An Adaptive PID Control Method and System for High-Voltage Mining Frequency Converters Based on Load Prediction

Abstract

This invention relates to the field of adaptive PID control technology for frequency converters, and particularly to an adaptive PID control method and system for high-voltage mining frequency converters based on load prediction. The method includes the following steps: system initialization, signal acquisition and processing, load prediction and determination, dynamic parameter optimization, and control execution and feedback. The system includes a load prediction module, an adaptive PID parameter adjustment module, a mining anti-interference compensation module, and an integrated coordinated control module. This invention fundamentally avoids large overshoots caused by insufficient response and unplanned equipment shutdowns due to overload and oscillations through predictive regulation, ensuring insulation safety under high-voltage environments, improving stability and electrical reliability under impact loads, reducing the technical threshold and time cost of on-site commissioning, resulting in significant economic benefits and providing dual safety guarantees for equipment and personnel.

Description

Technical Field

[0001] This invention relates to the field of adaptive PID control technology for frequency converters, specifically to an adaptive PID control method and system for high-voltage mining frequency converters based on load prediction. Background Technology

[0002] High-voltage mining frequency converters are core equipment for modern mine automation and energy-saving operation. Their control performance directly affects the operating efficiency and reliability of important equipment such as underground mining and transportation. To achieve precise control of the drive motor, using PID control technology to construct a closed loop is a common practice in the industry. Existing technologies mainly focus on how to set or adjust PID parameters, and can be divided into two main categories: The first type is the fixed-parameter PID control scheme. This scheme uses preset, constant proportional (P), integral (I), and derivative (D) parameters to form the control loop. The system collects signals such as motor speed and output current in real time, calculates the deviation from the set value, and performs calculations based on the deviation using fixed PID parameters, outputting control commands to the frequency converter drive unit to adjust the voltage and frequency. Its typical architecture is a closed-loop mode of "sensor signal acquisition module → PID controller → frequency converter power drive unit". This scheme is simple in structure and easy to implement, but it highly depends on the experience of the commissioning engineer for parameter preset, and is usually only suitable for mining scenarios where the load characteristics are relatively stable and change slowly, such as drainage pumps that operate at a fixed power for a long time.

[0003] The second category is the conventional adaptive PID control scheme. To cope with a certain degree of load variation, this scheme introduces a dynamic parameter adjustment mechanism. Its core is based on real-time deviation signals (including deviation magnitude and rate of change), using a specific adaptive algorithm (such as fuzzy control algorithm) to fine-tune the PID parameters online. Its typical structure adds a "deviation analysis module" and a "fuzzy decision unit," first analyzing and fuzzifying the deviation, and then outputting the corrected PID parameters according to a predefined fuzzy rule base. Compared to fixed-parameter PID, this type of scheme has achieved some application in mine hoists with continuously varying partial loads, enabling real-time parameter adjustment.

[0004] However, the aforementioned existing technologies still have significant shortcomings and defects when dealing with complex working conditions with typical mine production characteristics, mainly in the following three aspects: 1. Poor adaptability to load surges: The harsh mining environment and the sudden, impactful loads on critical equipment such as scraper conveyors and tunneling machines can cause loads to increase several times within tens of milliseconds due to sudden events such as coal blockage or large ore falling. Faced with such high-frequency, drastic load surges, fixed-parameter PID controllers, due to lag in control parameters and slow response (typically exceeding 200ms), are prone to severe speed overshoot (overshoot rate greater than 15%), leading to system oscillations or even triggering inverter overload protection and shutdown, severely impacting production continuity. While conventional adaptive PID controllers can dynamically adjust, their adjustment mechanism is based on an already occurring deviation, essentially remaining a lag response. They cannot provide sufficient control strength in the initial stages of load surges and are similarly unable to effectively suppress large overshoots.

[0005] 2. Lack of Load Trend Prediction: Conventional adaptive PID control logic adjusts the current output based on the current error, which is a passive follower control. It completely lacks the ability to predict future load changes. In scenarios with clear periodic and regular load fluctuations, such as mine hoists (e.g., lifting heavy loads and lowering light loads), this passive "following the drop but not the rise" mode causes the inverter output power to frequently chase the actual load changes instead of matching them in advance. This not only reduces the dynamic stability of the system but also exacerbates the switching losses and thermal stress of power modules such as IGBTs. Statistics show that this will increase the power module losses by about 10% to 15%, shortening the equipment's lifespan.

[0006] 3. Insufficient adaptability to the special environment of mining: Existing algorithms generally do not fully consider the special operating environment faced by mining frequency converters during their design. First, in terms of the electrical environment, the 1140V / 3300V high-voltage power supply system has extremely high requirements for insulation safety. However, when existing algorithms perform rapid parameter switching, they may directly or indirectly introduce voltage spikes and high-frequency harmonics, which will accelerate the aging of motor insulation over a long period of time, posing a safety hazard. Second, in terms of the physical environment, the underground space is narrow, and large electromechanical equipment is densely packed, resulting in strong electromagnetic interference. The signal acquisition circuits and filtering algorithms of existing solutions are not targeted enough, resulting in a large amount of noise in the feedback speed and current signals, with a deviation detection error rate of 8% to 10%. This seriously pollutes the "senses" of the control system, making it difficult for any advanced control algorithm to achieve its due accuracy.

[0007] No solutions have yet been proposed for the relevant technical issues. Summary of the Invention

[0008] To address the problems in related technologies, this invention proposes an adaptive PID control method and system for high-voltage mining frequency converters based on load prediction. This overcomes the aforementioned technical problems in existing related technologies. The purpose of this invention is to fundamentally avoid large overshoots caused by insufficient response and unplanned equipment shutdowns due to overload and oscillations through predictive regulation, ensuring insulation safety in high-voltage environments, improving stability and electrical reliability when facing impact loads, reducing the technical threshold and time cost of on-site commissioning, and achieving significant economic benefits. It also provides dual safety guarantees for equipment and personnel.

[0009] To achieve the above objectives, the present invention provides the following technical solution: an adaptive PID control method for high-voltage mine frequency converters based on load prediction, comprising the following steps: S1, System initialization: Configure the inverter rated voltage, motor rated power, load type and load prediction threshold, and load the benchmark PID parameter library and pre-trained Long Short-Term Memory (LSTM) network model corresponding to the load type. S2, Signal Acquisition and Processing: Real-time acquisition of the inverter's output current, motor speed, and bus voltage signals, followed by Kalman filtering and differential amplification to obtain the feedback signal after anti-interference processing; S3, Load Prediction and Judgment: Perform wavelet transform on the current signal after anti-interference processing to extract current change features; when the extracted current change features exceed a preset threshold, activate the LSTM model to predict the load change trend and load change rate within the next 100ms to 200ms time window. S4, Parameter Dynamic Optimization: Based on the load change trend, the load change rate, and the current speed deviation, the parameters of the PID controller are dynamically adjusted using the particle swarm optimization algorithm to generate an optimized PID parameter set; the adjustment is completed before the actual load change, and the total time is less than 50ms. S5, Control Execution and Feedback: The optimized PID parameter set is applied to the inverter drive unit to control the power module output; at the same time, the actual load is compared with the predicted load, and when the deviation exceeds the set threshold, the parameters of the LSTM model are corrected online.

[0010] Preferably, in step S3, the load prediction and determination specifically includes: (1) Acquire the inverter output current signal at a sampling frequency of 10kHz; (2) Analyze the current signal using wavelet transform and calculate the slope of the current change; (3) When the current change slope is detected to be greater than 5A / ms, it is determined to be a precursor to a sudden load change; (4) Input the historical and current data sequence including the current mutation characteristics into the LSTM model, and the LSTM model outputs the load prediction value and change rate within a specified future time.

[0011] Preferably, in step S4, the dynamic optimization of parameters specifically includes: (1) A load-parameter mapping library is preset, wherein the mapping library defines the baseline PID parameters according to the load range, and the load range includes at least: light load range, medium load range and heavy load range; (2) Using the load prediction value as the query basis, obtain the baseline PID parameters of the corresponding load range from the mapping library; (3) Using the benchmark PID parameters as the initial population, a fitness function is constructed by combining the load change rate and the current speed deviation, and the particle swarm optimization algorithm is run for iterative optimization; When a sudden increase in load is predicted, the optimization objective is to increase the proportional coefficient P and the integral coefficient I; when a sudden decrease in load is predicted, the optimization objective is to decrease the proportional coefficient P.

[0012] Preferably, the load range is defined as follows: light load is less than 30% of the rated load, medium load is 30% to 70% of the rated load, and heavy load is greater than 70% of the rated load.

[0013] Preferably, in step S2, the signal acquisition and processing further includes high-voltage anti-interference compensation, which specifically includes: (1) The collected speed and current signals are processed by differential transmission and Kalman filtering algorithm to achieve an electromagnetic interference suppression rate of more than 85%. (2) When outputting control parameters to the drive unit, a voltage spike suppression algorithm is executed to ensure that the output voltage fluctuation of the inverter does not exceed the motor insulation tolerance threshold under a high voltage environment of 1140V or 3300V.

[0014] Preferably, it also includes integrated security coordination control, which specifically includes: (1) Real-time monitoring of inverter power module temperature and system insulation resistance; (2) When the temperature of the power module or the insulation resistance exceeds the safe range, the PID controller parameters are automatically switched to the preset safe parameter set and an alarm is triggered.

[0015] Preferably, the benchmark PID parameter library is constructed separately for different types of mining equipment, and the types of mining equipment include at least scraper conveyors and hoists.

[0016] Preferably, in step S5, the optimized PID parameters are transmitted to the IGBT drive unit of the frequency converter through an optical fiber isolation interface.

[0017] To achieve the above objectives, the present invention also provides the following technical solution: An adaptive PID control system for high-voltage mining frequency converters based on load prediction is used to implement the above-mentioned method. The system includes: The load prediction module is used to acquire current signals, extract abrupt change features, and use an LSTM model to predict load changes. The adaptive PID parameter adjustment module has a built-in load-PID parameter mapping library and particle swarm optimization unit, which is used to dynamically optimize PID parameters based on prediction results. The mining anti-interference compensation module is used to perform Kalman filtering and differential processing on the acquired speed, current and voltage signals, and to perform voltage spike suppression. The integrated coordination control module coordinates the operation of the load prediction module, the adaptive PID parameter adjustment module, and the mine anti-interference compensation module, and switches to the safety control mode when the equipment status is abnormal.

[0018] Compared with the prior art, the beneficial effects of the present invention are: (1) This invention is an adaptive PID control method and system for high-voltage mining frequency converter based on load prediction. Through a three-level architecture of load prediction-parameter optimization-execution, and by using a dual-dimensional model that integrates current mutation feature extraction and LSTM trend prediction, the load mutation trend can be predicted 50-200ms in advance. The lag time of parameter adjustment is shortened to less than 50ms, and the response speed is improved by more than 75%. This fundamentally avoids the large overshoot caused by the system's insufficient response, effectively suppresses the motor speed overshoot rate to less than 3%, significantly improves the system's stability when facing impact loads, and effectively avoids unplanned equipment shutdowns caused by overload and oscillation. (2) This invention is an adaptive PID control method and system for high-voltage mine frequency converters based on load prediction. Through the synergistic effect of two levels, the control accuracy is fully optimized. First, at the signal source, a mine anti-interference compensation algorithm combining differential signal transmission and Kalman filtering is adopted to reduce the signal acquisition error rate caused by strong electromagnetic interference in the mine from 8%-10% in the prior art to below 2%. Second, at the control decision level, a high-precision load prediction model provides a reliable basis for parameter optimization. The combination of the two improves the matching degree between the output frequency of the frequency converter and the actual demand of the dynamically changing load to more than 98%. At the same time, the built-in voltage spike suppression function of the algorithm ensures the insulation safety in the 1140V / 3300V high-voltage environment and improves the overall electrical reliability of the system. (3) This invention is an adaptive PID control method and system for high-voltage mining frequency converters based on load prediction. It constructs a PID parameter mapping library based on different mining equipment load types, so that engineers do not need to perform tedious on-site PID parameter tuning for each piece of equipment. It can quickly adapt to at least 5 common mining equipment, greatly reducing the technical threshold and time cost of on-site debugging. It has strong adaptability and greatly reduces the mechanical shock of motors and the frequent and drastic adjustment of power modules caused by sudden load changes. It can extend the service life of power modules by about 25% and reduce the overall maintenance cost of equipment by about 20%, resulting in significant economic benefits. Attached Figure Description

[0019] Figure 1 This is an architecture diagram of the system of the present invention; Figure 2 This is a flowchart of the load prediction algorithm of the present invention; Figure 3 This is a flowchart illustrating the optimization of PID parameters for this invention. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Example 1

[0021] Please see Figure 1-3 This invention proposes a technical solution for an adaptive PID control method and system for high-voltage mining frequency converters based on load prediction: The adaptive PID control method for high-voltage mining frequency converters based on load prediction is applicable to mining frequency converters with voltage levels of 1140V or 3300V, and aims to cope with the impact load changes of equipment such as scraper conveyors and hoists. The method includes the following steps: S1, System Initialization: Configure the inverter's rated voltage (1140V or 3300V), motor's rated power, load type (e.g., select from options such as "scraper conveyor", "elevator", "main ventilator"), and load prediction threshold; load the baseline PID parameter library and pre-trained Long Short-Term Memory (LSTM) network model corresponding to the load type; for example, set a current change slope exceeding 5A / ms as a precursor to a sudden change requiring load prediction to be initiated. S2, Signal Acquisition and Processing: Real-time acquisition of the inverter's output current, motor speed, and bus voltage signals, followed by Kalman filtering and differential amplification to obtain the feedback signal after anti-interference processing; S3, Load Prediction and Judgment: Perform wavelet transform on the current signal after anti-interference processing to extract current change features; when the extracted current change features exceed a preset threshold, activate the LSTM model to predict the load change trend and load change rate within the next 100ms to 200ms time window. S4, Parameter Dynamic Optimization: Based on the load change trend, the load change rate, and the current speed deviation, the parameters of the PID controller are dynamically adjusted using the particle swarm optimization algorithm to generate an optimized PID parameter set; the adjustment is completed before the actual load change, and the total time is less than 50ms. S5, Control Execution and Feedback: The optimized PID parameter set is applied to the inverter drive unit to control the power module output; at the same time, the actual load is compared with the predicted load, and when the deviation exceeds the set threshold, the parameters of the LSTM model are corrected online.

[0022] Furthermore, in step S3, the load prediction and determination specifically includes: (1) Acquire the inverter output current signal at a sampling frequency of 10kHz; (2) Analyze the current signal using wavelet transform and calculate the slope of the current change; (3) When the current change slope is detected to be greater than 5A / ms, it is determined to be a precursor to a sudden load change; (4) Input the historical and current data sequence including the current mutation characteristics into the LSTM model, and the LSTM model outputs the load prediction value and change rate within a specified future time.

[0023] In this embodiment, the high-precision current signal processed by S2 is analyzed in the time and frequency domain through wavelet transform to extract features that characterize rapid load changes, such as the instantaneous current slope. This feature is continuously monitored, and once it is detected that it exceeds a preset threshold (e.g., 5A / ms), it is immediately determined that the load is about to undergo a sudden change. At this time, the system activates the loaded LSTM neural network prediction model. The LSTM model receives the recent current data sequence containing the sudden change feature points, and combines it with the long-term time-series dependency rules learned within its network to output the prediction results of the load change trend (increase or decrease) and rate of change in the next 100ms to 200ms period. For example, the model may output the prediction that "the load will increase to 1.8 times the rated value after 100ms".

[0024] Furthermore, in step S4, the dynamic optimization of parameters specifically includes: (1) A load-parameter mapping library is preset, wherein the mapping library defines the baseline PID parameters according to the load range, and the load range includes at least: light load range, medium load range and heavy load range; (2) Using the load prediction value as the query basis, obtain the baseline PID parameters of the corresponding load range from the mapping library; (3) Using the benchmark PID parameters as the initial population, a fitness function is constructed by combining the load change rate and the current speed deviation, and the particle swarm optimization algorithm is run for iterative optimization; When a sudden increase in load is predicted, the optimization objective is to increase the proportional coefficient P and the integral coefficient I; when a sudden decrease in load is predicted, the optimization objective is to decrease the proportional coefficient P.

[0025] In this embodiment, a load-parameter mapping library is maintained. The current load range (light, medium, heavy load) is determined based on the predicted load value, and the corresponding baseline PID parameters (P0, I0, D0) are extracted. These baseline parameters serve as the initial particle swarm for the particle swarm optimization algorithm. The objective function (fitness function) of the optimization algorithm comprehensively considers the predicted load change rate and the current rotational speed deviation. The algorithm performs rapid iterative optimization (time <50ms), and its optimization direction is guided by the load prediction trend: if the predicted load is a sudden increase, the optimization direction tends to increase the proportional coefficient P and the integral coefficient I to enhance the system's response capability in advance; if the predicted load is a sudden decrease, the proportional coefficient P tends to decrease to suppress potential overshoot in advance. Finally, a set of optimized PID parameters (P', I', D') is output.

[0026] Furthermore, the load range is defined as follows: light load is less than 30% of the rated load, medium load is 30% to 70% of the rated load, and heavy load is greater than 70% of the rated load.

[0027] Furthermore, in step S2, the signal acquisition and processing also includes high-voltage anti-interference compensation, which specifically includes: (1) The collected speed and current signals are processed by differential transmission and Kalman filtering algorithm to achieve an electromagnetic interference suppression rate of more than 85%. (2) When outputting control parameters to the drive unit, a voltage spike suppression algorithm is executed to ensure that the output voltage fluctuation of the inverter does not exceed the motor insulation tolerance threshold under a high voltage environment of 1140V or 3300V.

[0028] In this embodiment, sensors installed on the inverter output side and the motor side collect three-phase output current, motor speed, and DC bus voltage signals in real time at a sampling frequency of 10kHz. Considering the strong electromagnetic interference environment underground, the raw signal first passes through a differential signal transmission circuit to suppress common-mode interference. Subsequently, the signal is sent to a digital processing unit based on the Kalman filter algorithm to further filter out random noise and interference in specific frequency bands, ensuring the authenticity of the feedback signal and controlling the signal acquisition error rate to within 2%. The anti-interference compensation in this step is crucial to ensuring the accuracy of subsequent load prediction and parameter optimization.

[0029] Furthermore, it also includes integrated security coordination control, which specifically includes: (1) Real-time monitoring of inverter power module temperature and system insulation resistance; (2) When the temperature of the power module or the insulation resistance exceeds the safe range, the PID controller parameters are automatically switched to the preset safe parameter set and an alarm is triggered.

[0030] In this embodiment, key status parameters such as the core temperature of the inverter power module and the system's insulation resistance to ground are monitored in real time. Once abnormal conditions such as excessive temperature or decreased insulation resistance are detected, the system will ignore the current optimized parameters and immediately switch to a set of preset, conservative, and safe safety PID parameters to ensure that the equipment can still operate smoothly under fault or extreme conditions or safely shut down, and send alarm information to the host computer.

[0031] Furthermore, the benchmark PID parameter library is constructed separately for different types of mining equipment, including at least scraper conveyors and hoists.

[0032] Furthermore, in step S5, the optimized PID parameters are transmitted to the IGBT drive unit of the frequency converter through an optical fiber isolation interface.

[0033] In this embodiment, the controller parameters are updated to drive the power module to output voltage and frequency that match the predicted load. While the control is executed, the system continuously monitors the actual motor speed and load changes, comparing them with the predicted values ​​from step S3. If, over a period of time, the average deviation between the predicted and actual values ​​exceeds a set threshold (e.g., 5%), the online fine-tuning mechanism of the LSTM model is activated. New operating data is used to slightly correct the model parameters, allowing them to better adapt to the current operating characteristics of the equipment and achieve self-evolution of prediction accuracy.

[0034] Alternative implementation plan

[0035] Although the above embodiments describe preferred technical combinations in detail, the core architecture of the present invention allows for the substitution of some technical modules to form different implementation schemes under different application scenarios, cost constraints or performance requirements.

[0036] Load prediction algorithm alternatives: Option A (BP Neural Network Prediction): In applications with relatively simple load fluctuation patterns and high cost sensitivity, a BP neural network can be used to replace the LSTM network for load prediction. The BP network has a simple structure, and training and deployment costs can be reduced by approximately 10%. However, its ability to model long-term time-series dependencies is weak, leading to a potential drop in prediction accuracy to around 85% for the next 100-200ms. It is suitable for scenarios with significant load periodicity and non-drastic changes.

[0037] Option B (Sliding Window Prediction): For simple scenarios where load changes are extremely slow or no historical data is available, the sliding window prediction method can be used. This method predicts future trends based on the average load over a recent window (e.g., 500ms) or by linear extrapolation. This method has extremely low algorithm complexity and requires no model training, but it cannot identify or predict sudden load changes and is only suitable for situations where the load is almost constant.

[0038] Alternatives to PID parameter optimization algorithms: Option C (Fuzzy PID): Fuzzy control rules can be used to replace particle swarm optimization for parameter adjustment. Fuzzy PID has an extremely fast response speed (≤30ms), making it particularly suitable for applications with extremely high real-time requirements. However, its control performance heavily relies on a fuzzy rule base developed by expert experience. When the load fluctuation range exceeds the range covered by its rule base (e.g., ±50% of the rated value), the system has poor robustness and is prone to instability.

[0039] Option D (Expert System PID): An expert system based on "IF-THEN" rules can be built to adjust PID parameters. This option is intuitive to debug and easily integrates the field experience of engineers. The disadvantages are the complexity of building and maintaining the rule base, and the difficulty in adaptively learning the characteristics of new mining equipment or unknown operating conditions, resulting in insufficient flexibility.

[0040] Anti-interference solution alternatives: Option E (Optical Isolation + RC Filter): In environments with weak electromagnetic interference (such as ground test platforms) or where cost control is extremely stringent, opto-isolation combined with a passive RC filter circuit can be used for signal processing at a lower cost. This option can reduce hardware costs by about 30%, but its ability to suppress high-frequency interference is limited, and the overall interference suppression rate may only reach about 60%.

[0041] Option F (Adaptive Filtering): In complex electromagnetic environments where interference suppression rates are extremely high (≥90%), adaptive filtering algorithms such as LMS (Least Mean Square) can be used to replace Kalman filtering. Adaptive filtering can dynamically track interference characteristics and has excellent performance, but the algorithm has a large computational load and usually requires an additional dedicated processor (such as FPGA) to ensure real-time performance, resulting in an increase in system cost of about 25%. Example 2

[0042] To implement the above-mentioned adaptive PID control method for high-voltage mining frequency converters based on load prediction, this invention also includes an adaptive PID control system for high-voltage mining frequency converters based on load prediction, the system comprising: The load prediction module is used to acquire current signals, extract abrupt change features, and use an LSTM model to predict load changes. The adaptive PID parameter adjustment module has a built-in load-PID parameter mapping library and particle swarm optimization unit, which is used to dynamically optimize PID parameters based on prediction results. The mining anti-interference compensation module is used to perform Kalman filtering and differential processing on the acquired speed, current and voltage signals, and to perform voltage spike suppression. The integrated coordination control module coordinates the operation of the load prediction module, the adaptive PID parameter adjustment module, and the mine anti-interference compensation module, and switches to the safety control mode when the equipment status is abnormal.

[0043] In this embodiment, the load prediction module includes a high-speed AD sampling unit, a wavelet transform feature extraction unit, and an LSTM neural network prediction unit. The load prediction module is responsible for performing the current acquisition in step S2 and the feature extraction and load prediction functions in step S3.

[0044] The adaptive PID parameter adjustment module includes a non-volatile memory for storing load-parameter mapping libraries for different load types; and a high-speed computing unit for running the particle swarm optimization algorithm, which receives the prediction results from the load prediction module and the speed deviation from the feedback channel, and performs parameter optimization in step S4.

[0045] The mining anti-interference compensation module includes a differential signal conditioning circuit, a high-speed digital signal processor (DSP) and related filtering algorithms. The mining anti-interference compensation module is responsible for performing anti-interference processing on all input sensor signals (current, speed, voltage) in step S2, and integrates voltage spike suppression logic at the control command output terminal.

[0046] The integrated coordination control module is controlled by the main control MCU or CPU. The integrated coordination control module is responsible for scheduling the load prediction module, the adaptive PID parameter adjustment module, and the mine anti-interference compensation module to work in an orderly manner, realizing the timing coordination of "prediction-optimization-execution". At the same time, it obtains the status information of the power module temperature sensor and insulation monitor through the I / O interface or communication bus, and executes safety monitoring and parameter switching logic.

[0047] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An adaptive PID control method for high-voltage mining frequency converters based on load prediction, characterized in that, Includes the following steps: S1, System initialization: Configure the inverter rated voltage, motor rated power, load type and load prediction threshold, and load the benchmark PID parameter library and pre-trained Long Short-Term Memory (LSTM) network model corresponding to the load type. S2, Signal Acquisition and Processing: Real-time acquisition of the inverter's output current, motor speed, and bus voltage signals, followed by Kalman filtering and differential amplification to obtain the feedback signal after anti-interference processing; S3, Load Prediction and Judgment: Perform wavelet transform on the current signal after anti-interference processing to extract current change features; when the extracted current change features exceed a preset threshold, activate the LSTM model to predict the load change trend and load change rate within the next 100ms to 200ms time window. S4, Parameter Dynamic Optimization: Based on the load change trend, the load change rate, and the current speed deviation, the parameters of the PID controller are dynamically adjusted using the particle swarm optimization algorithm to generate an optimized PID parameter set; the adjustment is completed before the actual load change, and the total time is less than 50ms. S5, Control Execution and Feedback: The optimized PID parameter set is applied to the inverter drive unit to control the power module output; at the same time, the actual load is compared with the predicted load, and when the deviation exceeds the set threshold, the parameters of the LSTM model are corrected online.

2. The adaptive PID control method for high-voltage mine frequency converters based on load prediction according to claim 1, characterized in that, In step S3, the load prediction and determination specifically includes: (1) Acquire the inverter output current signal at a sampling frequency of 10kHz; (2) Analyze the current signal using wavelet transform and calculate the slope of the current change; (3) When the current change slope is detected to be greater than 5A / ms, it is determined to be a precursor to a sudden load change; (4) Input the historical and current data sequence including the current mutation characteristics into the LSTM model, and the LSTM model outputs the load prediction value and change rate within a specified future time.

3. The adaptive PID control method for high-voltage mine frequency converters based on load prediction according to claim 1, characterized in that, In step S4, the dynamic optimization of parameters specifically includes: (1) A load-parameter mapping library is preset, wherein the mapping library defines the baseline PID parameters according to the load range, and the load range includes at least: light load range, medium load range and heavy load range; (2) Using the load prediction value as the query basis, obtain the baseline PID parameters of the corresponding load range from the mapping library; (3) Using the benchmark PID parameters as the initial population, a fitness function is constructed by combining the load change rate and the current speed deviation, and the particle swarm optimization algorithm is run for iterative optimization; When a sudden increase in load is predicted, the optimization objective is to increase the proportional coefficient P and the integral coefficient I; when a sudden decrease in load is predicted, the optimization objective is to decrease the proportional coefficient P.

4. The adaptive PID control method for high-voltage mine frequency converters based on load prediction according to claim 3, characterized in that, The load range is defined as follows: light load is less than 30% of the rated load, medium load is 30% to 70% of the rated load, and heavy load is greater than 70% of the rated load.

5. The adaptive PID control method for high-voltage mine frequency converters based on load prediction according to claim 1, characterized in that, In step S2, the signal acquisition and processing further includes high-voltage anti-interference compensation, which specifically includes: (1) The collected speed and current signals are processed by differential transmission and Kalman filtering algorithm to achieve an electromagnetic interference suppression rate of more than 85%. (2) When outputting control parameters to the drive unit, a voltage spike suppression algorithm is executed to ensure that the output voltage fluctuation of the inverter does not exceed the motor insulation tolerance threshold under a high voltage environment of 1140V or 3300V.

6. The adaptive PID control method for high-voltage mine frequency converters based on load prediction according to claim 1, characterized in that, It also includes integrated security coordination control, which specifically includes: (1) Real-time monitoring of inverter power module temperature and system insulation resistance; (2) When the temperature of the power module or the insulation resistance exceeds the safe range, the PID controller parameters are automatically switched to the preset safe parameter set and an alarm is triggered.

7. The adaptive PID control method for high-voltage mine frequency converters based on load prediction according to claim 1, characterized in that, The benchmark PID parameter library is constructed separately for different types of mining equipment, including at least scraper conveyors and hoists.

8. The adaptive PID control method for high-voltage mine frequency converters based on load prediction according to claim 1, characterized in that, In step S5, the optimized PID parameters are transmitted to the IGBT drive unit of the frequency converter through an optical fiber isolation interface.

9. An adaptive PID control system for high-voltage mining frequency converters based on load prediction, characterized in that, The system for implementing the method according to any one of claims 1 to 8, the system comprising: The load prediction module is used to acquire current signals, extract abrupt change features, and use an LSTM model to predict load changes. The adaptive PID parameter adjustment module has a built-in load-PID parameter mapping library and particle swarm optimization unit, which is used to dynamically optimize PID parameters based on prediction results. The mining anti-interference compensation module is used to perform Kalman filtering and differential processing on the acquired speed, current and voltage signals, and to perform voltage spike suppression. The integrated coordination control module coordinates the operation of the load prediction module, the adaptive PID parameter adjustment module, and the mine anti-interference compensation module, and switches to the safety control mode when the equipment status is abnormal.

Images