Adaptive cut-off control of on-load tap changer based on current zero-crossing prediction

Abstract

An adaptive disconnection control on-load tap changer based on current zero-crossing prediction includes a current sensing unit located on the incoming side of the on-load tap changer, configured to collect real-time current waveform data flowing through the moving contact of the switch; and a voltage sensing unit configured to collect system voltage waveform data. This invention, through high-frequency current sampling, hardware-level fundamental frequency extraction and zero-crossing prediction, combined with precise calibration and timing compensation of the action delay of the digital actuator, successfully locks the contact separation moment within a preset microsecond-level window near the natural current zero-crossing point. This minimizes the instantaneous current value and arc energy at the moment of disconnection, effectively suppressing contact erosion and insulating oil degradation, and improving the electrical life and operational reliability of the on-load tap changer.

Description

Technical Field

[0001] This invention relates to the field of voltage regulating transformer equipment technology, and more specifically to an adaptive on-load tap changer based on current zero-crossing point prediction and its disconnection control method. Background Technology

[0002] Existing on-load tap changers typically consist of two core components: the switching switch and the tap selector. During the operation of the switching switch, a bridging phase and a subsequent current interruption phase are inevitable. To extinguish the arc generated at the moment of interruption, traditional solutions often employ a spring-energy-storing operating mechanism to drive the contacts to rapidly separate in the insulating oil, utilizing the impact and cooling effect of the oil flow to extinguish the arc, or using a vacuum interrupter to enclose the arc within a vacuum bubble. Ideally, the AC current interruption moment should be near the natural zero-crossing point of the current, because at this point, the electromagnetic energy stored in the circuit is minimal, and the arc is most easily extinguished. In existing on-load tap changers, the mechanism driving the switching switch is typically a spring-energy-storing operating mechanism. An electric motor stores energy in the spring; when a tap change command is received, the locking device unlocks, and the spring instantaneously releases its energy, ultimately driving the moving contact through a complex mechanical transmission chain including connecting rods, crank arms, and rotating shafts. The action delay of this mechanical system is significantly affected by various factors such as ambient temperature, oil viscosity, mechanical structure wear and aging, and load randomness, making it difficult to actively control the precise moment of contact separation. The contact separation moment and the phase of the AC current waveform flowing through the contacts have a random coupling relationship. If the contacts separate precisely near the current peak, a high-energy transient arc will ignite between them. This high-energy arc not only severely erodes the contact material and degrades the insulating oil, but also generates a large number of metal particles and carbides, severely shortening the electrical life and maintenance cycle of the on-load tap changer, and even causing a safety accident due to switch failure. Summary of the Invention

[0003] To address the technical problems existing in the prior art, as a first aspect of the present invention, an adaptive disconnection control on-load tap changer based on current zero-crossing point prediction is proposed, comprising: The multi-source sensing unit is used to collect real-time operating status data of the on-load tap changer, including: The current sensing unit is located on the incoming side of the on-load tap changer and is configured to collect real-time current waveform data i(t) flowing through the moving contact of the switching switch. The voltage sensing unit is configured to acquire system voltage waveform data u(t); The displacement sensing unit, mechanically coupled to the operating mechanism of the moving contact, is configured to monitor and output displacement data representing the actual motion trajectory of the moving contact in real time. A digital actuator, connected to the moving contact of a switching switch, is used to receive control commands and drive the moving contact to perform separation or engagement actions; The controller, communicatively connected to the multi-source sensing unit and the digital actuator, is configured to perform calculations based on the operating status data collected by the multi-source sensing unit to determine the predicted ideal natural zero-crossing time T0, the load power factor, and the calibrated actual action delay T of the digital actuator. 延时 Based on the ideal natural zero-crossing time T0, the load power factor, and the calibrated actual action delay T... 延时 Compensation is performed, and trigger control commands are generated and output to the digital actuator so that the moving contact can complete separation within the preset target opening window; The target switching window is dynamically determined based on the load power factor and is defined within a microsecond range before the ideal natural zero-crossing point T0.

[0004] Preferably, the current sensing unit is a flexible Rogowski coil, and its output signal is processed by a high-frequency integrator to provide real-time current waveform data i(t) with a bandwidth of not less than 1MHz.

[0005] Preferably, the digital actuator is a permanent magnet operating mechanism, which has a constant action delay characteristic that is less affected by environmental factors compared to spring energy storage mechanisms.

[0006] Preferably, the controller is a heterogeneous computing architecture that includes a field-programmable gate array module and an advanced reduced instruction set computer processor module; Among them, the field-programmable gate array module is configured to implement a fundamental wave extraction algorithm based on a second-order generalized integrator in hardware to lock the fundamental wave phase of the current in real time, and the advanced reduced instruction set computer processor module is configured to execute power factor calculation and online identification and correction algorithms for drive delay.

[0007] As a second aspect of the present invention, an adaptive disconnection control method for on-load tap changers based on current zero-crossing point prediction is also proposed. The adaptive disconnection control method for on-load tap changers includes the following steps: Step S1: Control the digital actuator to perform the full stroke action, and use the displacement sensing unit to record the actual separation time of the moving contact, so as to establish a mechanical motion delay reference parameter library T related to ambient temperature and / or the number of actions. 延时库 ; Step S2: When a gear shifting command is received, execute the following sub-steps: Step S21: Acquire and analyze the real-time current waveform data i(t) and system voltage waveform data u(t) of the current and historical cycles to calculate the total harmonic distortion rate of the current and the power factor angle of the load; Step S22: Based on the calculated power factor angle, dynamically determine a target lead ΔT. The target lead ΔT is defined as the target separation time T of the moving contact. 目标 The time lead of T relative to the ideal natural zero-crossing time T0, i.e., T 目标 =T0-ΔT; Step S3: Lock in the ideal natural zero-crossing time T0 of the fundamental frequency of the next current to be interrupted, and retrieve the reference parameter library T. 延时库 The search function retrieves the mechanism action delay T corresponding to the current operating condition. 延时 And calculate the trigger time T 触发 = T0-ΔT-T 延时 At the trigger time T 触发 Send a trigger pulse to the digital actuator; Step S4: Record and identify the actual moment T of the moving contact's rigid separation during this action. 实际 Calculate the timing deviation δ=T 实际 -(T0-ΔT), and use an adaptive filtering algorithm to take the timing deviation δ as feedback, to adjust the reference parameter library T. 延时库 The corresponding mechanism action delay T 延时 Perform online corrections and updates.

[0008] Preferably, in step S22: If the controller determines that the load is an inductive load, the target lead time ΔT is set to the first preset value; If the controller determines that the load is a resistive load, the target lead time ΔT is set to the second preset value; The first preset value is greater than the second preset value.

[0009] Preferably, in step S22: If the controller determines that the total harmonic distortion rate of the current exceeds the preset arc stability prediction threshold, it will use the zero-crossing point of the system voltage waveform data u(t) as an auxiliary reference to calculate the pseudo-current zero-crossing point as an alternative reference for the ideal natural zero-crossing point time T0.

[0010] Preferably, if a stable predicted value cannot be obtained through the voltage zero-crossing point assistance method, the controller automatically switches to the random phase switching mode and outputs a synchronous failure alarm signal.

[0011] Preferably, the adaptive filtering algorithm used in step S4 is the recursive least squares algorithm.

[0012] Preferably, step S5 is also included: Collect the arc voltage and arc current at the moment of switching, and calculate the arc energy E during this interruption process; and If the single or cumulative arc energy E exceeds the preset safety threshold, it is determined that the drive phase is out of sync, and subsequent gear shifting operations are blocked.

[0013] Compared with the prior art, the advantages of the present invention are as follows: This invention, through high-frequency current sampling, hardware-level fundamental frequency extraction and zero-crossing prediction, combined with precise calibration and timing compensation of the action delay of the digital actuator, successfully locks the contact separation moment within a preset microsecond-level window near the natural zero-crossing point of the current. This minimizes the instantaneous current value and arc energy at the moment of disconnection, effectively suppressing contact erosion and insulating oil degradation, and improving the electrical life and operational reliability of the on-load tap changer.

[0014] By introducing a displacement sensing unit as a feedback loop and employing adaptive algorithms such as recursive least squares, a closed-loop control system of perception, calculation, feedback, and updating is formed. This system can identify and correct mechanical motion delays caused by factors such as oil temperature, aging, and wear in real time, ensuring the long-term stability of timing control accuracy throughout the entire life cycle.

[0015] By detecting the power factor of the load in real time and dynamically adjusting the lead of the target breaking window according to the different physical characteristics of resistive and inductive loads, it ensures that sufficient time is reserved before the current crosses zero when breaking inductive loads to allow the contact gap to widen fully and the dielectric strength to fully recover, avoiding the risk of re-breakdown and current-cutting overvoltage after zero crossing, and making it more adaptable to resistive, inductive and capacitive loads. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the system structure of an on-load tap changer based on current zero-crossing prediction according to an embodiment of the present invention.

[0017] Figure 2 This is a schematic diagram of the sensor installation position and connection relationship according to an embodiment of the present invention.

[0018] Figure 3 This is a diagram showing the heterogeneous computing architecture and signal flow of the controller in an embodiment of the present invention.

[0019] Figure 4 This is a schematic diagram of the main flow of the adaptive cut-off control method according to an embodiment of the present invention. Attached Figure Description

[0020] 100. On-load tap changer; 110. Switching switch; 111. Moving contact; 120. Operating mechanism; 121. Reflective target; 10. Multi-source sensing unit; 11. Current sensing unit; 12. Voltage sensing unit; 13. Displacement sensing unit; 20. Digital actuator; 30. Controller. Detailed Implementation

[0021] To better understand the technical content of the present invention, specific embodiments are described below in conjunction with the accompanying drawings.

[0022] Example 1

[0023] Combination Figures 1 to 3 As shown, an adaptive on-load tap changer based on current zero-crossing prediction mainly includes a multi-source sensing unit 10, a digital actuator 20, and a controller 30.

[0024] Combination Figure 1 As shown, the multi-source sensing unit includes a current sensing unit 11, a voltage sensing unit 12, and a displacement sensing unit 13, which are used to collect the current waveform of the switching contact, the system voltage waveform, and the actual displacement trajectory of the moving contact in real time.

[0025] Optional, combined Figure 2 As shown, the current sensing unit 11 is a flexible Rogowski coil. This coil has an open flexible structure and can be directly wrapped around the insulating outer sheath of the inlet bushing of the on-load tap changer 100 without disconnecting the main circuit. It is used to detect the current flowing through the moving contact 111 and the stationary contact of the changeover switch 110.

[0026] Furthermore, its output is connected to a high-frequency integrator via a coaxial cable. This high-frequency integrator can be built into the front-end signal conditioning module of the controller 30, and can restore the differential signal di / dt output by the Rogowski coil, which is proportional to the time derivative of the measured current, to standard real-time current waveform data i(t) in real time for subsequent analog-to-digital conversion and processing.

[0027] As mentioned above, the flexible Rogowski coil can achieve high-precision current detection without disconnecting the main circuit, avoiding the drawbacks of traditional through-type current transformers that require power outages and wire disconnection during installation. It also poses no additional contact resistance or insulation risks to the primary system. In conjunction with a high-frequency integrator, the di / dt differential signal is restored to the standard current waveform i(t). This enables the controller to provide high-fidelity, low-latency real-time current data while ensuring the safety of personnel and equipment, thereby supporting the realization of microsecond-level current zero-crossing prediction.

[0028] Furthermore, the high-frequency integrator restores the differential signal di / dt output by the Rogowski coil to standard real-time current waveform data i(t), and sends it to the analog-to-digital conversion interface of the controller 30.

[0029] Optionally, the flexible Rogowski coil has an analog bandwidth of up to 1 MHz, enabling it to capture current waveform details, including high-order harmonics, without distortion.

[0030] The high analog bandwidth of 1MHz enables the current sensing unit 11 to completely retain the waveform details of up to dozens of harmonic components in the load current. Even under the condition that there is harmonic distortion in the actual power grid, the current zero-crossing phase information will not be distorted or delayed due to the bandwidth limitation of the sensing link, providing a high-quality signal source for subsequent fundamental wave extraction algorithms and accurate zero-crossing prediction.

[0031] Among them, the voltage sensing unit 12 is connected in parallel to the system bus side, and the displacement sensing unit 13 uses a laser beam to align with the reflective target 121 rigidly connected to the moving contact 111 to measure the displacement in real time.

[0032] In an optional embodiment, the voltage sensing unit 12 may be a voltage transformer.

[0033] Specifically, the primary winding of the voltage transformer is connected in parallel in the voltage transformer terminal box on the system bus side or line side to obtain the system voltage reference signal corresponding to the load current.

[0034] In this way, the voltage waveform on the system bus side or line side is the most stable and standard, which can accurately reflect the voltage phase on the power supply side, providing a reliable reference for subsequent power factor calculation and zero-crossing point auxiliary prediction. Its secondary winding outputs a standard low voltage signal u(t) to the controller 30.

[0035] In an optional embodiment, the displacement sensing unit 13 may be a high-speed laser displacement sensor, wherein the sensor is fixed on a dedicated bracket outside or inside the switching oil chamber, and the laser beam emitted by the sensor passes through the observation window or directly illuminates the reflective target 121 on the operating mechanism 120 that drives the moving contact 111 to move in the oil.

[0036] Thus, the high-speed laser displacement sensor positioned here can directly and without delay reflect the actual motion state of the moving contact 111, eliminating the interference of transmission chain gap on displacement measurement, thereby accurately capturing the micron-level displacement change at the moment of contact separation.

[0037] Furthermore, the high-speed laser displacement sensor continuously measures the absolute position change of the target at an extremely high sampling frequency (e.g., 20 kHz) and transmits the analog voltage signal or digital signal characterizing the displacement to the controller 30.

[0038] Furthermore, in combination Figure 3 As shown, the controller 30 adopts a heterogeneous architecture of field-programmable gate array and advanced reduced instruction set computer processor.

[0039] The controller 30 receives the sensor signal, and after being processed by the internal field-programmable gate array module and the advanced reduced instruction set computer processor module, it sends a trigger pulse to the digital actuator 20, which drives the operating mechanism 120 to move the moving contact 111 to complete the separation or engagement action in the insulating oil chamber.

[0040] Specifically, due to its parallel processing hardware characteristics, the field-programmable gate array module is specifically responsible for processing the high-frequency, real-time data stream from the current sensing unit 11, and it runs a fundamental wave extraction algorithm based on a second-order generalized integrator, which can output the real-time phase and frequency information of the current fundamental wave with a microsecond delay.

[0041] Specifically, the field-programmable gate array (FPGA) module is internally configured with pre-programmed hardware description language (HTAL) logic circuits. These logic circuits are used to perform parallel current fundamental frequency extraction operations. This logic circuit is built based on the principle of a second-order generalized integrator, and its function is equivalent to separating the sinusoidal fundamental frequency component from the distorted current in real time. The gate-level implementation of this hardware circuit is a standard mapping method for digital signal processing engineers in this field, so the specific gate circuit connections will not be described in detail here.

[0042] The Advanced Reduced Instruction Set Computing (ARPC) processor module interacts with the Field Programmable Gate Array (FPGA) module via a high-speed internal bus (such as the AXI bus) and is responsible for running complex floating-point arithmetic tasks, such as Fast Fourier Transform, Recursive Least Squares algorithm, and logical decision tasks.

[0043] In the heterogeneous architecture of this embodiment, the second-order generalized integrator phase-locked loop (PLL) running on the field-programmable gate array (FPGA) module is a time-domain adaptive filter that iterates on a sample-by-sample-point basis. Its inherent dynamic response time is approximately 1 to 2 power frequency cycles (20ms to 40ms), enabling it to quickly track changes in power grid frequency and phase. The Fast Fourier Transform (FFT) executed by the Reduced Instruction Set Computing (RISC) processor module updates the total harmonic distortion (THD) and power factor calculation results at fixed periodic intervals based on a sliding data window (typically with a window length of 20ms, one power frequency cycle).

[0044] Under normal operating conditions, the calculation results of the Fast Fourier Transform (FFT) and the phase-locked loop (PLL) results of the second-order generalized integrator mutually verify each other. When the power grid experiences transient disturbances or non-stationary distortions occur during the moment of interruption, the total harmonic distortion (THD) value will change abruptly. In this case, the controller 30 executes a cross-validation mechanism: the Advanced Reduced Instruction Set Computer (ARPC) processor module compares the THD value calculated by the latest FFT frame with the preset arc stability prediction threshold.

[0045] If the total harmonic distortion (THD) does not exceed the threshold, it indicates that the current waveform is still dominated by the fundamental wave, and the phase-locked loop (PLL) result of the second-order generalized integrator is reliable. The system continues to predict the zero-crossing point using the fundamental phase output by the second-order generalized integrator. If the THD exceeds the threshold, it indicates that the PLL accuracy of the second-order generalized integrator may have been affected by harmonic interference. In this case, the system triggers a pseudo-current zero-crossing point estimation mechanism, using the voltage zero-crossing point as an auxiliary reference for correction, to ensure that a reliable zero-crossing reference time can still be obtained under distorted operating conditions.

[0046] As mentioned above, by deploying the fundamental wave extraction and zero-crossing prediction tasks, which have extremely high real-time requirements and fixed operation modes, in the hardware parallel logic of the field-programmable gate array (FPGA) module, deterministic response with microsecond-level latency is achieved. Meanwhile, the computationally complex power factor calculation, delay correction algorithm, and decision-making tasks with numerous logical branches are handled by the high-instruction set computer (RISC) processor module. In this way, the heterogeneous architecture fully leverages the respective advantages of the two processors, ensuring the real-time performance and accuracy of current phase locking, while also meeting the flexibility requirements of adaptive decision-making under complex operating conditions.

[0047] Furthermore, the digital actuator 20, preferably a permanent magnet operating mechanism, directly receives the trigger control command from the controller 30 and drives the moving contact 111 to complete the separation or engagement action.

[0048] Thus, the permanent magnet operating mechanism directly receives the trigger pulse electrical signal from the controller 30. The permanent magnet provides the holding force for the closed or open state, and the electromagnetic force generated by the energization of the opening and closing coils overcomes the permanent magnet holding force and drives the iron core to move. The output shaft of the iron core is directly mechanically connected to the moving contact 111 of the switching switch 110 through the operating mechanism 120 (insulated pull rod).

[0049] Compared to the complex transmission chain of spring mechanisms, permanent magnet operating mechanisms reduce a large number of intermediate transmission links, thus their action delay T 延时 The sensitivity to changes in ambient temperature and mechanical wear is greatly reduced, providing a stable and repeatable execution basis for subsequent microsecond-level timing control.

[0050] It should be understood that, since the output interface of the permanent magnet operating mechanism (such as the mechanical dimensions and stroke of the output shaft) can be customized according to the input interface of the existing on-load tap changer 100, and its external contour dimensions are usually comparable to or more compact than the original spring mechanism, the original mechanism can be replaced in situ without changing the transformer tank body structure and the switch mounting flange.

[0051] Meanwhile, the power supply and drive signal cables required for the permanent magnet operating mechanism can be laid using the existing cable channels without requiring additional modifications to the surrounding insulation environment.

[0052] As described above, through high-frequency sampling by the current sensing unit 11 and real-time fundamental frequency extraction and zero-crossing prediction by the field-programmable gate array module, the controller 30 can accurately lock the ideal natural zero-crossing moment T0 several milliseconds before the actual contact separation action occurs. Simultaneously, combined with the inherent action delay T of the digital actuator 20... 延时 With precise calibration and timing compensation calculations, the system can actively guide and lock the moment of contact separation within a preset target breaking window, improving the passive state of random coupling of contact separation phase in traditional on-load tap changers, so as to suppress breaking arc energy to the maximum extent.

[0053] The target breaking window is a microsecond-level time period adjacent to the ideal natural zero-crossing point T0, and the time width of the target breaking window is dynamically adjusted based on the load power factor. By guiding and locking the contact separation moment within the microsecond-level target window before the current natural zero-crossing, the risk of repeated breakdown after the current zero-crossing can be avoided, and the arc energy at the moment of breaking can be minimized.

[0054] The following example illustrates the implementation process of current zero-crossing prediction in this scheme, using a set of specific numerical values.

[0055] Assuming the power grid frequency is 50Hz, the corresponding fundamental period is 20ms. The current sensing unit 11 continuously acquires the real-time current waveform data i(t) flowing through the moving contact 111 at a sampling rate of 1MHz, and sends the discrete sampling sequence to the field programmable gate array module in the controller 30.

[0056] The field-programmable gate array (FPGA) module internally runs a fundamental wave extraction algorithm based on a second-order generalized integrator. This algorithm performs real-time filtering and orthogonal decomposition on the input discrete current sample values ​​in hardware parallelism, outputting sine and orthogonal cosine components in phase with the fundamental current wave. Specifically, the FPGA module completes an iterative update of the state variables once in each sampling period and uses arctangent operation to calculate the instantaneous phase angle θ(t) of the fundamental current wave at the current moment in real time.

[0057] After the controller 30 receives the gear adjustment command and completes the decision, the field-programmable gate array module reads the fundamental phase angle θ at the current moment. 当前 Since the fundamental phase monotonically increases with a period of 2π, the phase corresponding to the next negative zero-crossing (i.e., the crossing point where the current changes from positive to negative) is π. Therefore, the remaining time Δt from the current moment to the next negative zero-crossing can be calculated. 过零 : If θ 当前 If π < Δt 过零 =(π-θ 当前 ) / (2πf; If θ 当前If Δt > π, then Δt 过零 =(3π-θ 当前 ) / (2πf; Where f is the fundamental frequency of the power grid being tracked in real time.

[0058] As an example, suppose that at the moment the field-programmable gate array module completes its decision, the calculated instantaneous phase angle θ of the fundamental current is... 当前 =0.3π radians (approximately 54°), real-time grid frequency f=50.00Hz. Therefore, the remaining time for the next negative zero-crossing is: Δt 过零 =(π-0.3π) / (2π×50.00)≈0.7π / (100π)≈7.00ms.

[0059] Therefore, the field-programmable gate array (FPGA) module precisely locks the next ideal natural zero-crossing time T0 as the current reference time plus 7.00 ms. This predicted value will be provided to subsequent timing-triggered calculations with microsecond-level accuracy.

[0060] Furthermore, a displacement sensing unit 13 is introduced as an online feedback element.

[0061] Thus, by introducing the displacement sensing unit 13 as a closed-loop feedback element, the system can compare the measured actual moment of separation with the preset target moment after each gear shifting action, calculate the timing deviation, and use an adaptive filtering algorithm to adjust the mechanism action delay T in the reference parameter library. 延时 Perform online corrections and updates.

[0062] Specifically, after an action is completed, the system compares T... 实际 With T 目标 The time series deviation δ is used to update the T in the benchmark parameter library online using a recursive least squares algorithm. 延时 value.

[0063] In a specific embodiment, the recursive least squares algorithm is used to correct the mechanism motion delay T in the reference parameter library online. 延时 The specific recursive steps are as follows: Step (1): Initialization When the system is first powered on or the baseline parameter library is established, the following initial parameters are assigned to the recursive least squares algorithm: The parameter vector to be estimated is θ̂0=T 延时库 The value in the middle corresponds to the initial delay value based on the current temperature and the number of actions. The covariance matrix P0 = σ²I, where σ² takes a large positive value (e.g., 10). 6 I is the identity matrix; The forgetting factor λ has a value of 0.98.

[0064] Step (2): Constructing observation data After each gear shifting action is completed, the system records the actual moment T of the action. 实际 And calculate the observed output value y n =T 实际 -T 触发 The observed value y n This actually reflects the total delay of this action (including the delay of the mechanism action and the delay of the control system). Input vector u n We set it to a constant of 1 because the delay parameter is a univariate estimate.

[0065] Step (3): Calculate the gain vector Based on the covariance matrix P of the previous moment n-1 Current input vector u n And the forgetting factor λ, calculate the gain vector K n : .

[0066] Step (4): Update parameter estimates Using the current observed output value y n The error between the predicted output and the parameter estimate at the previous time step is used to weight and correct the parameter estimate: ,in, ̂ is the updated mechanism motion delay T 延时 (new).

[0067] Step (5): Update the covariance matrix .

[0068] Step (6): Write back the benchmark parameter library The updated θ n Value as the new T 延时 Write back to the reference parameter library T 延时库 The corresponding temperature index and action count index are stored in the memory location to complete this closed-loop correction.

[0069] Through the above six-step recursive process, the recursive least squares algorithm can adaptively and smoothly correct the mechanism's action delay by using the measured timing deviation after each gear adjustment action, thereby maintaining timing control accuracy throughout the entire life cycle.

[0070] In this way, the system can automatically track and compensate for action delay drift caused by factors such as oil temperature changes, mechanical wear, and spring stress relaxation throughout the equipment's life cycle of several decades, ensuring the long-term stability and robustness of synchronous switching control accuracy.

[0071] Furthermore, the controller 30 calculates the power factor angle in real time and dynamically adjusts ΔT accordingly. This enables the implementation of differentiated timing strategies based on the different physical characteristics of the switching circuits.

[0072] Specifically, when interrupting inductive loads, by actively increasing ΔT, a longer dielectric strength recovery time is provided for the contact gap, thereby effectively suppressing the risk of re-breakdown caused by the rapid rise of transient recovery voltage after the current crosses zero; when interrupting resistive loads, a smaller ΔT is used to reduce unnecessary lead and avoid generating additional arc energy due to premature contact separation before the current drops sufficiently.

[0073] This improves the adaptability of on-load tap changers to different load types and their breaking reliability.

[0074] Example 2

[0075] Combination Figure 4 As shown, an adaptive disconnection control method based on current zero-crossing prediction is applied to the on-load tap changer in Example 1. The specific operation process is as follows: Step S1: Initialization and Self-establishment of the Baseline Parameter Library After the equipment is installed, debugged, or undergoes major maintenance, an initialization self-learning command is sent to the controller 30 via the host computer.

[0076] The controller 30 drives the digital actuator 20 (permanent magnet operating mechanism) to perform several full-stroke closing and opening actions under no-load conditions.

[0077] During this process, the displacement sensing unit 13 accurately records the rising edge T of the trigger pulse signal emitted from the field programmable gate array for each action. 触发 The moment T is reached when the contact displacement signal undergoes a sudden change (indicating actual contact separation). 实际 Controller 30 calculates the difference T between the two. 延时 =T 实际 -T 触发 .

[0078] At the same time, the controller 30 reads the current temperature value T from the temperature sensor installed in the oil chamber. 温度 And record the cumulative number of actions N that the mechanism has completed at this time.

[0079] By testing data at different temperature points multiple times, a two-dimensional lookup table, namely the mechanical motion delay reference parameter library T, was established inside the controller 30. 延时库 This parameter library uses temperature T as its base. 温度 The corresponding calibration delay value is stored using the number of actions N as an index, serving as the initial benchmark for subsequent dynamic compensation.

[0080] Step S2: Static Perception and Pattern Decision When the control system of the on-load tap changer receives a valid adjustment command, the controller 30 does not immediately drive the actuator to act, but first enters a brief analysis and decision window.

[0081] Sub-step S21: The Advanced Reduced Instruction Set Computer (ARPC) processor module reads the real-time current waveform data i(t) and system voltage waveform data u(t) for at least two complete power frequency cycles before the instruction arrives, and performs a Fast Fourier Transform operation.

[0082] The total harmonic distortion rate of the current and the phase difference between the voltage and the fundamental current are calculated, thereby obtaining the power factor angle φ of the load.

[0083] Sub-step S22: The controller 30 makes a mode decision based on the calculation results.

[0084] First, the controller 30 judges the calculated total harmonic distortion rate.

[0085] If the total harmonic distortion rate of the current is determined not to exceed the preset arc stability prediction threshold (e.g., set to 15%), it indicates that the current waveform quality is good and the zero-crossing point can be stably predicted. In this case, the controller 30 further performs an adaptive decision on the target lead based on the power factor angle φ: Since the power factor in this example is 0.82, which is less than the preset inductive load determination threshold of 0.9, the system determines that the current load is an inductive load.

[0086] According to a preset rule base, the target lead time ΔT is dynamically set to a first preset value, which can be selected as 500μs. If a resistive load with a power factor close to 1.0 is detected, ΔT will be set to a second preset value, which can be selected as 200μs.

[0087] It should be noted that the above settings reflect the targeted optimization of this solution for inductive loads. If a 200μs window is used indiscriminately, the contact gap will face the test of voltage peaks after the inductive load is disconnected, which will easily lead to re-breakdown. The 500μs advance allows sufficient insulation recovery time for this purpose.

[0088] Furthermore, if the total harmonic distortion rate of the current exceeds the preset arc stability prediction threshold (e.g., the total harmonic distortion rate is greater than 15%), it indicates that the current waveform has been severely distorted, and the direct prediction of the fundamental zero-crossing point will have a large error or be unstable.

[0089] Under this operating condition, the controller 30 activates the auxiliary prediction mechanism: using the zero-crossing point of the system voltage waveform data u(t) as an auxiliary reference.

[0090] Specifically, when calculating the pseudo-current zero-crossing point, it is necessary to obtain the voltage-current fundamental phase difference of the current load as the basis for calculation. Under conditions of severe current distortion, directly calculating this phase difference through Fast Fourier Transform may result in significant deviations due to harmonic interference.

[0091] To address this issue, controller 30 employs the following robust estimation strategy: The RISC-N (Reduced Instruction Set Computer) processor module maintains a circular buffer of several power frequency cycles to continuously store the calculated voltage-current phase difference values ​​for each normal cycle before distortion occurs. When the total harmonic distortion rate of the current exceeds a threshold, controller 30 extracts the historical average of the phase differences from the most recent normal cycles in the circular buffer as an estimate of the inherent voltage-current phase difference under the current operating condition. Under more stringent operating conditions, controller 30 can also extract the fundamental positive-sequence component from the three-phase voltage and current samples in real time based on the symmetrical component method. By calculating the phase difference between the positive-sequence voltage and positive-sequence current, the fundamental power factor angle is obtained, thereby effectively suppressing the interference of harmonics and asymmetrical components on the phase difference estimation.

[0092] Through the above strategy, even under severe operating conditions with severely distorted current waveforms, the system can still stably and reliably calculate the pseudo-current zero-crossing point, providing an effective timing reference for synchronous interruption control.

[0093] Furthermore, if the grid voltage itself also has severe distortion or transient disturbance, making it impossible to obtain a stable T0 prediction value even through the voltage zero-crossing point auxiliary method, then the controller 30 determines that the current synchronous interruption conditions are not met.

[0094] At this point, controller 30 automatically switches to safety protection mode, i.e., random phase switching mode, abandoning precise locking of the zero-crossing window and directly issuing trigger pulses according to the preset fixed delay to complete the necessary adjustment operation. Simultaneously, controller 30 outputs a synchronous failure alarm signal to the substation monitoring system through its communication interface, alerting maintenance personnel that there is a power quality problem or sensor malfunction in the current system, requiring inspection and maintenance.

[0095] Step S3: Dynamic Prediction and Timing Triggering The field-programmable gate array (FPGA) module uses its built-in second-order generalized integrator model to continuously track the phase of the fundamental current wave and lock the negative zero-crossing time T0 of the next fundamental current wave.

[0096] The Advanced Reduced Instruction Set Computer (ARPC) processor module reads the current oil temperature from T. 延时库 Search for the corresponding mechanism action delay.

[0097] The Advanced Reduced Instruction Set Computing (ARPC) computer processor module calculates the precise trigger timing according to a formula: T触发 =T0-ΔT-T 延时 .

[0098] Step S4, Phase Four: Motion Evaluation and Closed-Loop Correction During the execution of the action, the displacement sensing unit 13 continuously records the displacement data of the moving contact at a high sampling rate. The Advanced Reduced Instruction Set Computer (ARPC) module identifies abrupt changes in the displacement curve using an edge detection algorithm, thereby accurately determining the actual moment of rigidity T during this action. 实际 .

[0099] Calculation of timing skew in a LDPC (Reduced Instruction Set Computing) computer processor module: Timing deviation δ = T 实际 -T 目标 The recursive least squares algorithm module running in the Advanced Reduced Instruction Set Computer (ARPC) processor module receives the timing deviation δ.

[0100] The recursive least squares algorithm is a recursive identification algorithm with a forgetting factor. Its recursive process includes: First, a gain vector is calculated based on the weighted covariance matrix of the currently observed time-series deviation δ and the historical deviation data. Second, this gain vector is used to adjust the currently stored mechanism action delay value T. 延时 A weighted correction is performed so that the correction direction converges in the direction of reducing the timing bias δ; finally, the weighted covariance matrix is ​​updated for use in the next recursion.

[0101] The forgetting factor λ is preset to a fixed value between 0.95 and 0.99 to give higher calculation weight to recent deviation data, thereby effectively tracking the gradual change in action delay caused by oil temperature changes or mechanical wear.

[0102] Based on the current gain factor G of the recursive least squares algorithm (this factor is determined by the covariance of historical data; for example, assume G = 0.2), the delay value is weighted and corrected: T 延时 (New) = T 延时 (Old) + G × δ.

[0103] Updated T 延时 (New) was immediately written back to T 延时库 The old values ​​are overwritten in the corresponding storage locations for temperature and number of cycles. At this point, the closed-loop correction is complete.

[0104] Thus, the system will use this more precise delay time as T the next time it operates under similar conditions. 延时 Calculations are performed to determine the next trigger time T. 触发 The calculation is more accurate, T 实际It will be closer to T 目标 .

[0105] Step S5: Arc energy monitoring and interlocking protection steps To further enhance the system's safety and self-diagnostic capabilities, the method in this embodiment further includes steps for monitoring and evaluating the arc energy during the interruption process.

[0106] During the separation of the switching contacts and the burning of the electric arc, the controller 30 synchronously acquires the arc voltage u at both ends of the contacts through voltage detection circuits and current detection circuits arranged on both sides of the moving and stationary contacts, at a sampling rate of not less than 1MHz. 弧 (t) and the arc current i flowing through the contact 弧 (t).

[0107] Specifically, the voltage detection circuit includes a high-voltage isolation voltage divider resistor network connected in parallel between the moving and stationary contacts. The low-voltage signal after voltage division is sent to the controller via an isolation operational amplifier. The current detection circuit uses a broadband current transformer or shunt connected in series with the contacts, and its output signal is synchronously acquired after being processed by a signal conditioning circuit. The controller 30 synchronously samples the arc voltage u at a sampling rate of not less than 1MHz. 弧 (t) and arc current i 弧 (t), and perform product integration on the sampled values.

[0108] In a specific embodiment, the high-level reduced instruction set computer processor module of the controller 30 calculates the instantaneous arc voltage u based on the acquired data. 弧 (t) and arc current i 弧 (t), the arc energy E released during this interruption is calculated through integration, and the calculation formula is: E =∫u 弧 (t)·i 弧 The integral interval is from the instant the contacts separate until the arc current first drops to zero.

[0109] The controller 30 internally maintains an energy accumulation register to accumulate the arc energy generated by each successful gear shifting operation.

[0110] If the controller 30 detects that the single arc energy E exceeds the preset single energy safety threshold, or detects that the cumulative arc energy of several consecutive operations exceeds the preset cumulative energy safety threshold, it determines that there is a serious loss of synchronization or abnormal contact opening performance in the current drive phase. The controller 30 will actively block subsequent gear adjustment operations and output a blocking alarm signal to the substation monitoring system through the communication interface to prevent the equipment from continuing to operate under fault conditions and causing damage.

[0111] In an exemplary gear shifting operation embodiment, the power grid frequency is set to 50Hz; the current oil temperature is set to 45℃; the cumulative number of operations is set to 1230; the load characteristics are: inductive load, power factor 0.82; and the total harmonic distortion rate of the current is set to 2.8%.

[0112] Execution result of step S1: During the system initialization phase, under operating conditions of 45℃ oil temperature and approximately 1200 actuations, the initial mechanism action delay T corresponding to the reference parameter library is obtained through self-learning calibration. 延时 =8.500ms.

[0113] Step S2 execution process: The moment when controller 30 receives the gear shifting command is recorded as t0.

[0114] The Advanced Reduced Instruction Set Computer (ARPC) processor module reads the current and voltage data of the first two cycles of t0 and analyzes them using Fast Fourier Transform to find that the power factor is 0.82 (indicating an inductive load) and the total harmonic distortion rate is 2.8% (below the 15% threshold, which can be stably predicted).

[0115] Depending on the load type, the target lead time ΔT is set to the first preset value of 500 μs.

[0116] Step S3 execution process: The field-programmable gate array (FPGA) module calculates the fundamental phase of the current in real time and finds that at time t0, the instantaneous phase angle of the fundamental wave is 0.3π radians and the real-time grid frequency is 50.00Hz.

[0117] Calculate the remaining time for the next negative zero crossing. Since the fundamental phase has a period of 2π, the phase corresponding to the next negative zero crossing is π. The remaining time Δt1 = (π - 0.3π) / (2π×50) = 0.7π / 100π = 7.00ms.

[0118] In order to obtain sufficient time for calculation and action preparation, the system chooses to skip the upcoming zero-crossing point and instead target the zero-crossing point of the next cycle as the target break window.

[0119] Therefore, the predicted ideal natural zero-crossing time T0 = t0 + Δt1 + 20ms (one complete power frequency cycle) = t0 + 27.00ms.

[0120] Advanced Reduced Instruction Set Computing (ARPC) computer processor module from T 延时库 Search for the T corresponding to the current working condition in the middle. 延时 = 8.500ms.

[0121] Calculate the trigger time T 触发 =T0-ΔT-T 延时= (t0+27.00ms) -0.500ms-8.500ms=t0+18.000ms.

[0122] Execution result of step S4: At time t0+18.00ms, the controller 30 sends a trigger pulse to the digital actuator 20.

[0123] The displacement sensing unit 13 continuously monitors the displacement of the moving contact at a sampling rate of 20kHz. During the operation, the controller 30 identifies an abrupt change in the displacement curve at t0+26.540ms (indicating actual contact separation) using an edge detection algorithm, thereby accurately determining the actual moment of contact separation T during this operation. 实际 =t0+26.540ms.

[0124] According to T 实际 The actual mechanical delay of this action can be deduced: T 实际延时 =T 实际 -T 触发 = 8.540ms.

[0125] Target separation time T 目标 =T0-ΔT= (t0+27.00ms)-0.500ms=t0+26.500ms.

[0126] Timing deviation δ=T 实际 -T 目标 = (t0+26.540ms)-(t0+26.500ms) =+40μs.

[0127] By T 触发 As can be seen from the calculation formula, δ is numerically equal to the deviation between the actual delay and the calibrated delay: δ = (T 实际 -T 触发 )- (T 目标 -T 触发 ) =T 实际延时 -T 延时 =8.540ms-8.500ms=+40μs.

[0128] The recursive least squares algorithm uses δ=+40μs as the feedback quantity to perform weighted correction on the mechanism's motion delay. Assuming the current gain factor is 0.2, then T 延时 (New) = 8.500ms + 0.2 × 0.040ms = 8.508ms, the updated delay value is written back to T. 延时库 The corresponding storage location for temperature and number of times.

[0129] Execution result of step S5: During this interruption process, the controller synchronously collects the arc voltage and arc current, and calculates the single arc energy as E = 0.85 joules, which does not exceed the preset single-attack safety threshold (e.g., 2.0 joules). The accumulated energy register also does not exceed the limit, and the system maintains normal operation, waiting for the next adjustment command.

[0130] Through the above-mentioned continuous cycle of prediction, execution, feedback and correction, this invention tightly couples the microsecond-level power electronic control requirements with the millisecond-level mechanical action characteristics, minimizes the dispersion of mechanical components, and stabilizes the accuracy of long-term synchronous switching within a very small error band.

Claims

1. An adaptive turn-off control on-load tap changer based on current zero-crossing prediction, characterized in that, include: The multi-source sensing unit (10) is used to collect real-time operating status data of the on-load tap changer (100), including: The current sensing unit (11) is located on the incoming side of the on-load tap changer (100) and is configured to collect real-time current waveform data i(t) flowing through the moving contact (111) of the switching switch (110). The voltage sensing unit (12) is configured to collect system voltage waveform data u(t); The displacement sensing unit (13) is mechanically coupled to the operating mechanism (120) of the moving contact (111) and is configured to monitor and output displacement data representing the actual motion trajectory of the moving contact (111) in real time. The digital actuator (20) is connected to the moving contact (111) of the switch (110) and is used to receive control commands and drive the moving contact (111) to perform separation or engagement actions; The controller (30) is communicatively connected to the multi-source sensing unit (10) and the digital actuator (20). The controller (30) is configured to perform calculations based on the operating status data collected by the multi-source sensing unit (10) to determine the predicted ideal natural zero-crossing time T0, the load power factor, and the actual action delay T of the digital actuator (20) after calibration. 延时 Based on the ideal natural zero-crossing time T0, the load power factor, and the calibrated actual action delay T... 延时 Compensation is performed, and trigger control commands are generated and output to the digital actuator (20) so that the moving contact (111) can complete separation within the preset target opening window; The target switching window is dynamically determined based on the load power factor and is defined within a microsecond range before the ideal natural zero-crossing point T0.

2. The adaptive off control OLTC based on current zero-crossing prediction as claimed in claim 1, wherein, The current sensing unit (11) is a flexible Rogowski coil, and its output signal is processed by a high-frequency integrator to provide real-time current waveform data i(t) with a bandwidth of not less than 1MHz.

3. The adaptive off control OLTC based on current zero-crossing prediction as claimed in claim 1, wherein, The digital actuator (20) is a permanent magnet operating mechanism.

4. The adaptive off control OLTC based on current zero-crossing prediction as claimed in claim 1, wherein, The controller (30) is a heterogeneous computing architecture that includes a field-programmable gate array module and an advanced reduced instruction set computer processor module; Among them, the field-programmable gate array module is configured to implement the fundamental wave extraction algorithm based on the second-order generalized integrator in hardware to lock the fundamental wave phase of the current in real time, and the advanced reduced instruction set computer processor module is configured to execute the online identification and correction algorithm for power factor calculation and drive delay.

5. An adaptive disconnection control method for on-load tap changers based on current zero-crossing prediction, applied to any of the on-load tap changers described in claims 1 to 4, characterized in that, Includes the following steps: Step S1: Control the digital actuator (20) to perform the full stroke action, and use the displacement sensing unit (13) to record the actual separation time of the moving contact (111) to establish a mechanical motion delay reference parameter library T related to ambient temperature and / or the number of actions. 延时库 ; Step S2: When a gear shifting command is received, execute the following sub-steps: Step S21: Acquire and analyze the real-time current waveform data i(t) and system voltage waveform data u(t) of the current and historical cycles to calculate the total harmonic distortion rate of the current and the power factor angle of the load; Step S22: Based on the calculated power factor angle, dynamically determine a target lead ΔT. The target lead ΔT defines the target separation time T of the moving contact (111). 目标 The time lead of T relative to the ideal natural zero-crossing time T0, i.e., T 目标 =T0-ΔT; Step S3: Lock in the ideal natural zero-crossing time T0 of the fundamental frequency of the next current to be interrupted, and retrieve the reference parameter library T. 延时库 The search function retrieves the mechanism action delay T corresponding to the current operating condition. 延时 And calculate the trigger time T 触发 = T0-ΔT-T 延时 At the trigger time T 触发 A trigger pulse is sent to the digital actuator (20); Step S4: Record and identify the actual moment T of the moving contact (111) being snapped during this action. 实际 Calculate the timing deviation δ = T 实际 -(T0-ΔT), and use an adaptive filtering algorithm to take the timing deviation δ as feedback, to adjust the reference parameter library T. 延时库 The corresponding mechanism action delay T 延时 Perform online corrections and updates.

6. The method according to claim 5, characterized in that, In step S22: If the controller (30) determines that the load is an inductive load, the target lead time ΔT is set to the first preset value; If the controller (30) determines that the load is a resistive load, the target lead time ΔT is set to the second preset value; The first preset value is greater than the second preset value.

7. The method of claim 5, wherein, In step S22: If the controller (30) determines that the total harmonic distortion rate of the current exceeds the preset arc stability prediction threshold, then based on the zero-crossing point of the system voltage waveform data u(t) as an auxiliary reference, a pseudo-current zero-crossing point is calculated as an alternative reference for the ideal natural zero-crossing point time T0.

8. The method of claim 7, wherein, If a stable predicted value cannot be obtained through the voltage zero-crossing auxiliary method, the controller (30) will automatically switch to the random phase switching mode and output a synchronous failure alarm signal.

9. The method of claim 5, wherein, The adaptive filtering algorithm used in step S4 is the recursive least squares algorithm.

10. The method of claim 5, wherein, It also includes step S5: Collect the arc voltage and arc current at the moment of switching, and calculate the arc energy E during this interruption process; as well as If the single or cumulative arc energy E exceeds the preset safety threshold, it is determined that the drive phase is out of sync, and subsequent gear shifting operations are blocked.

Images