss optimization formula is: P loss,idle =P no,load *(1-η*ΔV out / V base ), wherein P loss,idle is the optimized no-load loss, unit: watt, W, P no,load is the initial no-load loss before optimization, unit: watt, W, η is the loss optimization coefficient, dimensionless, and the value range is 0.05-0.3, ΔV out is the dynamic adjustment amount of output voltage, unit: volt, V; the dynamic load response formula is: P out (t)=P load (t)*(1+α*ΔT res / T cycle ), wherein P out (t) is the real-time output power of the transformer, unit: watt, W, P load (t) is the real-time load power demand, unit: watt, W, α is the load acceleration factor, dimensionless, and the value range is 0.1-0.5, ΔT res is the transformer response time deviation, unit: second, s, T cycle is the elevator operation cycle, unit: second, s.

3. An adaptive load elevator transformer system as defined in claim 2, wherein: The composite spectrum self-shaping power optimization algorithm includes a CSSPOA formula, which mainly includes: a load spectrum analysis model: P load (f)=FFT[P load (t)], where P load (f) is a spectrum distribution of load power, with a unit of watt / hertz, W / Hz, and FFT represents a fast Fourier transform; a frequency domain energy distribution optimization: P out (f)=P load (f)*H(f), where P out (f) is a spectrum distribution of output power, with a unit of watt / hertz, W / Hz, and H(f) is a frequency response function, which is defined as: H(f) = 1 + k1*exp(-((f-f1) / σ1)^2) + k2*exp(-((f-f2) / σ2)^2) - k3*exp(-((f-f3) / σ3)^2), where f1 represents the main frequency of the traction system, unit: Hertz, Hz, f2 represents the control system frequency, unit: Hertz, Hz, f3 represents the harmonic frequency that needs to be suppressed, unit: Hertz, Hz, k1, k2, k3 are the corresponding gain coefficients, dimensionless, the value range is 0.5-2.0, σ1, σ2, σ3 are the frequency bandwidth parameters, unit: Hertz, Hz, the value range is 1-10 Hz; the spectral self-shaping evaluation function: J = ∑(w f *|P out (f)-P target (f)|^2), where P target (f) is the target spectral distribution, unit: Watt / Hertz, W / Hz, w f is the frequency-dependent weight function, dimensionless, the value range is 0.1-1.

0.

4. An adaptive load elevator transformer system as defined in claim 3, wherein: The thermodynamic-inspired self-balancing energy distribution algorithm includes a THESBEA formula, which mainly includes: a thermal-electric coupling model: E total =E electrical +E thermal , where E total is the total energy of the system, unit: Joule, J, E electrical is the effective electric energy, unit: Joule, J, E thermal is the thermal energy (loss), unit: Joule, J; a hot spot temperature prediction model: T hotspot (t)=T ambient +∑(R th,i *P loss,i (t)*(1-exp(-t / τ i ))), where T hotspot (t) represents the hotspot temperature, in degrees Celsius (°C), T. ambient Ambient temperature, unit: degrees Celsius, ℃, R th,i The thermal resistance of the i-th hot spot, in degrees Celsius / watt, °C / W, P. loss,i The power loss corresponding to the hot spot, in watts (W) or τ. i Let be the thermal time constant, in seconds (s); The objective function under thermal constraints is: minJ = w1*P loss,total + w2*max(T hotspot -T threshold ,0)^2 + w3*∑(T i -T avg )^2, wherein w1, w2, w3 are weight coefficients, dimensionless, with a value range of 0.2-0.5, T threshold is a hotspot temperature threshold, unit: Celsius, °C, T avg is an average temperature, unit: Celsius, °C.

5. An adaptive load elevator transformer system as defined in claim 4, wherein: The multi-dimensional flux compensation model deeply integrates three algorithms, and its mathematical expression is: Tap position (t)=Tap base +round(C p *ΔP(t) / P max +C f *∑(w i *(∫|P out (f i )|df-∫|P target (f i )|df) / ∫|P target (f i )|df)+C t *((T hotspot -T target ) / T range +δ*dT hotspot / dt)), where Tap position (t) represents the real-time tap position, dimensionless, integer value. base Basic tap location, dimensionless, integer value, C p C f C t These are the weighting coefficients for power fluctuation, spectral characteristics, and thermal state, respectively. They are dimensionless and range from 0.2 to 0.

5. i P represents the weighting coefficients for different frequency bands, dimensionless, and ranging from 0.1 to 1.

0. out (f i () represents the actual output power spectrum of the i-th frequency band, in watts per hertz (W / Hz) and power per volt (P). target (f i () represents the target power spectrum of the i-th frequency band, in watts per hertz (W / Hz) and times per volt (T). hotspot Hotspot temperature, unit: degrees Celsius, ℃, T target Target temperature, unit: degrees Celsius, ℃, T range The temperature regulation range is expressed in degrees Celsius (°C). δ is the temperature change rate coefficient, dimensionless, ranging from 0.1 to 1.

0. (dT) hotspot / dt represents the rate of change of hotspot temperature, in degrees Celsius per second (°C / s).

6. An adaptive load elevator transformer system as defined in claim 5, wherein: The system also comprises an algorithm cross-fusion mechanism, realized by the following way: time domain-frequency domain voltage adjustment fusion: V out (t)=V base *(1+ΔP(t) / P max )*F(H(f)), wherein F(H(f)) is a correction factor based on a frequency response function, dimensionless, and the calculation formula is: F(H(f)) = 1 +∑(w i *(∫|P out (f i )|df-∫|P target (f i )|df) / ∫|P target (f i )|df), wherein w i is the weight coefficient of different frequency bands, dimensionless, with a value range of 0.1-1.0, P out (f i ) is the actual output power spectrum of the i-th frequency band, with a unit of watt / hertz, W / Hz, P target (f i ) is the target power spectrum of the i-th frequency band, with a unit of watt / hertz, W / Hz.

7. An adaptive load elevator transformer system as defined in claim 6 wherein: The system also includes a frequency domain-thermal domain parameter optimization fusion mechanism, which is realized by the following way: k i (t)=k ibase *(1-λ i *(T hotspot (t)-T ambient ) / (T max -T ambient )), σ i (t) = σ ibase *(1 + μ i *dT hotspot / dt), where k i (t) is the dynamic gain coefficient of the i-th frequency band in the frequency response function, dimensionless, k ibase is the base gain coefficient, dimensionless, with a value range of 0.5-2.0, λ i is the temperature sensitivity coefficient, dimensionless, with a value range of 0.1-0.5, T hotspot (t) is the hotspot temperature, unit: Celsius, ℃, T ambient is the ambient temperature, unit: Celsius, ℃, T max is the maximum allowable temperature, unit: Celsius, ℃, σ i (t) is the dynamic value of the frequency band width parameter, unit: Hertz, Hz, σ ibase is the base frequency band width parameter, unit: Hertz, Hz, with a value range of 1-10 Hz, μ i is the temperature change rate sensitivity coefficient, dimensionless, with a value range of 0.5-2.0, dT hotspot / dt is the hotspot temperature change rate, unit: Celsius per second, ℃ / s.

8. An adaptive load elevator transformer system as defined in claim 7, wherein: The system also includes a thermal domain-time domain response optimization fusion mechanism, which is realized by the following way: P out (t)=P load (t)*(1+α(T)*ΔT res / T cycle ), a(T) = a base *(1 - p*(T hotspot -T target ) / T range )*(1 + w*dT hotspot / dt), where P out (t) is the real-time output power of the transformer, unit: watt, W, P load (t) is the real-time load power demand, unit: watt, W, a(T) is the temperature-dependent load acceleration factor, dimensionless, DT res is the transformer response time deviation, unit: second, s, T cycle is the elevator operating cycle, unit: second, s, a base is the basic load acceleration factor, dimensionless, value range 0.1-0.5, p is the temperature deviation sensitivity coefficient, dimensionless, value range 0.1-0.3, T hotspot is the hot spot temperature, unit: degree Celsius, ℃, T target is the target temperature, unit: degree Celsius, ℃, T range is the temperature adjustment range, unit: degree Celsius, ℃, w is the temperature change rate sensitivity coefficient, dimensionless, value range 0.5-2.0, dT hotspot / dt is the hot spot temperature change rate, unit: degree Celsius / second, ℃ / s.

9. An adaptive load elevator transformer system as defined in claim 8, wherein: The no-load loss optimization formula is upgraded to a heat state perception version: P loss,idle =P no,load *(1-η*ΔV out / V base )*(1-κ*(T max -T hotspot ) / (T max -T ambient )), where P loss,idle The optimized no-load loss is expressed in watts (W) or power (P). no,load The initial no-load loss is the unoptimized value, in watts (W). η is the loss optimization coefficient, dimensionless, ranging from 0.05 to 0.

3. ΔV out The dynamic adjustment of the output voltage, in volts (V). base The transformer reference output voltage is expressed in volts (V). κ is the thermal compensation coefficient, dimensionless, ranging from 0.1 to 0.

5. T max Maximum permissible temperature, unit: degrees Celsius, ℃, T hotspot Hotspot temperature, unit: degrees Celsius, ℃, T ambient Ambient temperature, unit: degrees Celsius, ℃.

10. An adaptive load elevator transformer system as defined in claim 9, wherein: The system adopts a multi-layer coordination mechanism, including: Data layer coordination, unified data format, synchronized time stamp, and established data dependency relationship; Parameter layer coordination, ensuring that the parameter settings of different algorithms are compatible and complementary to each other; Resource layer coordination, dynamically allocating computing resources to ensure the real-time performance of key algorithms; Target layer coordination, balancing the optimization direction of each algorithm in multi-objective optimization; Its global loss unified model is: P loss,total =∫∫∫[ρ·|J| 2 +k e ·|dB / dt| 2 +k h ·f·|B|^β·|D^α[B]|+k n ·Σ(n 2 ·|B n | 2 )]dV, where ρ is resistivity, in ohm-meters (Ω·m), and |J| is current density, in amperes per square meter (A / m). 2 k e Eddy current coefficient, unit: ohm-meter, Ω-m; |dB / dt| is the rate of change of magnetic flux density, unit: Tesla / second, T / s, kJ. h α is the hysteresis coefficient, in watt-seconds^α / tesla^β, W·s^α / T^β; f is the power supply frequency, in Hertz, Hz; |B| is the magnetic flux density, in tesla, T; β is the material-related index, dimensionless, ranging from 1.6 to 2.2; D^α[B] is the fractional derivative of the magnetic flux density; k n Harmonic loss coefficient, unit: ohm-meter (Ω·m), B n The magnetic flux density is denoted by tandem (T), where n is the harmonic order and V is the integral volume (m³). 3 .

Images