Plunger-type fracturing pump control system

Abstract

The application discloses a plunger type fracturing pump control system and relates to the technical field of fracturing pump control. In the system operation, through real-time collection of plunger displacement signals, pump cavity pressure signals, inlet and outlet valve position signals, driving motor current and rotating speed signals, and pump body vibration and temperature signals, a multi-source sensing acquisition module is connected, a rectifier is arranged in the sensor power supply link to perform rectification and voltage stabilization treatment on the input power supply, control instructions for driving the execution mechanism are generated, the multi-source sensing acquisition module is connected to perform time-frequency domain feature extraction and operation state mode recognition on the collected signals, and the degradation trend of the plunger sealing wear and the valve seat leakage rate is predicted based on the recognition result. The self-adaptive regulation and control module and the state diagnosis and prediction module are connected respectively to perform balanced distribution of the load pressure of each pump and collaborative control of the total displacement under the multi-pump parallel operation condition, and maintenance decision and operation strategy adjustment are performed according to the state prediction result.

Description

Technical Field

[0001] This invention relates to the field of fracturing pump control technology, specifically to a plunger-type fracturing pump control system. Background Technology

[0002] In the field of oil and gas field enhancement and renovation, fracturing is a key technology for achieving efficient reservoir development, and the performance of its core equipment, the plunger fracturing pump, directly determines operational efficiency and construction safety. The plunger fracturing pump pressurizes fracturing fluid to the formation through the reciprocating motion of the plunger. Its control system needs to precisely regulate key parameters such as pumping pressure, flow rate, and stroke rate. In modern well site equipment, to ensure stable power supply quality to the actuators and sensors within the control system, rectifiers are typically configured in some control links to shape the external power supply. Inductors are also placed in the PWM circuit controlling the drive motor current to suppress electromagnetic disturbances, thereby improving the stability of plunger movement regulation. As the core unit for realizing intelligent operation and status monitoring of the fracturing pump, the control accuracy and response speed of the plunger fracturing pump control system have a decisive impact on ensuring the quality of fracturing operations.

[0003] In existing technologies, plunger-type fracturing pump control systems mostly employ PID control strategies based on fixed thresholds, which are ill-suited to adapt to the sudden load changes and nonlinear disturbances caused by complex downhole conditions. Traditional control systems exhibit lag in their coordinated control of pump and valve opening / closing sequences and plunger movement phases, easily leading to pressure pulsations and water hammer, thus accelerating pump body fatigue. Furthermore, existing systems generally lack signal stabilization mechanisms matched with inductor-type anti-interference components. Under high vibration and high impact environments, noise accumulation in some sensor links makes it difficult for the system to achieve real-time diagnosis of pump unit operating status, thus failing to effectively predict potential faults such as plunger seal wear or valve seat leakage. This can easily lead to unplanned downtime due to component failure during long-term continuous operation. In addition, traditional control architectures do not adequately consider load distribution optimization during multi-pump parallel operation, resulting in insufficient flow and pressure coordination among pumps, severely impacting the overall efficiency and stability of fracturing operations.

[0004] Therefore, plunger-type fracturing pumps face multiple technical challenges under complex operating conditions, such as insufficient control precision, lack of state awareness, and low collaborative efficiency. Furthermore, existing systems lack effective utilization of rectifier and inductor structures in terms of power shaping, drive anti-interference, and signal stabilization processing. There is an urgent need for a control system solution that can achieve precise regulation, intelligent diagnosis, and collaborative optimization. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a plunger-type fracturing pump control system, which solves the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a plunger-type fracturing pump control system, comprising a multi-source sensor acquisition module, an adaptive control module, a state diagnosis and prediction module, and a collaborative optimization decision-making module;

[0007] The modules can achieve one-way or two-way data interaction connection through industrial communication bus or data link;

[0008] The multi-source sensor acquisition module is used to acquire plunger displacement signals, pump chamber pressure signals, inlet and outlet valve position signals, drive motor current and speed signals, and pump body vibration and temperature signals in real time.

[0009] The adaptive control module is connected to the multi-source sensor acquisition module, which is used to perform feedforward compensation of the plunger motion trajectory and dynamic correction of the pump valve opening and closing phase based on the acquired real-time signals, and to generate control commands to drive the actuator.

[0010] The status diagnosis and prediction module is connected to the multi-source sensor acquisition module to extract time-frequency domain features and identify operating status patterns of the acquired signals, and to predict the degradation trend of plunger seal wear and valve seat leakage rate based on the identification results.

[0011] The collaborative optimization decision-making module is connected to the adaptive control module and the state diagnosis and prediction module respectively. It is used to balance the load pressure of each pump and coordinate the total discharge under the condition of multi-pump parallel operation, and to make maintenance decisions and adjust the operation strategy based on the state prediction results.

[0012] Preferably, the multi-source sensing acquisition module includes a high-precision magnetostrictive displacement sensor, a high-frequency dynamic pressure sensor, a Hall effect valve position sensor, a motor current and speed sensor, and a triaxial vibration and infrared temperature sensor.

[0013] A high-precision magnetostrictive displacement sensor is installed on the side of the plunger rod to detect the real-time displacement of the plunger and output an analog voltage signal of 4 millivolts per millimeter. Its power supply terminal is equipped with a rectifier to rectify and regulate the external power supply.

[0014] A high-frequency dynamic pressure sensor is embedded in the pump cavity wall. It adopts a MEMS piezoresistive sensing element with a sampling frequency of no less than 20 kHz and a range covering 0 to 150 MPa.

[0015] Hall effect valve position sensors are arranged outside the inlet and outlet valve cores to detect the opening and closing status of the valve cores and output switching signals.

[0016] The motor current and speed sensors are integrated inside the drive motor controller. The current measurement accuracy is 0.5% of the full scale, and the speed measurement resolution reaches 1 revolution per minute.

[0017] The triaxial vibration and infrared temperature sensors are fixed to the pump body surface via a magnetic base. The vibration measurement frequency range is 5 Hz to 10 kHz, and the temperature measurement range is 0 degrees Celsius to 120 degrees Celsius. The vibration signal front-end filtering network includes an inductor as a filtering element to suppress common-mode interference.

[0018] Preferably, the adaptive control module includes a feedforward compensation unit, a phase correction unit, and a command generation unit;

[0019] The feedforward compensation unit receives the plunger displacement signal and pump chamber pressure signal transmitted by the multi-source sensor acquisition module, and calculates the compensation amount of the plunger motion acceleration through a nonlinear model based on the pump chamber fluid compressibility and pipeline flow resistance characteristics.

[0020] The phase correction unit receives inlet and outlet valve position signals and plunger displacement signals. By detecting the phase deviation between the actual opening time of the valve core and the dead point position of the plunger, it dynamically adjusts the speed setpoint of the drive motor to keep the phase difference between the opening time of the valve core and the beginning time of the plunger return stroke within the range of 5 to 8 degrees.

[0021] The instruction generation unit integrates an improved fuzzy PID controller. This controller uses the compensation amount output by the feedforward compensation unit and the speed correction amount output by the phase correction unit as feedforward inputs, and the deviation between the real-time pump chamber pressure and the set pressure as feedback inputs. After reasoning from the fuzzy rule base and self-tuning of the PID parameters, it outputs a PWM pulse signal with an adjustable duty cycle to drive the actuator.

[0022] Preferably, the fuzzy rule base of the improved fuzzy PID controller contains 49 rules, the input variables are pressure deviation and deviation change rate, its universe of discourse is divided into 7 fuzzy subsets, and the output variables are the correction amounts of the proportional coefficient, integral coefficient and derivative coefficient of the PID controller.

[0023] The controller uses the Mamdani method for fuzzy inference and the centroid method for defuzzification. The controller sampling period is set to 1 millisecond.

[0024] Preferably, the state diagnosis prediction module includes a feature extraction unit, a pattern recognition unit, and a degradation prediction unit;

[0025] The feature extraction unit performs wavelet packet decomposition on the vibration signal and extracts the energy proportion of 8 frequency bands as the vibration feature vector. At the same time, it performs peak detection and waveform factor calculation on the pump cavity pressure signal and extracts the pressure pulsation coefficient and waveform distortion rate as the pressure feature vector.

[0026] The pattern recognition unit uses a support vector machine classifier, which inputs a combined feature vector consisting of vibration feature vector and pressure feature vector into a trained classification model to identify the pump operating status as normal, slight piston wear, early valve seat leakage, or abnormal cavitation in the pump chamber, with a classification accuracy of no less than 98%.

[0027] Based on the identified abnormal conditions, the degradation prediction unit uses a time series prediction algorithm, taking historical wear or leakage rate data as input, and predicts the growth curve of plunger seal wear and the upward trend of valve seat leakage rate in the next 24 hours through an exponential smoothing model.

[0028] Preferably, the time series prediction algorithm adopts a cubic exponential smoothing model, whose smoothing coefficient is adaptively adjusted by the mean square error minimization criterion. The model is retrained every 4 hours, and the training data spans 30 days.

[0029] Preferably, the collaborative optimization decision module includes a load balancing unit, a displacement coordination unit, and a maintenance decision unit;

[0030] When multiple pumps are running in parallel, the load balancing unit collects the outlet pressure and drive motor power of each pump in real time. Taking the minimization of the pressure deviation of each pump as the objective function, it dynamically adjusts the stroke set value of each pump through the gradient descent algorithm to keep the total pressure imbalance of the parallel pump group within 3%.

[0031] The displacement coordination unit calculates the displacement deviation and generates a stroke compensation command based on the total displacement demand and the actual displacement of each pump. This command is superimposed on the stroke setting value output by the load balancing unit and then sent to each adaptive control module.

[0032] The maintenance decision unit receives wear and leakage rate prediction data transmitted by the status diagnosis prediction module. When the predicted wear exceeds the preset threshold of 2 mm or the leakage rate exceeds 5 liters per minute, an early warning signal is generated and it is recommended to replace the component in the next maintenance window. At the same time, the load distribution ratio of the current pump is automatically adjusted to reduce its operating intensity.

[0033] Preferably, the gradient descent algorithm employs a momentum acceleration strategy, with an initial learning rate of 0.01, a momentum factor of 0.9, an iteration step size of 100 milliseconds, and updates the set values ​​for each pump stroke after every 10 iterations.

[0034] Preferably, the system operates within a multi-timescale hierarchical framework, which includes a strategic layer, a tactical layer, and an operational layer.

[0035] The strategic layer is used to set long-term supply and demand goals and game theory frameworks on a monthly timescale;

[0036] The tactical layer is used to plan major logistics routes and inventory strategies on a weekly timescale.

[0037] The operation layer is used to perform equilibrium analysis output by the collaborative optimization decision module and state prediction output by the state diagnosis prediction module on a daily time scale.

[0038] Preferably, the system also includes a data communication bus for connecting the multi-source sensor acquisition module, the adaptive control module, the state diagnosis and prediction module, and the collaborative optimization decision-making module;

[0039] The data communication bus adopts the industrial Ethernet protocol, with a data transmission cycle of 10 milliseconds, and supports dual-ring network redundancy design.

[0040] This invention provides a plunger-type fracturing pump control system, which has the following beneficial effects:

[0041] (1) During system operation, the system collects the plunger displacement signal, pump chamber pressure signal, inlet and outlet valve position signal, drive motor current and speed signal, and pump body vibration and temperature signal in real time. It connects to the multi-source sensor acquisition module, sets a rectifier in the sensor power supply link to rectify and stabilize the input power supply, and generates control commands for driving the actuator. It connects to the multi-source sensor acquisition module to extract time and frequency domain features and identify the operating status mode of the collected signals. Based on the identification results, it predicts the degradation trend of plunger seal wear and valve seat leakage rate. It connects to the adaptive control module and the status diagnosis prediction module respectively to balance the load pressure of each pump and coordinate the total displacement control under the condition of multi-pump parallel operation. It makes maintenance decisions and adjusts the operation strategy based on the status prediction results.

[0042] (2) The plunger-type fracturing pump control system proposed in this invention achieves a comprehensive and fine-grained closed-loop mechanism for acquiring and controlling operational information, which is difficult to achieve with previous control architectures, by constructing a multi-source sensor acquisition module, an adaptive control module, a state diagnosis and prediction module, and a collaborative optimization decision-making module. The system introduces a rectifier in the sensor power supply link to stabilize the power quality, and configures inductors in the vibration signal conditioning circuit and the PWM drive circuit of the control actuator to effectively suppress high-frequency interference, so that the multi-source signal acquisition and execution unit control has higher stability and anti-interference capability. Compared with the traditional system that relies on a single pressure signal and a fixed threshold for control, this invention ensures the real-time and accurate acquisition of multi-dimensional data such as plunger displacement, pump chamber pressure, valve position timing, vibration spectrum, and temperature field, providing a solid foundation for subsequent control and diagnosis, and fundamentally solving the problems of control lag and weak state recognition capability caused by insufficient information dimensions in the prior art.

[0043] (3) This invention significantly enhances the system's adaptability to downhole operating condition fluctuations, valve opening and closing deviations, and plunger stroke disturbances through an adaptive control chain composed of feedforward compensation, phase correction, and fuzzy PID self-tuning strategies. The feedforward compensation unit performs dynamic acceleration compensation based on a nonlinear fluid model, the phase correction unit corrects the deviation between plunger movement and valve core opening and closing in real time, and the fuzzy PID controller automatically adjusts the proportional, integral, and derivative parameters, so that the pumping pressure and displacement remain stable and respond quickly under complex operating conditions. In addition, the control power supply with the addition of a rectifier has higher quality, and the drive current with the addition of an inductor is more stable, so that the control signal transmission link maintains reliability in the well site environment with strong vibration and high impact, thereby making up for the defects of traditional control methods that are susceptible to interference and difficult to cope with nonlinear fluctuations, effectively suppressing common engineering problems such as pressure pulsation, water hammer, and pump body impact load, and improving the control accuracy and action coordination of the plunger fracturing pump.

[0044] (4) The state diagnosis prediction module and collaborative optimization decision module of this invention construct a pump group-level intelligent prediction and collaborative control system rarely seen in the industry. Based on wavelet packet energy characteristics, pressure waveform components and support vector machine identification strategies, the system realizes real-time identification of fault symptoms such as plunger wear, valve seat leakage and cavitation, and combines exponential smoothing prediction algorithm to make forward-looking judgment on the degradation trend of key components. The collaborative optimization unit further realizes dynamic optimization allocation of pressure and flow when multiple pumps are running in parallel through gradient descent load balancing and displacement collaborative regulation. Due to the integration of inductor-type anti-interference structure in the communication bus and signal link, the system has stronger anti-electromagnetic interference capability in long-distance data transmission and pump group collaborative scheduling. Compared with the existing method of pump group scheduling that relies solely on manual experience or fixed rules, this invention not only improves operational stability, but also significantly reduces the risk of unplanned downtime caused by component over-limit operation, and realizes an overall improvement in predictive maintenance, collaborative regulation and safe operation capabilities. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the overall technical architecture of the plunger-type fracturing pump control system proposed in this invention;

[0046] Figure 2 This is a schematic diagram of the core principle framework of the adaptive control module in this invention;

[0047] Figure 3 This is a schematic diagram of the multi-level interaction relationship and data flow between the state diagnosis and prediction module and the collaborative optimization decision-making module in this invention; Detailed Implementation

[0048] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0049] Example 1

[0050] This invention provides a plunger-type fracturing pump control system; please refer to [link / reference]. Figure 1 It includes a multi-source sensing acquisition module, an adaptive control module, a state diagnosis and prediction module, and a collaborative optimization decision-making module;

[0051] The multi-source sensor acquisition module is used to acquire plunger displacement signals, pump chamber pressure signals, inlet and outlet valve position signals, drive motor current and speed signals, and pump body vibration and temperature signals in real time.

[0052] The adaptive control module is connected to the multi-source sensor acquisition module, which is used to perform feedforward compensation of the plunger motion trajectory and dynamic correction of the pump valve opening and closing phase based on the acquired real-time signals, and to generate control commands to drive the actuator.

[0053] The status diagnosis and prediction module is connected to the multi-source sensor acquisition module to extract time-frequency domain features and identify operating status patterns of the acquired signals, and to predict the degradation trend of plunger seal wear and valve seat leakage rate based on the identification results.

[0054] The collaborative optimization decision-making module is connected to the adaptive control module and the state diagnosis and prediction module respectively. It is used to balance the load pressure of each pump and coordinate the total discharge under the condition of multi-pump parallel operation, and to make maintenance decisions and adjust the operation strategy based on the state prediction results.

[0055] In this embodiment, the system comprises a multi-source sensor acquisition module, an adaptive control module, a condition diagnosis and prediction module, and a collaborative optimization decision-making module. These modules interact and transmit commands via an industrial Ethernet bus. The multi-source sensor acquisition module is deployed on the fracturing pump body and drive unit, used to collect real-time physical quantities such as plunger displacement, pump chamber pressure, inlet and outlet valve positions, motor current and speed, pump body vibration, and temperature. The adaptive control module is installed in the control cabinet; its core function is to generate feedforward compensation and phase correction quantities based on real-time sensor data and output control commands to drive the actuators. The condition diagnosis and prediction module runs on an industrial server platform, using feature extraction and pattern recognition of historical and real-time data to predict component degradation trends. The collaborative optimization decision-making module is integrated into the system's main controller, responsible for load balancing, displacement coordination, and maintenance strategy formulation under multi-pump parallel operation conditions.

[0056] Example 2

[0057] This embodiment is an explanation based on Embodiment 1. Please refer to it. Figure 2 and Figure 3 Specifically, the multi-source sensing acquisition module includes a high-precision magnetostrictive displacement sensor, a high-frequency dynamic pressure sensor, a Hall effect valve position sensor, a motor current and speed sensor, and a triaxial vibration and infrared temperature sensor. The high-precision magnetostrictive displacement sensor is fixed to the side of the plunger rod using a stainless steel clamp. Its measuring rod is parallel to the plunger rod axis, with a sensing range of 0 to 1200 mm. The output analog voltage signal has a sensitivity of 4 mV / mm, and the signal is sampled by a 24-bit analog-to-digital converter before being transmitted to the adaptive control module.

[0058] The high-frequency dynamic pressure sensor employs a MEMS piezoresistive sensing element. Its diaphragm is flush with the pump cavity wall and sealed with high-temperature epoxy resin. The measurement range covers 0 to 150 MPa, with a sampling frequency set to 20 kHz. The output signal is conditioned by a charge amplifier and then sent to an anti-aliasing filter. The Hall effect valve position sensor is located on the outer housing of the inlet and outlet valve cores. Its permanent magnet is linked to the valve core. When the valve core opens or closes, the Hall element outputs a switching signal, which is transmitted to the control unit via an opto-isolator. The motor current and speed sensors are integrated into the drive motor controller. Current measurement uses a closed-loop Hall principle with an accuracy of 0.5% of full scale. Speed ​​measurement is achieved through encoder pulse counting with a resolution of 1 revolution per minute. The triaxial vibration and infrared temperature sensors are fixed to the pump body's stress-concentrated areas via magnetic bases. The vibration measurement frequency range is 5 Hz to 10 kHz, and the temperature measurement range is 0°C to 120°C. Data is uploaded via the CAN bus protocol.

[0059] The specific implementation of the adaptive control module is shown in the attached figure. Figure 2 As shown, its hardware is based on a multi-core processor architecture, and the software runs a real-time operating system. The module is internally divided into a feedforward compensation unit, a phase correction unit, and an instruction generation unit. The feedforward compensation unit receives the plunger displacement signal and pump chamber pressure signal transmitted from the multi-source sensor acquisition module, and calculates the plunger motion acceleration compensation amount using a nonlinear model based on the pump chamber fluid compressibility and pipeline flow resistance characteristics. This model considers the fluid elastic modulus and the flow channel damping coefficient, and its differential equation is described as: ;

[0060] The parameters are defined as follows:

[0061] P represents the real-time pressure value inside the plunger cavity, in MPa;

[0062] t represents time, in seconds (s);

[0063] β represents the equivalent bulk elastic coefficient of fracturing fluid, which is used to characterize the relationship between volume change and pressure change when fracturing fluid is compressed;

[0064] V represents the equivalent working volume of the plunger cavity, in m³;

[0065] R represents the overall flow resistance characteristic coefficient of the pipeline, which is used to characterize the energy loss characteristics of fluid in the pipeline.

[0066] Q represents the instantaneous flow rate of fracturing fluid, in m³ / s;

[0067] This represents the rate of change of pressure over time;

[0068] This represents the rate of change of volume over time.

[0069] All of the above parameters can be calculated using on-site sensor data, equipment structural parameters, and system calibration parameters.

[0070] The range of values ​​for the above parameters can be determined through experimental calibration based on different fracturing conditions and equipment models.

[0071] The fluid volume change rate is proportional to the derivative of the plunger displacement, and the compensation output is an acceleration adjustment value in meters per second squared. The phase correction unit synchronously acquires inlet and outlet valve position signals and plunger displacement signals, identifies the actual valve core opening time using a zero-point detection algorithm, and calculates the phase deviation with the plunger dead point position. The dynamic correction logic uses a target phase difference of 5 to 8 degrees and adjusts the drive motor speed setpoint through a proportional regulator with an adjustment period of 10 milliseconds. The instruction generation unit integrates an improved fuzzy PID controller, whose inputs include feedforward compensation, phase correction, and real-time pressure deviation. The controller's fuzzy rule base contains 49 rules. The universe of discourse for the input variables, pressure deviation and deviation change rate, is divided into seven fuzzy subsets: negative large, negative medium, negative small, zero, positive small, positive medium, and positive large. The output variables are the correction values ​​for the proportional coefficient, integral coefficient, and derivative coefficient. Fuzzy inference uses the Mamdani method, and defuzzification uses the centroid method. The controller sampling period is 1 millisecond. The final output PWM pulse signal has a duty cycle range of 0 to 100% and a frequency of 20 kHz, which drives the IGBT power module through optocoupler isolation.

[0072] The specific implementation of the state diagnosis and prediction module is attached. Figure 3As shown, it is deployed on an industrial server, and data storage uses a time-series database. The module consists of a feature extraction unit, a pattern recognition unit, and a degradation prediction unit. The feature extraction unit performs four-level wavelet packet decomposition on the vibration signal, extracts the energy proportion of eight frequency bands to form a vibration feature vector, and calculates the ratio of peak value to root mean square value of the pump chamber pressure signal as a waveform factor, and derives the pressure pulsation coefficient. The pattern recognition unit uses a support vector machine classifier with a radial basis function kernel. The training data covers four types of samples: normal state, slight piston wear state, early valve seat leakage state, and abnormal pump chamber cavitation state, with a classification accuracy of no less than 98%. Based on the recognition results, the degradation prediction unit uses a cubic exponential smoothing model for time series prediction. The smoothing coefficient of this model is adaptively adjusted according to the mean square error minimization criterion, and it is retrained every 4 hours, with a training data time span of 30 days. The prediction output is the piston seal wear growth curve and the valve seat leakage rate increase trend in the next 24 hours. The data is transmitted to the collaborative optimization decision module in JSON format.

[0073] In this embodiment, the specific implementation of the collaborative optimization decision-making module is combined with the appendix. Figure 3 This module operates within the multi-task scheduling environment of the main controller. It includes a load balancing unit, a displacement coordination unit, and a maintenance decision unit. When multiple pumps are running in parallel, the load balancing unit collects real-time data on the outlet pressure of each pump and the power of its drive motor. Using the minimization of pressure deviation among the pumps as the objective function, it dynamically adjusts the stroke setpoints of each pump through a gradient descent algorithm. The algorithm employs a momentum acceleration strategy, with an initial learning rate of 0.01, a momentum factor of 0.9, and an iteration step size of 100 milliseconds. The setpoints are updated every 10 iterations to ensure that the overall pressure imbalance of the parallel pump group is controlled within 3%. The displacement coordination unit calculates the deviation based on the total displacement requirement and the actual displacement of each pump, generating a stroke compensation command. This command is superimposed on the stroke setpoint output by the load balancing unit in an adder, and the result is sent to each adaptive control module via the PROFIBUS bus. The maintenance decision unit receives wear and leakage rate data transmitted from the status diagnosis and prediction module. When the predicted wear exceeds 2 mm or the leakage rate exceeds 5 liters per minute, it triggers an early warning signal and generates maintenance suggestions. At the same time, the unit automatically adjusts the current pump load distribution ratio, reducing its stroke setpoint by 20% until the maintenance window arrives.

[0074] Example 3

[0075] This embodiment is an explanation based on Embodiment 1. Please refer to it. Figure 1Specifically, this embodiment addresses the multi-pump parallel operation scenario of a plunger fracturing pump control system, further elaborating on the extended implementation of its collaborative optimization decision module. In large-scale fracturing operations, multiple plunger pumps need to work collaboratively to meet high-pressure, high-displacement requirements. This embodiment enhances the overall system stability and energy efficiency by strengthening load balancing and displacement coordination mechanisms.

[0076] Based on Example 1, the load balancing unit of the collaborative optimization decision module introduces a pressure-power coupling model. This model establishes a mapping relationship between the outlet pressure of each pump and the power of the drive motor, and uses the least squares method to fit coefficients, real-time correcting the objective function of the gradient descent algorithm. The load balancing unit collects pressure and power data of each pump every 50 milliseconds, calculates the deviation between the pressure-weighted average and the power-weighted average, and uses this as an additional constraint term for the objective function. During the iteration process, the gradient descent algorithm simultaneously optimizes the pressure balance and power balance, with the learning rate dynamically adjusted according to the magnitude of the deviation, ranging from 0.005 to 0.02. The momentum factor is maintained at 0.9, the iteration step size is shortened to 50 milliseconds, and the stroke setpoint is updated after every 5 iterations. This implementation further reduces the total pressure imbalance of the parallel pump group from 3% to 1.5%, while reducing the motor power fluctuation amplitude by 40%.

[0077] The displacement coordination unit introduces a feedforward-feedback composite strategy in total displacement control. The feedforward section calculates the initial stroke allocation ratio based on the total displacement demand and the rated displacement of each pump. The feedback section monitors the actual displacement of each pump in real time using a high-precision flow meter. Displacement deviations are corrected by a proportional-integral regulator to generate stroke compensation commands. These compensation commands are superimposed on the stroke setpoint output by the load balancing unit in a digital adder. The superposition result is processed by a limiter to ensure that the stroke setpoint does not exceed the pump's mechanical limits. The displacement coordination unit has a data update cycle of 20 milliseconds, and the flow meter signal is transmitted through a 4mA to 20mA current loop, achieving 90% anti-interference capability. This implementation improves the total displacement control accuracy to 98%, and ensures smooth displacement transition without overshoot during multi-pump switching.

[0078] Building upon the early warning mechanism in Example 1, the maintenance decision-making unit adds health status assessment and dynamic load adjustment functions. Health status assessment is based on wear and leakage rate data output by the condition diagnosis and prediction module, combined with the pump's cumulative operating time and number of start-stop cycles, to calculate a comprehensive health index. The health index ranges from 0 to 100, triggering a level-two early warning when the index falls below 70. Dynamic load adjustment adjusts the pump's load distribution ratio in real time according to the health index. For every 1-point decrease in the health index, the pump's stroke setpoint decreases by 0.5%, with a maximum adjustment of 30%. Simultaneously, the maintenance decision-making unit generates a maintenance priority list and recommends optimal maintenance time windows to avoid simultaneous pump shutdowns. This implementation reduces unplanned downtime by 60% and extends the average service life of the pump by approximately 15%.

[0079] The communication architecture in this embodiment adopts a dual-ring network redundancy design, with a data synchronization cycle of 10 milliseconds, and all control commands are timestamped for verification. Through the above enhancements, the system achieves higher-precision pressure balancing, displacement coordination, and intelligent maintenance under complex multi-pump parallel operation conditions, providing reliable technical support for large-scale fracturing operations.

[0080] In this embodiment, by extending the collaborative optimization decision module of the plunger-type fracturing pump control system, the load distribution accuracy, displacement coordination capability, and dynamic adjustment performance driven by health status are significantly improved in multi-pump parallel operation scenarios. The load balancing unit adopts a pressure-power coupling model and a gradient descent algorithm with dynamically adjusted learning rate, reducing the pressure imbalance from 3% to 1.5% and the motor power fluctuation amplitude by 40%. The displacement coordination unit utilizes a feedforward-feedback composite mechanism and a high-speed flow acquisition link to achieve 98% displacement control accuracy and makes the displacement transition during pump group switching smoother. In particular, to ensure the stable operation of the gradient descent optimization process and the flow acquisition link in a strong electromagnetic interference environment, an inductor is added to the current loop input of the displacement coordination unit to improve signal immunity, and a rectifier is configured in the sensor power supply branch of the load balancing unit to shape the power supply and avoid parameter deviations caused by power supply drift; this makes the optimization algorithm robust in harsh well site environments. The maintenance decision unit further introduces a dynamic load adjustment strategy driven by a health index, which automatically reduces the pump participation rate of severely worn pumps and generates an optimal maintenance sequence, thereby reducing unplanned downtime by approximately 60% and extending the average pump life by approximately 15%. The combined effect of these hardware and software enhancement measures enables the system to have higher stability, reliability, and predictive operation capabilities in multi-pump parallel fracturing operations.

[0081] Example 4

[0082] This embodiment is an explanation based on Embodiment 1. Please refer to it. Figure 1 Specifically, this embodiment focuses on the condition diagnosis and prediction module of a plunger-type fracturing pump control system, detailing its adaptive optimization and fault prediction enhancement functions under extreme operating conditions. The condition diagnosis and prediction module improves the accuracy and timeliness of early fault identification by introducing multimodal data fusion and deep learning technologies.

[0083] Based on wavelet packet decomposition in Example 1, the feature extraction unit adds an empirical mode decomposition (EMD) method to handle non-stationary vibration signals. EMD adaptively decomposes the vibration signal into eight intrinsic mode functions (EMFs), calculating the sample entropy and energy entropy of each component as additional features. Simultaneously, a short-time Fourier transform is performed on the pump chamber pressure signal to extract the energy concentration in the 0.5 kHz to 2 kHz frequency band as a cavitation characteristic index. The feature vector dimension is expanded from 8 to 16 dimensions, data standardization uses the Z-Score method, and the processing cycle is 100 milliseconds. This implementation improves the sensitivity of the features to minor faults by 25%, especially accelerating the detection of early valve seat leakage by 3 hours.

[0084] The pattern recognition unit introduces a convolutional neural network (CNN) as an auxiliary classifier on top of the support vector machine (SVM) classifier. The CNN input consists of a time-frequency graph of the vibration signal and a waveform graph of the pressure signal. The network structure includes two convolutional layers, two pooling layers, and one fully connected layer. The outputs of the SVM and CNN are fused using a weighted voting mechanism, with the weights dynamically adjusted based on historical accuracy. The fused classification accuracy improved from 98% to 99.5%, and the misclassification rate for slight wear on the plunger decreased by 50%. The classification model is updated online every 24 hours, and automatic annotation of training data is achieved through an expert system rule base.

[0085] In this embodiment, to address the condition diagnosis requirements of plunger fracturing pumps under extreme operating conditions such as high temperature, high vibration, and frequent changes in fluid disturbance, empirical mode decomposition (EMD), short-time Fourier transform (SFT), multimodal convolutional neural networks (MNN), and dynamic weight fusion mechanisms are introduced. This significantly improves the identifiability of weak fault features and the stability of the classification model. The feature extraction link incorporates an inductor at the vibration signal acquisition front end to form an anti-interference filter network, suppressing high-frequency mechanical noise and electromagnetic disturbances, making the time-frequency features obtained from EMD and SFT more stable and reliable. Simultaneously, a rectifier is added to the sensor power supply branch to stabilize and shape the power supply, improving the power stability of the signal source under severe operating conditions and reducing feature drift caused by power supply fluctuations. Through the above-mentioned joint optimization of software and hardware, this embodiment improves the sensitivity of features to minor leaks, initial wear and early cavitation signs by about 25%, and advances the detection of valve seat leaks by about 3 hours. The recognition framework that integrates support vector machines and convolutional neural networks improves the classification accuracy to 99.5% and reduces the false positive rate by more than 50%, which greatly enhances the system's early fault identification capability and prediction accuracy in continuous fracturing operation scenarios, thereby effectively supporting predictive maintenance strategies with higher timeliness.

[0086] Example 5

[0087] For information on plunger-type fracturing pump control systems, please refer to [link / reference]. Figure 2Specifically, the degradation prediction unit's triple exponential smoothing model, based on Example 1, incorporates a Long Short-Term Memory (LSTM) network for sequence prediction. The LTM network input is a sequence of wear or leakage rates over the past 30 days, with 32 hidden layer units and a loss rate of 0.2%. The prediction results are fused with the exponential smoothing model output using a Kalman filter, and the fusion weights are adaptively calculated based on the prediction error covariance. This implementation reduces the average absolute error of predictions for the next 24 hours to 0.1 mm for wear and 0.2 liters per minute for leakage rate, improving prediction timeliness to 6 hours. The degradation prediction unit retrains the LTM network every 2 hours, with a training data sliding window of 30 days.

[0088] The hardware platform of the condition diagnosis and prediction module has been upgraded to an industrial server with GPU acceleration, achieving an inference latency of less than 10 milliseconds. All diagnostic results are visualized through a digital twin model, displaying the pump's health status and predicted trends in real time. This embodiment, through the aforementioned enhancements, significantly improves the system's fault prediction capabilities under extreme conditions such as high temperature and high vibration, providing more accurate data support for preventative maintenance.

[0089] In this embodiment, based on the cubic exponential smoothing model of the degradation prediction unit, a long short-term memory network is introduced to perform deep sequence prediction of wear and leakage rate. Kalman filtering is used to fuse the prediction results, allowing the predicted values ​​to adaptively adjust according to the error covariance. Through this structure, the average absolute error of wear prediction for the next 24 hours is reduced to 0.1 mm, the leakage rate prediction error is reduced to 0.2 L / min, and the prediction lead is increased to 6 hours. To ensure the stability of the sequence input data, a rectifier is added to the data preprocessing link of the degradation prediction unit to regulate the voltage of the sensor front-end power supply, ensuring that the training data of the long short-term memory network remains reliable under high temperature and high noise conditions. Simultaneously, an inductor is added to the high-speed acquisition channel of the state diagnosis prediction module to form an anti-interference filtering network, effectively suppressing high-frequency noise in vibration and pressure signals, enabling the GPU inference model to obtain accurate input vectors even in noisy environments. Based on the above-mentioned hardware and software enhancement scheme, the system's fault prediction stability and accuracy under extreme conditions are significantly improved, prediction lag and false alarms are significantly reduced, the reliability of preventive maintenance decisions is further enhanced, and the risks of long-term continuous fracturing operations are effectively controlled.

[0090] 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. A plunger fracturing pump control system characterized by: The module comprises a multi-source sensing acquisition module, an adaptive control module, a state diagnosis and prediction module, and a collaborative optimization decision module. The multi-source sensing acquisition module is used for real-time acquisition of plunger displacement signals, pump cavity pressure signals, inlet and outlet valve position signals, driving motor current and speed signals, and pump body vibration and temperature signals. The adaptive control module is connected to the multi-source sensing acquisition module and is used for feedforward compensation of plunger motion trajectory and dynamic correction of pump valve opening and closing phase based on the acquired real-time signals, and generates control instructions for driving the actuator. The state diagnosis and prediction module is connected to the multi-source sensing acquisition module and is used for time-frequency domain feature extraction and running state pattern recognition of the acquired signals, and degradation trend prediction of plunger sealing wear and valve seat leakage rate based on the recognition results. The collaborative optimization decision module is connected to the adaptive control module and the state diagnosis and prediction module, and is used for balanced distribution of pump load pressure and collaborative control of total displacement under multi-pump parallel operation conditions, and maintenance decision and operation strategy adjustment according to the state prediction results.

2. The plunger frac pump control system of claim 1, wherein: The multi-source sensing acquisition module comprises a high-precision magnetostrictive displacement sensor, a high-frequency dynamic pressure sensor, a Hall effect valve position sensor, a motor current and speed sensor, and a three-axis vibration and infrared temperature sensor. The high-precision magnetostrictive displacement sensor is installed on the side of the plunger rod and is used to detect the real-time displacement of the plunger and output an analog voltage signal of 4 millivolts per millimeter. The power supply end is configured with a rectifier to rectify and stabilize the external power supply. The high-frequency dynamic pressure sensor is embedded in the pump cavity wall and uses a MEMS piezoresistive sensing element. Its sampling frequency is not less than 20 kHz, and the range covers 0 to 150 MPa. The Hall effect valve position sensor is arranged outside the inlet and outlet valve core and is used to detect the opening and closing state of the valve core and output a switching value signal. The motor current and speed sensor is integrated in the driving motor controller. The current measurement accuracy is 0.5% of the full range, and the speed measurement resolution is 1 revolution per minute. The three-axis vibration and infrared temperature sensor is fixed to the pump body surface through a magnetic base. The vibration measurement frequency range is 5 Hz to 10 kHz, the temperature measurement range is 0 to 120 degrees Celsius, and the vibration signal front-end filter network contains an inductor as an anti-common-mode interference filter element.

3. The plunger frac pump control system of claim 1, wherein: The adaptive control module comprises a feedforward compensation unit, a phase correction unit, and an instruction generation unit. The feedforward compensation unit receives the plunger displacement signals and pump cavity pressure signals transmitted by the multi-source sensing acquisition module, and calculates the compensation amount of plunger motion acceleration based on the nonlinear model of pump cavity fluid compressibility and pipeline flow resistance characteristics. The phase correction unit receives the inlet and outlet valve position signals and plunger displacement signals, dynamically adjusts the speed given value of the driving motor by detecting the phase deviation of the actual opening time of the valve core and the dead point position of the plunger, so that the phase difference between the valve core opening time and the plunger backstroke starting time is kept within 5 to 8 degrees. The instruction generating unit is integrated with an improved fuzzy PID controller, which takes the compensation quantity output by the feedforward compensation unit and the rotational speed correction quantity output by the phase correction unit as feedforward inputs, and takes the real-time pump cavity pressure deviation from the set pressure as a feedback input, and outputs a PWM pulse signal with adjustable duty cycle to the driving actuator through the reasoning of the fuzzy rule base and the self-tuning of the PID parameters. The rotational speed correction quantity is a motor target rotational speed adjustment value calculated by the phase correction unit based on the valve position signal and the plunger displacement signal.

4. The plunger frac pump control system of claim 3, wherein: The fuzzy rule base of the improved fuzzy PID controller contains 49 rules, the input variables are the pressure deviation and the rate of change of the deviation, and the domain is divided into 7 fuzzy subsets, and the output variables are the correction amounts of the proportional coefficient, the integral coefficient and the differential coefficient of the PID controller. The fuzzy reasoning of the controller uses the Mamdani method, the de-fuzzification uses the gravity method, and the sampling period of the controller is set to 1 millisecond.

5. The plunger frac pump control system of claim 1, wherein: The state diagnosis and prediction module includes a feature extraction unit, a pattern recognition unit and a degradation prediction unit. The feature extraction unit performs wavelet packet decomposition on the vibration signal, extracts 8 frequency band energy ratios as vibration feature vectors, and simultaneously performs peak detection and waveform factor calculation on the pump cavity pressure signal, extracts the pressure pulsation coefficient and the waveform distortion rate as pressure feature vectors. The pattern recognition unit uses a support vector machine classifier, inputs the combined feature vector composed of the vibration feature vector and the pressure feature vector into the trained classification model, identifies the pump operating state as normal state, plunger slight wear state, valve seat early leakage state or pump cavity abnormal cavitation state, and the classification accuracy is not less than 98%. The degradation prediction unit uses a time series prediction algorithm based on the identified abnormal state, takes historical wear amount or leakage rate data as input, and predicts the plunger seal wear growth curve and valve seat leakage rate rising trend in the next 24 hours through an exponential smoothing model.

6. The plunger frac pump control system of claim 5, wherein: The time series prediction algorithm uses a cubic exponential smoothing model, the smoothing coefficient is adaptively adjusted by the least squares error minimization criterion, the model is retrained every 4 hours, and the training data time span is 30 days.

7. The plunger frac pump control system of claim 1, wherein: The collaborative optimization decision module includes a load balancing unit, a displacement coordination unit and a maintenance decision unit. The load balancing unit collects the outlet pressure and driving motor power of each pump in real time when multiple pumps are running in parallel, takes the minimization of the pressure deviation of each pump as the objective function, dynamically adjusts the stroke setting value of each pump through the gradient descent algorithm, and controls the total pressure imbalance of the parallel pump group within 3%. The displacement coordination unit calculates the displacement deviation and generates a stroke compensation instruction based on the total displacement demand and the actual displacement of each pump, which is superimposed with the stroke setting value output by the load balancing unit and then sent to each adaptive control module. The maintenance decision unit receives the wear amount and leakage rate prediction data transmitted by the state diagnosis and prediction module, generates a warning signal and recommends replacing the components in the next maintenance window when the predicted wear amount exceeds the preset threshold of 2 millimeters or the leakage rate exceeds 5 liters per minute, and automatically adjusts the load distribution ratio of the current pump to reduce its operating intensity.

8. The plunger frac pump control system of claim 7, wherein: The gradient descent algorithm adopts a momentum acceleration strategy, the initial value of the learning rate is 0.01, the momentum factor is set to 0.9, and the algorithm iteration step is 100 milliseconds, and the pump stroke setting value is updated after every 10 iterations.

9. The plunger frac pump control system of claim 1, wherein: The system runs in a multi-time scale hierarchical framework, including a strategic layer, a tactical layer and an operational layer. The strategic layer is used to set long-term supply and demand targets and game frameworks at a monthly time scale. The tactical layer is used to plan main logistics trunk lines and inventory strategies at a weekly time scale. The operational layer is used to execute the equilibrium analysis output by the collaborative optimization decision module and the state prediction output by the state diagnosis prediction module at a daily time scale.

10. The plunger frac pump control system of claim 1, wherein: The system also includes a data communication bus for connecting the multi-source sensing acquisition module, the adaptive control module, the state diagnosis prediction module and the collaborative optimization decision module. The data communication bus adopts an industrial Ethernet protocol, the data transmission period is 10 milliseconds, and supports a dual-ring network redundancy design.

Images