Intelligent combination switch

Abstract

The invention relates to the technical field of electric energy quality control, and discloses an intelligent composite switch, which comprises a high-frequency transient sensing module, a power frequency actual measurement module, a dual-mode composite compensation unit and a predictive control and self-learning core, according to the switch, a high-frequency transient sensing module is used for capturing a high-frequency transient signal (P-wave) caused by load action in real time and extracting a transient feature vector of the high-frequency transient signal. The predictive control and self-learning core uses a dynamic feature library to match the features; matching succeeds: the core retrieves the corresponding physical delay time and the power frequency impact amplitude, and controls the compensation unit to execute prior compensation at the moment after the P-wave is generated, namely when the power frequency impact (S-wave) is actually generated, so as to actively counteract the impact; and if the matching fails, the core triggers the power frequency actual measurement module to actively observe the S-wave which is not compensated this time. According to the method, predictive elimination of reactive power impact is realized, voltage fluctuation is avoided, and a newly added or changed load in a power grid is adapted through self-learning.

Description

Technical Field

[0001] This invention relates to the field of power quality control technology, specifically to intelligent composite switches. Background Technology

[0002] In modern power systems, especially in industrial environments such as metallurgy, chemical industry, and electrified railways, impact loads such as electric arc furnaces, rolling mills, and large motors are widely used. During startup or operation, these loads draw enormous amounts of reactive power from the grid in a very short time, generating severe reactive power surges. These surges are a major cause of severe voltage fluctuations, flicker, and even momentary drops at the point of common coupling (PCC), seriously threatening the stable operation of the power grid and the safety of other sensitive users on the same line.

[0003] To suppress this reactive power surge and stabilize grid voltage, existing power quality control schemes typically employ various dynamic var compensators, such as static var compensators (SVCs) based on thyristor-controlled reactors (TCRs) and thyristor-switched capacitors (TSCs), or static synchronous compensators (STATCOMs) based on voltage source inverters (VSCs).

[0004] Existing technical solutions suffer from common technical deficiencies when dealing with rapid reactive power surges. The core problem lies in the fact that these compensation devices all employ reactive or passive control logic. Specifically, they must first detect the voltage deviation or reactive power change caused by the reactive power surge at the power frequency (50 / 60Hz) level before the control system can calculate the required compensation amount and finally drive the power electronic devices to execute the compensation action. An inherent response delay inevitably exists throughout the entire process from the occurrence of the surge to the compensation output. For millisecond-level rapid surges, this delay causes the compensation action to always lag behind the surge itself, failing to suppress the most severe initial voltage drop at the moment of the surge.

[0005] Furthermore, when using traditional switching capacitor banks (such as TSC or MSC) to provide large-capacity compensation, the switching operation itself can easily cause secondary impacts or disturbances on the power grid, further deteriorating power quality. At the same time, the control strategies of these traditional devices are usually based on fixed power grid and load parameters. When the power grid structure or load characteristics (such as the addition of large equipment in a factory) change, their compensation effect will decrease, lacking the ability to dynamically adapt to unknown load conditions. Summary of the Invention

[0006] The purpose of this invention is to provide an intelligent composite switch that solves the problem that traditional reactive power compensation devices have a delayed response and can only passively compensate after a reactive power surge occurs, rather than actively predicting and eliminating the surge in advance.

[0007] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides an intelligent composite switch, comprising:

[0008] A high-frequency transient sensing module monitors connected power lines, captures high-frequency transient signals on the power lines, and generates transient feature vectors based on the high-frequency transient signals.

[0009] A predictive control and self-learning core is connected to the high-frequency transient sensing module; the predictive control and self-learning core is internally configured with a dynamic feature library and a pattern matching engine; the pattern matching engine is used to match the transient feature vector with the reference feature vector stored in the dynamic feature library; the predictive control and self-learning core is used for:

[0010] If a match is successful, the predicted impact parameters that match the transient feature vector are retrieved from the dynamic feature library. The predicted impact parameters include physical delay time and predicted impact amplitude.

[0011] A dual-mode composite compensation unit is connected to the predictive control and self-learning core; the dual-mode composite compensation unit is used to receive instructions from the predictive control and self-learning core, and is used for:

[0012] Based on the physical delay time and the triggering time of the high-frequency transient signal, at the predicted occurrence time of the power frequency impact, an a priori compensation operation corresponding to the predicted impact amplitude is performed on the power line.

[0013] To achieve the above objectives, the present invention also provides a reactive power compensation method for an intelligent composite switch, comprising the following steps:

[0014] Monitor power lines, capture high-frequency transient signals on the power lines, and generate transient feature vectors;

[0015] The transient feature vector is matched with the reference feature vector stored in the dynamic feature library;

[0016] If a match is successful, the predicted impact parameters that match the transient feature vector are retrieved from the dynamic feature library. The predicted impact parameters include physical delay time and predicted impact amplitude.

[0017] Based on the physical delay time and the triggering time of the high-frequency transient signal, the predicted occurrence time of the power frequency impact is determined, and the compensation unit is controlled to perform a priori compensation operation at the predicted occurrence time.

[0018] In one specific embodiment, the intelligent composite switch further includes:

[0019] The power frequency measurement module is connected to the predictive control and self-learning core.

[0020] The predictive control and self-learning core is also used for:

[0021] If the matching fails, the power frequency measurement module is triggered to enter the active observation mode;

[0022] The power frequency measurement module is used for:

[0023] In the active observation mode, the actual power frequency impact event that occurs on the power line immediately following the high-frequency transient signal is monitored and captured, and the actual impact parameters are calculated and generated.

[0024] The actual impact parameters include the time difference between the triggering time of the high-frequency transient signal and the starting time of the actual power frequency impact event, which is used as the measured physical delay time.

[0025] The predictive control and self-learning core is also used for:

[0026] The transient feature vectors that failed to match are associated with the actual impact parameters generated by the power frequency measurement module, and updated to the dynamic feature library as new mapping entries.

[0027] In one specific embodiment, the reactive power compensation method further includes the following steps:

[0028] If the matching fails, the power frequency measurement module will be triggered to enter active observation mode;

[0029] The power frequency measurement module monitors and captures actual power frequency impact events, calculates and generates actual impact parameters, including the measured physical delay time.

[0030] The transient feature vectors that failed to match are associated with the actual impact parameters and updated to the dynamic feature library as new mapping entries.

[0031] Preferably, the dual-mode composite compensation unit includes:

[0032] Fast transient unit;

[0033] Steady-state baseline unit;

[0034] The predictive control and self-learning core is also used to perform smooth handover control, which includes:

[0035] While controlling the steady-state baseline unit to input the steady-state compensation amount, the fast transient unit is controlled to output a compensation correction amount that is equal in magnitude and opposite in direction to the steady-state compensation amount, so as to maintain the continuity of the total compensation output of the dual-mode composite compensation unit.

[0036] Preferably, the high-frequency transient sensing module includes:

[0037] A high-bandwidth, non-invasive sensor is used to couple the current signal of the power line;

[0038] An analog high-pass filter is used to filter out the power frequency and low-order harmonic components in the current signal to obtain the high-frequency transient signal.

[0039] The feature extraction processor is used to perform time-frequency domain analysis on the high-frequency transient signal to extract energy, dominant frequency, attenuation coefficient and spectral bandwidth to form the transient feature vector.

[0040] Preferably, the power frequency measurement module includes:

[0041] The instantaneous reactive power calculation unit is used to calculate the instantaneous reactive power based on the collected three-phase voltage and three-phase current and instantaneous reactive power theory.

[0042] The steady-state change extraction unit is used to perform a moving average window calculation on the instantaneous reactive power to obtain the average reactive power;

[0043] An impact event parameterization processor is used to detect a step change in the average reactive power value and calculate the amplitude and duration of the step change to form part of the actual impact parameters.

[0044] In summary, the present invention has at least one of the following beneficial technical effects:

[0045] 1. This invention captures high-frequency transient signals that precede load action by setting up a high-frequency transient sensing module, and uses the dynamic feature library in the predictive control and self-learning core to retrieve the power frequency impact prediction parameters corresponding to the transient feature vector, especially including the key physical delay time. This mechanism enables the system to generate and execute prior compensation instructions in advance before the actual power frequency impact event occurs, thereby solving the inherent response delay problem of traditional compensation schemes that rely on power frequency signal detection.

[0046] 2. By setting up an online self-learning and calibration mechanism, this invention can automatically trigger the power frequency measurement module to capture the subsequent actual power frequency impact event when the system encounters an unknown transient feature not stored in the feature library. The system then establishes a new mapping relationship between the new transient feature vector and the parameters of the actual impact event and updates the dynamic feature library. This closed-loop learning capability enables the switch to dynamically adapt to new or changing loads in the power grid without manual intervention, ensuring the long-term accuracy of compensation.

[0047] 3. The dual-mode composite compensation unit used in this invention combines the fast response capability of a fast transient unit (such as STATCOM) with the large-capacity compensation capability of a steady-state baseline unit (such as TSC). Through predictive control and smooth handover control executed by the self-learning core, the system can accurately control the fast transient unit to output a compensation correction amount of equal magnitude and opposite direction at the moment the steady-state baseline unit is engaged, thereby maintaining the continuity of the total compensation amount and effectively avoiding secondary voltage fluctuations or impacts that may be caused when traditional capacitor banks are switched on or off. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the system architecture of the intelligent composite switch of the present invention;

[0049] Figure 2 This is a flowchart illustrating the internal structure and signal processing of the high-frequency transient sensing module of the present invention.

[0050] Figure 3 This is a block diagram showing the internal structure and functional structure of the power frequency measurement module of the present invention.

[0051] Figure 4 This is a schematic diagram of the internal structure and connection of the dual-mode composite compensation unit of the present invention;

[0052] Figure 5 This is a functional block diagram of the predictive control and self-learning core of the present invention;

[0053] Figure 6 This is a flowchart illustrating the overall operation of the intelligent composite switch of the present invention. Detailed Implementation

[0054] See attached document Figure 1 The intelligent composite switch provided by the present invention may include: a high-frequency transient sensing module, a power frequency measurement module, a dual-mode composite compensation unit, and a predictive control and self-learning core.

[0055] The high-frequency transient sensing module, the power frequency measurement module, and the dual-mode composite compensation unit are all coupled to the power line.

[0056] The high-frequency transient sensing module is electrically connected to the predictive control and self-learning core. The high-frequency transient sensing module is used to monitor power lines in real time, separate and extract high-frequency transient signals from the power lines, generate corresponding transient feature vectors based on the high-frequency transient signals, and then send the transient feature vectors to the predictive control and self-learning core.

[0057] The power frequency measurement module is electrically connected to the predictive control and self-learning core. The power frequency measurement module is used to monitor the power frequency parameters of the power line, and when it receives a trigger command from the predictive control and self-learning core, it captures the actual power frequency impact events that occur on the power line, parameterizes them, and sends them to the predictive control and self-learning core.

[0058] The dual-mode composite compensation unit is electrically connected to the predictive control and self-learning core. The dual-mode composite compensation unit receives compensation commands from the predictive control and self-learning core and performs reactive power compensation operations on the power lines.

[0059] The predictive control and self-learning core is connected to the high-frequency transient sensing module, the power frequency measurement module, and the dual-mode composite compensation unit, respectively. The predictive control and self-learning core has a built-in dynamic feature library, which is used to store the mapping relationship between transient feature vectors and power frequency impact parameters (especially including physical delay time).

[0060] The predictive control and self-learning core is configured to query the dynamic feature library after receiving transient feature vectors from the high-frequency transient sensing module.

[0061] If the transient feature vector matches the dynamic feature library, the corresponding power frequency impact prediction parameter is retrieved from the dynamic feature library. The predictive control and self-learning core generates a priori compensation instructions based on the power frequency impact prediction parameter and controls the dual-mode composite compensation unit to perform a reactive power compensation operation that is inversely related to the power frequency impact prediction parameter.

[0062] If the transient feature vector fails to match the dynamic feature library, the predictive control and self-learning core sends a trigger command to the power frequency measurement module to trigger the power frequency measurement module to record the actual power frequency impact event that occurs immediately after the high-frequency transient signal. After receiving the parameters of the actual power frequency impact event returned by the power frequency measurement module, the predictive control and self-learning core establishes a new mapping relationship between the transient feature vector and the parameters of the actual power frequency impact event, and updates the dynamic feature library.

[0063] See attached document Figure 2 In this embodiment, the high-frequency transient sensing module specifically includes: a high-bandwidth non-invasive sensor, an analog high-pass filter, an analog-to-digital conversion unit, and a feature extraction processor.

[0064] High-bandwidth, non-invasive sensors are coupled to power lines to acquire raw current signals from the power lines. In one specific implementation, the high-bandwidth non-invasive sensor is a Rogowski coil or a high-frequency current transformer (HFCT). The selected sensor has a bandwidth of not less than 100 kHz and does not undergo magnetic saturation under high power frequency current, to ensure that the weak high-frequency transient components superimposed on the power frequency current can be linearly reproduced.

[0065] The input of the analog high-pass filter is connected to the output of the high-bandwidth non-invasive sensor. The analog high-pass filter is configured with a fixed cutoff frequency. Cutoff frequency Set to a frequency significantly higher than the fundamental frequency of the power system. (50Hz or 60Hz) and its main low-order harmonic frequencies. In this embodiment, the following is set: ≥2kHz. Analog high-pass filter 12 is used to filter out the raw current signal. The power frequency components and low-order harmonic components in the signal are removed, resulting in a pure high-frequency transient signal. Its mathematical expression is:

[0066] ;

[0067] in, This is a high-pass filtering operation.

[0068] The analog-to-digital converter unit is connected to an analog high-pass filter for use at a sampling rate of For simulated high-frequency transient signals Discretization sampling is performed to generate discrete time series. Sampling rate Assuming that the Nyquist sampling theorem is satisfied, i.e. ,in, The sampling rate is the expected highest effective frequency of the transient signal. In this embodiment, the sampling rate is... Set to no less than 50kHz.

[0069] The feature extraction processor is connected to the analog-to-digital conversion unit. The feature extraction processor has pre-set triggering logic and feature calculation logic.

[0070] First, the feature extraction processor performs transient event triggering and interception operations. The trigger threshold is set to... When a discrete time series is detected The absolute value exceeds The time is determined as the start time of the transient event. The feature extraction processor extracts data from... Start time window Data segments within (total) (Sampling points), constituting the transient signal sequence to be analyzed, specifically represented as:

[0071] ;

[0072] Subsequently, the feature extraction processor performs joint time-frequency domain analysis on the transient signal sequence to calculate and generate transient feature vectors. The transient feature vector contains four physical feature components, represented as:

[0073] ;

[0074] The specific calculation definitions for each characteristic component are as follows:

[0075] Energy characteristics : Characterizes the total energy intensity of a transient signal within a time window.

[0076] ;

[0077] in, For time window The total number of sampling points within the area.

[0078] Frequency domain characteristics : The dominant frequency representing the point of maximum power spectral density in a transient signal. The feature extraction processor... The spectrum is obtained by performing a Fast Fourier Transform (FFT). And calculate the power spectral density. The index is then obtained from the power spectral density. Then, the system converts this index into the corresponding physical frequency based on the sampling rate and the number of FFT points, which can be expressed as:

[0079] ;

[0080] ;

[0081] in, This represents the total number of sampling points for performing the FFT. The frequency domain characteristics reflect the resonant properties of the internal circuit structure of the load.

[0082] Attenuation characteristics : Characterizes the rate of decay of the transient signal oscillation amplitude. Feature extraction processor extracts Hilbert envelope And using the exponential function The exponential coefficients obtained by fitting the envelope using the least squares method are... Specifically, it is expressed as:

[0083] ;

[0084] The attenuation rate characteristic reflects the damping characteristics of the load circuit.

[0085] Bandwidth characteristics The effective bandwidth characterizes the dispersion of frequency components in a transient signal, and its calculation formula is as follows:

[0086] ;

[0087] in, The center frequency of the spectrum.

[0088] The feature extraction processor sends the calculated transient feature vector to the predictive control and self-learning core through a digital communication interface for subsequent feature matching and querying.

[0089] See attached document Figure 3 In this embodiment, the power frequency measurement module is used as an online calibration unit for the intelligent composite switch. Its core function is to accurately capture and parameterize the actual power frequency reactive power impact (S-wave) caused by the load action after receiving the trigger command from the predictive control and self-learning core, so as to provide true data for self-learning.

[0090] The power frequency measurement module includes: a power frequency sensing unit 21, a power frequency sampling unit 22, an instantaneous reactive power calculation unit 23, a steady-state change extraction unit 24, and an impact event parameterization processor 25.

[0091] The power frequency sensing unit includes voltage transformers (PTs) and current transformers (CTs), which are coupled to the power line for synchronously acquiring power frequency three-phase voltage signals. , , and three-phase current signal , , .

[0092] The power frequency sampling unit is connected to the power frequency sensing unit and is used to synchronously sample and convert three-phase voltage and current analog signals to digital signals, generating discrete digital voltage sequences. and digital current sequence .

[0093] The instantaneous reactive power calculation unit, connected to the power frequency sampling unit, performs calculations based on instantaneous reactive power theory (pq theory). In one embodiment, the instantaneous reactive power calculation unit first performs a Clarke transformation on the three-phase voltage and current, converting them to... In a stationary coordinate system:

[0094] ;

[0095] ;

[0096] Subsequently, the instantaneous reactive power calculation unit calculates the instantaneous reactive power. (or its discrete form) ):

[0097] ;

[0098] this The signal includes the fundamental frequency component, harmonic components, and actual reactive power surge fluctuations.

[0099] The steady-state change extraction unit is connected to the instantaneous reactive power calculation unit. The steady-state change extraction unit is used to extract the instantaneous reactive power of high-frequency oscillations. Smoothing is performed to extract the average reactive power value reflecting the power frequency cycle. In this embodiment, the steady-state change extraction unit extracts the data over one power frequency cycle. The sliding average window implementation:

[0100] ;

[0101] in, It is the integral variable.

[0102] After processing The signal, after filtering out high-order harmonic interference, presents a smooth curve that clearly reflects the steady-state step change of reactive power.

[0103] The impact event parameterization processor is connected to the steady-state change extraction unit and the predictive control and self-learning core. The impact event parameterization processor receives a trigger command from the core (the command is sent at...). That is, the time when the high-frequency transient signal occurs. It is then activated and enters active observation mode.

[0104] Impact event parameterization processor continuously monitors The signal detects the step start time of reactive power by setting a threshold for the rate of change. and the end of the step jump .

[0105] Subsequently, the impact event parameterization processor calculates and generates parameters of the actual power frequency impact event, which are organized into a vector. :

[0106] ;

[0107] in:

[0108] reactive power step amplitude :

[0109] ;

[0110] Impact duration :

[0111] ;

[0112] Physical delay time This parameter is a key calibration parameter of the present invention, defined as the physical time difference from the occurrence of the high-frequency transient signal to the actual occurrence of the power frequency impact.

[0113] ;

[0114] Finally, the shock event parameterization processor will contain a vector of the three parameters mentioned above. Send to the predictive control and self-learning core for updating the dynamic feature library.

[0115] See attached document Figure 4 In this embodiment, the dual-mode composite compensation unit includes a fast transient unit and a steady-state baseline unit connected in parallel. Both the fast transient unit and the steady-state baseline unit are coupled in parallel to the access point of the power line and each receives control commands from the predictive control and self-learning core.

[0116] A fast transient unit (FTTU) is used to perform reactive power compensation operations defined by a priori compensation commands. In this embodiment, the FTTU is a voltage source inverter (VSC) device based on fully controlled power electronic devices (such as IGBTs), specifically a static synchronous compensator (STATCOM) or an active power filter (APF) with reactive power compensation function. The response speed of the FTTU is defined as the time from receiving the command to outputting a stable compensation current, which is less than a preset threshold (e.g., less than 5 milliseconds), enabling it to perform reactive power compensation operations within a physical delay time (…). The deployment of compensation actions was completed within a specified timeframe.

[0117] The steady-state baseline unit, with a total compensation capacity greater than that of the fast transient unit, is used to take over the subsequent steady-state reactive power compensation supply after the transient impact ends. In this embodiment, the steady-state baseline unit can specifically be multiple sets of thyristor-switched capacitor banks (TSC) or contactor-switched capacitor banks (MSC). The steady-state baseline unit provides a large capacity of baseline reactive power in a grouped switching manner through predictive control and self-learning core control. Its function is to release the compensation capacity of the fast transient unit, allowing the fast transient unit to return to standby state and prepare for the compensation of the next transient impact.

[0118] The fast transient unit and the steady-state baseline unit perform smooth handover control under the coordination of the predictive control and self-learning core. When the predictive control and self-learning core determines that the compensation load needs to be transferred from the fast transient unit to the steady-state baseline unit (e.g., after prior compensation has been completed and the power frequency impact has stabilized), the predictive control and self-learning core calculates the amount of compensation that needs to be applied to the steady-state baseline unit. .

[0119] At the same time that the steady-state baseline unit is activated. The predictive control and self-learning core issues a compensation correction instruction to the fast transient unit. This instruction causes its output to have a size of And the compensation correction amount in the opposite direction (i.e.

[0120] set up In the moments before deployment, This refers to the moment after the input. At any given time, the system's total compensation ,in, This is the output of the fast transient unit. At that moment, the steady-state baseline unit was put into operation. Meanwhile, the output of the fast transient unit becomes .

[0121] Therefore, in The total system compensation at time t is:

[0122] ;

[0123] Through this collaborative control, the total compensation amount of the system exist Maintaining continuous operation at all times avoids secondary voltage jumps or disturbances to power lines caused by capacitor bank switching.

[0124] See attached document Figure 5 The predictive control and self-learning core is the control and decision-making center of this intelligent composite switch, and can be implemented by a digital signal processor (DSP), microcontroller (MCU), or field-programmable gate array (FPGA). In this embodiment, the predictive control and self-learning core functionally includes: a dynamic feature library, a pattern matching engine, and a control logic unit.

[0125] The dynamic feature library is a non-volatile storage unit. It is used to store the mapping relationship between high-frequency transient feature vectors and power frequency impact parameters. Each entry in the library is represented by a reference feature vector. and a corresponding predicted impact parameter composition.

[0126] Reference feature vector The data structure is consistent with the feature vectors generated by the high-frequency transient sensing module, for example... Predicting impact parameters The data structure is consistent with the parameter vector generated by the power frequency measurement module, that is:

[0127] ;

[0128] in, The predicted reactive step amplitude, For the predicted duration of the impact, This is the physical delay time for this load characteristic.

[0129] The pattern matching engine is connected to a dynamic feature library. Its function is to receive the current transient feature vector from the high-frequency transient sensing module. and compare it with the features stored in the dynamic feature library. reference feature vectors ( High-speed comparison is performed.

[0130] In this example, the comparison is performed by calculating the weighted Euclidean distance. To achieve:

[0131] ;

[0132] in, This is the current transient feature vector; For the first in the library One reference feature vector; This is a diagonal weight matrix. (Weight matrix) diagonal elements Used to define different feature components (such as) or The importance of ) in the matching process, The larger the value, the greater the contribution of that feature component to the matching.

[0133] The pattern matching engine calculates all Minimum distance after and compare it with a preset matching threshold. Compare them.

[0134] like If the match is successful, the index of the best matching entry will be output. .

[0135] like If the match fails, it is considered a failed match.

[0136] The control logic unit receives the determination result from the pattern matching engine and executes one of the following two logical paths:

[0137] Path A: Match successful (Priority compensation executed)

[0138] When a successful match and the index of the best matching entry are received At that time, the control logic unit retrieves the corresponding predicted impact parameters from the dynamic feature library. To obtain the key and .

[0139] The control logic unit immediately generates a priori compensation instruction. This instruction defines the compensation amount. and execution time (in (This refers to the triggering time of the high-frequency transient signal).

[0140] The command is sent to the dual-mode composite compensation unit, controlling its fast transient unit to activate before the actual occurrence of the power frequency impact (S-wave). The compensation amount is executed at any time. Reactive power compensation operation.

[0141] After the prior compensation is executed, the control logic unit is also responsible for subsequent smooth handover control. When it is determined that the steady-state baseline unit needs to be activated (e.g., when activating...), the control logic unit will then... While sending an input command to the steady-state baseline unit, the control logic unit also sends a compensation correction command to the fast transient unit. This is to maintain the continuity of the total compensation output.

[0142] Path B: Matching failed (active observation and self-learning performed)

[0143] When a matching failure signal is received, the control logic unit enters active observation mode.

[0144] The control logic unit will use the current transient feature vector of the current failed match. Marked as features to be calibrated It is temporarily stored in the cache.

[0145] The control logic unit immediately sends a trigger command (along with the P-wave trigger time) to the power frequency measurement module. ), activate its S-wave capture function.

[0146] The control logic unit remains in a waiting state until it receives the actual power frequency impact event parameters returned by the power frequency measurement module. (Including actual measurements) , , ).

[0147] The control logic unit will temporarily store the features to be calibrated. Compared with the actual impact parameters received Perform pairing and establish new mapping relationships. ).

[0148] Finally, control logic unit 53 will pair this new pairing ( , As a new entry, it is written into the dynamic feature library, completing a self-learning and library update.

[0149] See attached document Figure 6 The operation process of the intelligent composite switch starts with the monitoring of the high-frequency transient sensing module.

[0150] In operation, the high-frequency transient sensing module continuously monitors the power line. When a high-frequency transient signal (P-wave) occurs on the power line due to a specific load action (such as the starting of a large motor or the ignition of an electric arc furnace), the high-frequency transient sensing module detects it at all times. The signal is captured, and the current transient feature vector is calculated immediately. It is then sent to the predictive control and self-learning core.

[0151] Predictive control and self-learning core received Then, its internal pattern matching engine executes the matching query. This query results in two mutually exclusive working paths: path A (successful match) or path B (failed match).

[0152] See attached document Figure 6 Path A (prior compensation and smooth transition of known load) corresponds to the detected current transient feature vector. The case where a matching entry exists in the dynamic feature library.

[0153] The specific steps of this process are as follows:

[0154] Step S110: The high-frequency transient sensing module in P-waves are detected at all times, and transient feature vectors are generated. It is then sent to the predictive control and self-learning core.

[0155] Step S120: Calculation of the pattern matching engine within the predictive control and self-learning core. Compared with all reference feature vectors in the dynamic feature library Weighted Euclidean distance The pattern matching engine determines that there exists an index of a best-matching entry. such that its minimum distance Less than the preset matching threshold The pattern matching engine outputs a successful match signal and the index of the best matching entry to the control logic unit. .

[0156] Step S130: The control logic unit indexes the best matching entry. The corresponding predicted impact parameters are retrieved from the dynamic feature library. This parameter contains at least two key data points: the predicted power frequency reactive power impact amplitude. and the physical latency unique to this load .

[0157] Step S140: The control logic unit immediately generates a priori compensation instruction. This instruction targets the fast transient unit (e.g., STATCOM) in the dual-mode composite compensation unit. This instruction defines the compensation amount. And the precise timing of compensation execution. .

[0158] Step S150: Set the execution time for the control logic unit. This setting ensures that the compensation action occurs synchronously with the actual occurrence of the power frequency impulse (S-wave). The control logic unit sends this prior compensation command to the fast transient unit.

[0159] Step S160: Fast transient unit in to Compensation operations are being prepared during this period. At time 1, the fast transient unit instantaneously injects a magnitude of 1 into the power line. The reactive power.

[0160] At the same time (Right now The reactive power demand generated by the actual power frequency reactive power impulse (S-wave) predicted by the P-wave reaching the power line is: .

[0161] Because the fast transient unit outputs simultaneously The compensation amount is the net reactive power change applied to the power line. Therefore, during the occurrence of S-waves, the reactive power of the power line remains stable, and voltage fluctuations are suppressed.

[0162] Step S170: Perform smooth handover control. After the transient shock (S-wave) stabilizes (e.g., in... The next scheduled time The load has entered a steady state of operation, but it still requires... Reactive power compensation. To release the capacity of the fast transient unit, the control logic unit initiates a handover process.

[0163] Step S180: In At any given time, the control logic unit sends a command to the steady-state baseline unit (e.g., TSC) to activate one or more sets of capacitors, providing... The amount of compensation ( Selected as equal to or close to ).

[0164] Step S190: In At the same time, the control logic unit sends a compensation correction command to the fast transient unit, causing its output compensation amount to change from... Reduce to .

[0165] Therefore, in The total compensation amount on power line 100 before and after the specified time. Maintain continuity,

[0166] The handover process is complete; the steady-state baseline unit takes over the steady-state compensation task, while the output of the fast transient unit decreases (in...). When the time drops to 0), it returns to standby mode, preparing for the next P-wave detection and prior compensation.

[0167] See attached document Figure 6 Path B (active observation and self-learning of unknown load) corresponds to the detected transient feature vector. No matching entry was found in the dynamic feature library (i.e., it belongs to unknown load or newly connected load).

[0168] The specific steps of this process are as follows:

[0169] Step S210: The high-frequency transient sensing module at time... High-frequency transient signals (P-waves) on power lines are captured, and transient feature vectors are generated. And send it to the predictive control and self-learning core.

[0170] Step S220: Calculation of the pattern matching engine within the predictive control and self-learning core. The distance to all reference feature vectors in the dynamic feature library. At this point, the minimum weighted Euclidean distance is calculated. Greater than or equal to the preset matching threshold (Right now The pattern matching engine generates a match failure signal and sends that signal along with the current... It is passed to the control logic unit.

[0171] Step S230: In response to the matching failure signal, the control logic unit does not generate a priori compensation instruction (at this time, the dual-mode composite compensation unit remains in standby mode and does not execute any action), but immediately enters the active observation mode. The control logic unit will then... Marked as feature vector to be calibrated It is then temporarily stored in the core's high-speed cache memory, awaiting pairing with subsequent truth data.

[0172] Step S240: The control logic unit sends a trigger observation command to the power frequency measurement module. This command includes the occurrence time of the high-frequency transient signal. (As a time reference point). This instruction activates the impact event parameterization processor in the power frequency measurement module.

[0173] Step S250: The power frequency measurement module is in a high-sensitivity monitoring state, measuring the average reactive power of the power line. Real-time tracking is performed. Since no prior compensation was performed in step S230, the actual power frequency impulse (S-wave) immediately following the P-wave will naturally act on the power line. The power frequency measurement module detects... The step change, recording the start time of the step change. and the end of the step jump And measure the step amplitude.

[0174] Step S260: The power frequency measurement module calculates the parameters of this impact event and generates the parameter vector of the actual power frequency impact event. .

[0175] In this step, the physical delay time The calculation is strictly based on the occurrence time received in step S240. The start time measured in step S250 The difference is:

[0176] ;

[0177] Should The value precisely quantifies the time interval between the current unknown load emitting a high-frequency transient precursor and the actual generation of power frequency reactive power impact. The power frequency measurement module will calculate the value... It is then fed back to the predictive control and self-learning core.

[0178] Step S270: The control logic unit receives... Then, the feature vector to be calibrated, temporarily stored in step S230, is retrieved from the cache memory. The control logic unit performs feature-parameter association operations, which... Set as the new reference feature vector ,Will Set as the corresponding predicted impact parameters .

[0179] Step S280: The control logic unit sends a write command to the dynamic feature library to associate the above-mentioned pairs. It is stored as a new mapping entry in the dynamic feature library.

[0180] At this point, the self-learning process ends. When the same or similar transient feature vectors reappear on the power line... When the load action occurs, the pattern matching engine will be able to calculate This automatically redirects to path A, utilizing the information learned in step S260. and This achieves precise prior compensation. The process enables the system to adaptively calibrate for new loads or loads with changing characteristics in the power grid, maintaining compensation accuracy without manual intervention.

Claims

1. An intelligent composite switch, characterized in that, include: The high-frequency transient sensing module is used to separate and extract high-frequency transient signals from power lines in real time and generate corresponding transient feature vectors. The power frequency measurement module is used to monitor the power frequency parameters of the power line and capture actual power frequency impact events; Dual-mode composite compensation unit for reactive power compensation operation; The predictive control and self-learning core has a built-in dynamic feature library, which is used to store the mapping relationship between the transient feature vector and the power frequency impact parameters including physical delay time. The predictive control and self-learning core is used for: When the transient feature vector successfully matches the dynamic feature library, the corresponding power frequency impact prediction parameter is retrieved, and a priori compensation instruction is generated to control the dual-mode composite compensation unit to perform a reactive power compensation operation that is opposite to the power frequency impact prediction parameter. When the transient feature vector fails to match the dynamic feature library, the power frequency measurement module is triggered to record the actual power frequency impact event that occurs immediately after the high-frequency transient signal, and establishes a new mapping relationship between the transient feature vector and the parameters of the actual power frequency impact event and updates the dynamic feature library.

2. The intelligent composite switch according to claim 1, characterized in that, The parameters of the actual power frequency impact event include: reactive power step amplitude, impact duration, and the time difference from the time of occurrence of the high-frequency transient signal to the time of occurrence of the actual power frequency impact event, which is recorded as the physical delay time.

3. The intelligent composite switch according to claim 1, characterized in that, The transient feature vector is composed of joint time-frequency domain features, including at least: energy features based on the total energy of the transient signal, frequency domain features based on the power spectral density, and attenuation rate features based on the signal envelope.

4. The intelligent composite switch according to claim 1, characterized in that, The predictive control and self-learning core is also used for: Calculate the weighted Euclidean distance between the currently extracted transient feature vector and the reference feature vector stored in the dynamic feature library; If the weighted Euclidean distance is less than the preset matching threshold, the match is considered successful.

5. The intelligent composite switch according to claim 1, characterized in that, The high-frequency transient sensing module includes a high-bandwidth non-invasive sensor and a high-pass filter. The cutoff frequency of the high-pass filter is set to be higher than the power line frequency and its main harmonic frequencies. The high-pass filter is used to filter out the power frequency component.

6. The intelligent composite switch according to claim 1, characterized in that, The power frequency measurement module specifically includes: The instantaneous reactive power calculation unit is used to calculate the instantaneous reactive power based on the instantaneous reactive power theory, thereby completing the monitoring of the power frequency parameters; The steady-state change extraction unit is used to extract the steady-state change of the actual power frequency impact event by performing a sliding average window processing on the instantaneous reactive power, thereby completing the capture of the actual power frequency impact event.

7. The intelligent composite switch according to claim 1, characterized in that, The predictive control and self-learning core executes an active observation mode after the transient feature vector fails to match the dynamic feature library. The active observation mode specifically involves: The current transient feature vector is marked as a feature to be calibrated. After the power frequency measurement module completes the capture of the actual power frequency impact event, the feature to be calibrated is paired with the measured parameters of the captured actual power frequency impact event and stored in the dynamic feature library.

8. The intelligent composite switch according to claim 1, characterized in that, The dual-mode composite compensation unit includes: A fast transient unit, whose response speed is higher than a preset threshold, is used to perform the reactive power compensation operation defined by the prior compensation instruction; The steady-state baseline unit, with a capacity greater than that of the fast transient unit, is used to take over the subsequent steady-state reactive power compensation supply after the reactive power compensation operation is performed, thereby releasing the compensation capacity of the fast transient unit.

9. The intelligent composite switch according to claim 8, characterized in that, The fast transient unit is a static synchronous compensator or an active power filter; The steady-state baseline unit is a thyristor-switched capacitor bank or a contactor-switched capacitor bank.

10. The intelligent composite switch according to claim 8, characterized in that, The predictive control and self-learning core is also used to perform smooth handover control, which specifically includes: At the moment when the steady-state baseline unit is activated, the fast transient unit is controlled to output a compensation correction amount that is equal in magnitude and opposite in direction to the activation amount of the steady-state baseline unit, so as to maintain the continuity of the total compensation amount.

Images