A screw machine interlock control method

Abstract

This invention relates to the field of screw compressor control technology, and more particularly to a screw compressor interlocking control method. The method includes connecting the variable frequency drive (VFD) operation signal of a first-stage screw compressor to an automated control system (SIS), connecting the real-time operating current signal of a two-stage screw compressor to a central control system (DCS), and simultaneously acquiring real-time process operating parameters of the two-stage screw compressor. Based on the historical operating data and real-time process operating status of the two-stage screw compressor, the method predicts process disturbance events within a preset time window and extracts the interlocking risk disturbance characteristics corresponding to these events. The method inputs the VFD operation signal, real-time operating current signal, real-time process operating parameters, and interlocking risk disturbance characteristics of the first-stage screw compressor to a pre-constructed screw compressor interlocking risk prediction model to obtain the interlocking risk prediction results for the two-stage compressor within the preset time window. Through this invention, dynamic adaptation of the interlocking control logic to production load, process fluctuations, and equipment operating status is achieved.

Description

Technical Field

[0001] This invention relates to the technical field of screw compressor control, and more particularly to a screw compressor interlocking control method. Background Technology

[0002] In continuous production scenarios such as chemical and petrochemical industries, two-stage screw compressors are core process power equipment, and their operational stability and safety directly determine the continuous operation capability of the production system. To ensure equipment safety, two-stage screw compressors are typically equipped with an automated control system (SIS) for implementing safety interlocking control, and a central control system (DCS) for monitoring equipment operating status. Production processes explicitly require that after the first-stage screw compressor stops, the second-stage screw compressor must simultaneously shut down to avoid damage accidents such as negative pressure in the equipment cavity or rotor dry running.

[0003] Existing interlocking control schemes for two-stage screw compressors typically integrate the operating status signal of the first-stage compressor into a system-wide automation (SIS) system. Based on preset, fixed logic rules, they achieve interlocked shutdown control of both stages. Simultaneously, they monitor operating status parameters such as compressor operating current locally to provide reference for equipment operation and maintenance. However, in actual production applications, existing interlocking control schemes employ a reactive, after-the-fact response mode. They cannot anticipate interlocking risks based on dynamic changes in the production process, leading to delayed, erroneous, or non-responsive interlocking actions. This makes it difficult to reliably ensure the synchronous shutdown safety of both stages, posing a safety hazard to the continuous and stable operation of the equipment. Summary of the Invention

[0004] This invention provides a screw compressor interlocking control method, which can effectively solve the problems in the background art.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A screw compressor interlock control method, wherein the screw compressor is a two-stage screw compressor, comprising a first-stage screw compressor and a second-stage screw compressor connected in sequence, and is equipped with an automatic control system (SIS) for performing safety interlock control, and a central control system (DCS) for monitoring operating status. The method includes: The variable frequency operation signal of the single-stage screw compressor is connected to the automatic control SIS system through a dual-channel redundancy method, and the real-time operating current signal of the two-stage screw compressor is connected to the central control DCS system through a redundant transmission circuit, so as to synchronously collect the real-time process operation parameters of the two-stage screw compressor. Based on the historical operating data and real-time process operating status of the two-stage screw compressor, process disturbance events within a preset time window are predicted, and the interlocking risk disturbance characteristics corresponding to the process disturbance events are extracted. The variable frequency operation signal, real-time operating current signal, real-time process operation parameters, and interlock risk disturbance characteristics of the first-stage screw compressor are input into the pre-constructed screw compressor interlock risk prediction model to obtain the interlock risk prediction results of the two-stage compressor within a future preset time window. Based on the preset interlocking safety threshold system, it is determined whether the interlocking risk prediction result exceeds the limit. If it does not exceed the limit, the original interlocking strategy is maintained. If it exceeds the limit, the risk deviation degree is calculated, and the matching interlocking control level and timing constraints are determined. The inherent parameters of the compressor equipment, real-time operating data, risk deviation and process condition characteristics are input into the pre-built interlocking control model, and a two-level synchronous interlocking control strategy and an abnormal condition graded protection strategy are output and executed.

[0006] Furthermore, the variable frequency operation signal of the single-stage screw compressor is connected to the automatic control SIS system through a dual-channel redundancy method, and the real-time operating current signal of the two-stage screw compressor is connected to the central control DCS system through a redundant transmitter circuit. Real-time process operating parameters of the two-stage screw compressor are simultaneously acquired, including: An intermediate relay with contact health online monitoring is added to the inverter operation output terminal of the first-stage screw compressor. The two independent auxiliary contacts of the intermediate relay are respectively connected to the dual-channel redundant digital input module of the automatic control SIS system to complete the dual-channel redundant acquisition of the inverter operation signal. In the secondary circuit of the three-phase current transformer of the main motor of the two-stage screw compressor, an integrated current transmitter group with self-diagnostic function is configured. Each current transmitter group contains two independent transmission units, which connect the two independent standard signals to the analog input redundant channel of the central control DCS system to complete the redundant acquisition of real-time operating current signals.

[0007] Furthermore, the real-time process operating parameters include primary exhaust pressure, secondary intake pressure, rotor temperature, motor winding temperature, intake valve opening, and lubricating oil pressure.

[0008] Furthermore, the system predicts process disturbance events within a preset time window and extracts the interlocking risk disturbance features corresponding to these events, including: Collect multi-dimensional source data that affects the interlock risk of two-stage screw compressors. The multi-dimensional source data includes historical process interlock records, equipment failure data, real-time production process parameters, upstream and downstream equipment operating status data, and production plan scheduling data. The multi-dimensional source data is preprocessed to complete missing value imputation, outlier removal, time-series alignment and normalization. Using the preprocessed multi-dimensional source data as the training set, a time-series machine learning model is constructed and trained to obtain a process disturbance prediction model; The real-time collected process operation data and production scheduling data are input into the process disturbance prediction model, which outputs the process disturbance events and corresponding interlocking risk disturbance characteristics within a preset time window.

[0009] Furthermore, the preset interlocking safety threshold system includes three interlocking control levels that increase sequentially: early warning level, control level, and emergency stop level. Each interlocking control level corresponds to an independent risk probability threshold, control action logic, and timing constraints.

[0010] Furthermore, the timing constraints include at least the maximum permissible lag time for the interlocking shutdown of the primary screw compressor and the secondary screw compressor, and the response time window for abnormal operating condition protection actions.

[0011] Furthermore, the interlocking control model is based on a deep reinforcement learning model, with the safety, response speed and operation stability of the interlocking control as the reward function, and the equipment safety operation boundary and production process requirements of the two-stage screw compressor as the constraint conditions. The two-stage synchronous interlocking control strategy includes the shutdown command timing of the two-stage compressors and the selection of emergency cut-off trigger conditions for the shutdown circuit; The abnormal operating condition graded protection strategy includes graded execution logic corresponding to the interlock control level, such as early warning issuance, load adjustment, intake valve pre-control, and interlock shutdown.

[0012] Furthermore, the screw compressor interlocking risk prediction model is a composite architecture of a temporal convolutional network with a two-level equipment coupling attention mechanism combined with a gated recurrent unit. The model output layer has two parallel branches. The first branch outputs the interlocking failure probability, and the second branch outputs the equipment damage risk level. The model is trained using the historical time-series operation data of the two-stage screw compressor throughout its entire life cycle as the training set, and a multi-task joint loss function is used for training.

[0013] Furthermore, the risk deviation is calculated using the formula: D=α×(P-P0) / P0+β×(L-L0) / L0, where D is the risk deviation, P is the real-time interlock failure probability, P0 is the interlock failure probability threshold corresponding to the warning level, L is the real-time equipment damage risk level, L0 is the equipment damage risk level threshold corresponding to the warning level, α is the risk weight of the interlock failure probability, β is the risk weight of the equipment damage risk level, and α>β. The weight values ​​are set according to the degree of influence of the parameters on interlock safety.

[0014] Furthermore, the interlocking risk disturbance characteristics include the pressure change gradient, load change magnitude, degree of parameter deviation from rated operating range, duration of disturbance event, upstream and downstream operating condition matching degree, and interlocking risk correlation degree corresponding to the disturbance event.

[0015] The technical solution of this invention achieves the following technical effects: By using a dual-channel redundant approach to access the variable frequency operation signal of the primary screw compressor and a redundant transmission circuit to access the real-time operating current signal of the compressor, real-time process operating parameters are simultaneously acquired. These three elements work together to construct a comprehensive sensing link covering the electrical operating status, load status, and process operating status of the equipment. The dual-channel redundant architecture avoids signal loss and interlocking logic input distortion caused by single-acquisition circuit failures at the hardware level. Comprehensive data acquisition provides complete and reliable status information for interlocking control, eliminating the risk of interlocking maloperation or failure to operate due to signal failure or incomplete status perception. Furthermore, the process disturbance prediction mechanism identifies potential risks within a preset time window in advance, extracting interlocking risks. By analyzing the characteristics of potential disturbances and combining them with multi-source real-time operational data, an interlocking risk prediction model is used to quantify and output the interlocking risk prediction results for future periods. This enables early prediction and quantitative assessment of interlocking risks, allowing the invention to identify and prevent interlocking risks at their nascent stage and plan the synchronous interlocking control strategy for the two-stage compressor in advance. Risk quantification and classification are achieved through a preset interlocking safety threshold system. Based on the risk deviation, the interlocking control level and timing constraints are dynamically matched. Combined with inherent equipment parameters, real-time operational data, and process characteristics, the interlocking control model outputs an adaptive two-stage synchronous interlocking control strategy and an abnormal condition classification protection strategy, achieving dynamic adaptation of the interlocking control logic to production load, process fluctuations, and equipment operating status.

[0016] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating a screw compressor interlocking control method according to the present invention. Detailed Implementation

[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0021] like Figure 1 As shown, the present invention discloses a screw compressor interlocking control method, wherein the screw compressor is a two-stage screw compressor, comprising a first-stage screw compressor and a second-stage screw compressor connected in sequence, and is equipped with an automatic control SIS system for performing safety interlocking control, and a central control DCS system for monitoring the operating status. The method specifically includes the following steps: Step S1: Connect the variable frequency operation signal of the first-stage screw compressor to the automatic control SIS system through a dual-channel redundancy method, and connect the real-time operating current signal of the two-stage screw compressor to the central control DCS system through a redundant transmission circuit, and synchronously collect the real-time process operation parameters of the two-stage screw compressor. Step S2: Based on the historical operating data and real-time process operating status of the two-stage screw compressor, predict process disturbance events within a preset time window in the future, and extract the interlocking risk disturbance features corresponding to the process disturbance events; Step S3: Input the variable frequency operation signal, real-time operating current signal, real-time process operation parameters, and interlock risk disturbance characteristics of the first-stage screw compressor into the pre-constructed screw compressor interlock risk prediction model to obtain the interlock risk prediction results of the two-stage compressor within a future preset time window; the screw compressor interlock risk prediction model is based on a time-series deep learning model and is constructed using the historical time-series operation data of the two-stage screw compressor throughout its entire life cycle as the training set; the interlock risk prediction results include at least the interlock failure probability and the equipment damage risk level; Step S4: Based on the preset interlocking safety threshold system, determine whether the interlocking risk prediction result exceeds the limit. If it does not exceed the limit, maintain the original interlocking strategy. If it exceeds the limit, calculate the risk deviation degree and determine the matching interlocking control level and timing constraints. Step S5: Input the inherent parameters of the compressor equipment, real-time operating data, risk deviation and process condition characteristics into the pre-built interlocking control model, output a two-level synchronous interlocking control strategy and an abnormal condition graded protection strategy, and execute them.

[0022] In this embodiment, a dual-channel redundant approach is used to access the inverter operation signal of the primary screw compressor, and a redundant transmitter circuit is used to access the real-time operating current signal of the compressor. Real-time process operating parameters are simultaneously acquired, and these three components work together to construct a comprehensive sensing link covering the electrical operating status, load status, and process operating status of the equipment. The dual-channel redundant architecture avoids signal loss and interlocking logic input distortion caused by single-acquisition circuit failures at the hardware level, eliminating the risk of interlocking maloperation or failure to operate due to signal failure or incomplete status perception. A process disturbance prediction mechanism identifies potential risk factors within a preset time window in advance, extracts interlocking risk disturbance characteristics, and then combines these with multi-source real-time operating data to... The interlock risk prediction model quantifies and outputs the interlock risk prediction results for future periods, enabling early prediction and quantitative assessment of interlock risks. This allows the invention to identify and prevent interlock risks at their nascent stage, and to plan the synchronous interlock control strategy for the two-stage compressor in advance. By setting up an interlock safety threshold system, the invention completes the risk quantification and classification, dynamically matches the interlock control level and timing constraints based on the risk deviation, and combines the inherent parameters of the equipment, real-time operating data, and process characteristics. Through the interlock control model, it outputs an adaptive two-stage synchronous interlock control strategy and an abnormal condition classification protection strategy, realizing the dynamic adaptation of the interlock control logic to production load, process fluctuations, and equipment operating status.

[0023] In some embodiments of the present invention, in existing interlocking control schemes, the operating status signal of the first-stage screw compressor is accessed to the automatic control SIS system in a single-channel or redundant manner without state monitoring. This results in a single-point fault risk in the signal acquisition circuit, and latent contact faults cannot be identified online. Consequently, the interlocking logic input signal of the automatic control SIS system is misaligned with the actual operating status of the equipment, leading to interlocking failure or maloperation events. The operating current signal of the two-stage screw compressor is accessed to the central control DCS system in a single-channel transmission manner. The transmission unit has no self-diagnostic capability for faults and insufficient signal redundancy, causing the central control DCS system to be unable to obtain the actual operating load status of the equipment and thus unable to provide reliable auxiliary criteria for interlocking control. At the same time, there is no unified timing synchronization mechanism between the electrical operating signals and process operating parameters. The timing deviation of multi-source data is too large, resulting in timing misalignment of the input data for the subsequent interlocking risk prediction model, and the prediction results cannot match the actual operating status of the equipment. To address the above deficiencies, this embodiment constructs an integrated signal acquisition method that integrates dual-channel redundant inverter operating signal acquisition and access, redundant self-diagnostic current signal acquisition and access, and multi-source parameter synchronous acquisition, providing a reliable, synchronous, and complete basic data source for interlocking control.

[0024] Specifically, an intermediate relay with online contact health monitoring is added to the inverter output terminal of the primary screw compressor. The coil circuit of the intermediate relay is connected in series to the inverter's operating status output dry contact, ensuring that the excitation state of the intermediate relay coil completely corresponds to the actual operating state of the inverter. The intermediate relay is equipped with two independent, electrically isolated normally open auxiliary contacts, and the switching capacity of the two auxiliary contacts matches the input characteristics of the digital input module of the automatic control SIS system. The intermediate relay has a built-in online contact health monitoring unit, which collects the circuit electrical characteristic parameters of the two auxiliary contacts in real time. The monitoring unit is connected to the industrial system. The field bus uploads the monitoring data to the logic operation unit of the automatic control SIS system in real time, realizing online real-time monitoring of the health status of the contacts; the two independent auxiliary contacts of the intermediate relay are connected to the dual-channel redundant digital input module of the automatic control SIS system through independent shielded control cables. The two auxiliary contacts correspond to the A channel and B channel of the redundant digital input module, respectively. The A channel and B channel realize hardware redundancy and logic voting within the automatic control SIS system. The automatic control SIS system executes a two-out-of-one shutdown interlock trigger logic on the input signals of the two channels, and at the same time executes a two-out-of-two normal operation status confirmation logic. This configuration eliminates the impact of single-channel contact or channel failure on interlocking logic. When a single channel fails, the other channel can still trigger the shutdown interlock normally, avoiding interlock failure. Only when both channels simultaneously confirm the operating status can the equipment be determined to be in normal operating condition, avoiding interlock malfunction caused by a single channel's false signal. The online contact health monitoring unit can identify the trend of contact performance degradation before the contact completely fails, triggering maintenance warnings in advance and eliminating potential fault risks in signal acquisition.

[0025] More specifically, in the secondary circuits of the three-phase current transformers of the primary and secondary main motors of the two-stage screw compressor, integrated current transmitter groups with self-diagnostic functions are configured respectively. For each phase of the current transformer secondary circuit of each main motor, a corresponding current transmitter group is configured. Each current transmitter group contains two independent, electrically isolated transmitter units. The input sides of the two transmitter units are connected in parallel to the secondary circuit of the corresponding current transformer, with the input range matching the rated secondary parameters of the corresponding current transformer. The output sides of both transmitter units output standard DC signals, with the output range matching the rated operating parameters of the corresponding main motor. Each transmitter unit has a built-in self-diagnostic module, which monitors the transmitter unit in real time. The self-diagnostic module monitors the operating status of the input circuit, output circuit, power supply, and internal processing unit. It feeds back the fault status to the central control DCS system in real time via standard fault signals, enabling online self-diagnosis and reporting of transmitter unit faults. The standard signals output from the two independent transmitter units of each current transmitter group are connected to the analog input redundancy channel of the central control DCS system via independent shielded signal cables. The two standard signals correspond to channels A and B of the analog input redundancy channel, respectively. The central control DCS system performs real-time comparison of the input signals from the two channels. When the deviation between the two signals exceeds a preset range, a signal anomaly warning is triggered, and the effective value from the two signals is selected for subsequent logical operations. This configuration enables hardware redundancy acquisition of current signals. When a single transmitter unit fails, the other can still provide a valid signal, avoiding signal acquisition interruption. The self-diagnostic module can identify typical faults of the transmitter unit in real time, and the central control DCS system can directly identify fault signals and remove invalid data, avoiding misjudgment of equipment status caused by fault signals. The two-signal deviation comparison and effective value selection mechanism avoids signal distortion caused by transmitter unit parameter drift and prevents misjudgment of equipment load status.

[0026] The system synchronously acquires real-time process operating parameters of the two-stage screw compressor, including primary exhaust pressure, secondary intake pressure, rotor temperature, motor winding temperature, intake valve opening, and lubricating oil pressure. For each process parameter, corresponding redundant measuring instruments are configured. All measuring instruments for process parameters are connected to the corresponding redundant input channels of the central control DCS system via hard-wiring. The central control DCS system is configured with a unified timing synchronization trigger unit for the acquisition of all electrical signals and process operating parameters. This timing synchronization trigger unit performs synchronous sampling of all signals at a fixed acquisition cycle, controlling the timing deviation of all acquired signals within a preset range. This configuration achieves redundant acquisition of all process parameters, ensuring continuous parameter acquisition even if a single instrument fails.

[0027] In some embodiments of the present invention, existing interlocking control schemes for two-stage screw compressors adopt a reactive control mode with no pre-process disturbance prediction step, triggering interlocking actions solely based on the real-time operating status of the equipment. A few schemes with disturbance prediction capabilities use only the process operating data of a single piece of equipment as the basis for prediction, failing to incorporate multi-dimensional source data such as upstream and downstream operating conditions and production scheduling that affect equipment interlocking risks. Therefore, the prediction results cannot cover the interlocking risk triggers throughout the entire production process. This embodiment constructs a comprehensive method encompassing multi-dimensional source data acquisition, interlocking risk adaptability preprocessing, risk correlation model training, real-time disturbance prediction, and risk feature extraction. This provides pre-judgment inputs strongly correlated with interlocking risks for proactive interlocking risk prevention and control. The specific implementation is as follows: Step S21: Collect multi-dimensional source data affecting the interlocking risk of the two-stage screw compressor. The source data includes the following five categories: The first category is historical process interlocking records, which come from the historical interlocking event logs of the automatic control SIS system, including the interlocking trigger time, triggering reason, action execution result, and equipment operating parameters before and after the interlocking; the second category is equipment fault data, which comes from the equipment lifecycle management system, including historical equipment fault types, fault occurrence time, equipment operating parameters before and after the fault, and fault repair records; the third category is real-time production process parameters, which come from the synchronous acquisition of the central control DCS system through the aforementioned integrated signal acquisition method. The data includes: 1) Inverter operation signals of the first-stage screw compressor, real-time operating current of the two-stage screw compressor, first-stage discharge pressure, second-stage intake pressure, rotor temperature, motor winding temperature, intake valve opening, and lubricating oil pressure; 2) Operating status data of upstream and downstream equipment, sourced from the central control DCS system, including buffer tank pressure at the intake front end of the two-stage screw compressor, operating status of the intake purification equipment, process pipeline pressure at the exhaust back end, operating status of downstream gas-using equipment, and valve opening signals; 3) Production planning and scheduling data, sourced from the production execution system, including production load adjustment plans, equipment start-up and shutdown scheduling plans, and process parameter adjustment instructions. All source data carries a unique timestamp, and the timestamp accuracy is consistent with the acquisition cycle of the timing synchronization trigger unit in the aforementioned integrated signal acquisition method.

[0028] Step S22: Preprocess the collected multi-dimensional source data, including missing value imputation, outlier removal, time series alignment, and normalization. For discrete data, including historical process interlock records, equipment fault data, and equipment operating status switch signals, imputation is performed using historical data of the same type under the same operating conditions. For continuous time series data, including process operating parameters, current signals, and pressure and temperature analog data, linear interpolation combined with equipment safety operating boundaries is used for imputation, ensuring that the imputed data does not exceed the equipment's rated operating range for the corresponding parameters. Outlier identification is performed using the 3σ criterion combined with equipment safety operating boundaries. For data exceeding the 3σ interval and the equipment's rated operating range, outliers are identified. According to the data, outliers are identified and removed; data exceeding the 3σ range but within the rated operating range of the equipment are marked as operating condition fluctuation data and retained; based on the synchronous sampling time of the timing synchronization trigger unit in the aforementioned integrated signal acquisition method, timing resampling is performed on all multi-dimensional source data to ensure that the sampling time of all data is completely aligned with the reference sampling time, and the timing deviation of all data after alignment does not exceed the reference sampling period; for continuous data, the minimum-maximum normalization method is used to map the data to the [0,1] interval, with the mapping reference being the minimum and maximum rated operating values ​​of the corresponding parameters of the equipment; for discrete data, one-hot encoding is used for normalization processing.

[0029] Step S23: Using the preprocessed multi-dimensional source data as the training set, construct and train a time-series machine learning model to obtain a process disturbance prediction model. This is specifically divided into the following sub-steps: Step S231: Based on the preprocessed multi-dimensional source data, divide the training set, validation set and test set according to a preset ratio; label the historical data in the training set, including the occurrence time, type and magnitude of the process disturbance event, as well as the interlocking risk correlation degree corresponding to the disturbance event. The interlocking risk correlation degree is assigned a value based on the probability of the disturbance event triggering interlocking action or equipment failure in history, with a value range of [0,1]. Step S232: A time-series machine learning model is adopted, specifically a model architecture that combines a bidirectional long short-term memory network with an attention mechanism. The model input layer is preprocessed multi-dimensional source data, the hidden layer is a bidirectional long short-term memory network layer, the attention mechanism layer is used to assign higher weights to feature dimensions that are highly correlated with the interlocking risk, and the output layer is the prediction result of process disturbance events within a future preset time window. Step S233: Using the labeled training set as input and the actual occurrence of process disturbance events as labels, the model is trained using the cross-entropy loss function combined with the weighted coefficient of interlocking risk correlation. During training, the model accuracy is verified using the validation set. When the prediction accuracy of the model on the validation set reaches a preset threshold, training is stopped, and a fixed process disturbance prediction model is obtained. The preset threshold is set according to the safety requirements of the production process. The training weight of high interlocking risk disturbances is strengthened by the weighted coefficient of interlocking risk correlation, thereby improving the model's ability to predict high-risk disturbance events.

[0030] Step S24: After performing the same preprocessing operation as in step S22, the process operation data, upstream and downstream equipment operation status data, and production scheduling data issued in real time by the central control DCS system and the production execution system are input into the fixed process disturbance prediction model. The model outputs process disturbance events within a future preset time window. The output includes the occurrence time of the disturbance event, the disturbance type, the disturbance magnitude, and the corresponding interlocking risk correlation. The future preset time window is set according to the production process interlocking response time requirements.

[0031] Step S25: For the process disturbance events output by the model, extract disturbance features that are strongly correlated with the interlocking risk of the two-stage screw compressor. The extracted features include: the pressure change gradient, load change amplitude, degree of parameter deviation from the rated operating range, duration of the disturbance event, matching degree of upstream and downstream operating conditions, and correlation degree of interlocking risk. After the extracted interlocking risk disturbance features are time-series bound with the real-time collected equipment operating data, they are transmitted to the subsequent interlocking risk prediction stage.

[0032] In this embodiment, by collecting multi-dimensional source data covering all causes of interlocking risks, standardizing preprocessing adapted to interlocking control scenarios, and designing a time-series prediction model strongly correlated with interlocking risks, the shortcomings of existing technologies, such as single prediction basis, incomplete risk coverage, and insufficient accuracy in predicting high-risk disturbances, are addressed. This enables proactive prediction of process disturbances, reserving a response time window for interlocking risk prevention and control. The output interlocking risk disturbance characteristics provide accurate input for the risk prediction process, driving the interlocking control of two-stage screw compressors to shift from reactive post-event response to proactive pre-event prevention and control.

[0033] As one embodiment, in step S3, the screw compressor interlocking risk prediction model is constructed based on a time-series deep learning model and using historical time-series operating data of the two-stage screw compressor throughout its entire lifecycle as the training set. The specific construction method of the screw compressor interlocking risk prediction model is as follows: Step S31: Integrate the frequency converter operation signal of the first-stage screw compressor, the real-time operating current signal of the two-stage screw compressor, the real-time process operation parameters, and the interlocking risk disturbance characteristics obtained in the previous steps into an interlocking risk prediction input feature set; wherein, the frequency converter operation signal is the effective status signal after dual-channel redundant voting of the automatic control SIS system, the real-time operating current signal is the effective value after comparison of redundant channels of the central control DCS system, the real-time process operation parameters are the effective values ​​of redundant measurements acquired synchronously, and the interlocking risk disturbance characteristics are the strongly correlated risk characteristics after time-series binding; all input features carry timestamps consistent with the front-end acquisition stage to maintain time synchronization.

[0034] Step S32: Collect historical time-series operation data of the two-stage screw compressor throughout its entire life cycle to construct a model training dataset. The historical time-series operation data of the entire life cycle includes time-series data of all operating conditions during the equipment commissioning stage, normal operation stage, operating condition fluctuation stage, fault pre-warning stage, interlocking action stage, and shutdown maintenance stage. The data dimensions are completely consistent with the input feature set constructed in step S31. Perform the same preprocessing operation as in step S22 on the training dataset, and label the dataset. The labeling content includes the actual occurrence of interlocking failure and the actual occurrence of equipment damage under the corresponding time-series data. The labeling labels include interlocking failure event labels and equipment damage risk level labels. Divide the labeled dataset into training set, validation set, and test set according to a preset ratio.

[0035] Step S33: Based on a temporal deep learning model, specifically a composite architecture of a temporal convolutional network with a two-level device coupling attention mechanism and a gated recurrent unit, a screw compressor interlocking risk prediction model is constructed. The model input layer is the input feature set constructed in step S31. The first hidden layer is a temporal convolutional network layer used to extract local temporal correlation features of multi-source features. The second hidden layer is a gated recurrent unit layer used to capture long-term risk evolution trends. The two-level device coupling attention mechanism layer is embedded after the gated recurrent unit layer and is used to assign differentiated weights to the coupled operation features of the first-stage screw compressor and the second-stage screw compressor, strengthening the two-stage device coupling. The model is designed to identify risk characteristics related to the operation of the screw compressor. The output layer has two parallel branches: the first branch outputs the probability of interlock failure within a future preset time window, and the second branch outputs the equipment damage risk level for the same period. Using the training set defined in step S32 as input, and the labeled interlock failure event tags and equipment damage risk level tags as supervision tags, a multi-task joint loss function is employed for model training. During training, the model's prediction accuracy is verified using a validation set. When the model's prediction accuracy on the validation set reaches a preset threshold, training is stopped, resulting in a fixed screw compressor interlock risk prediction model. The preset threshold is set according to chemical production safety regulations and equipment interlock protection requirements.

[0036] Step S34: Input the real-time acquired and time-synchronized first-stage screw compressor frequency conversion operation signal, real-time operating current signal, real-time process operation parameters, and interlock risk disturbance characteristics extracted in real-time by the central control DCS system and the automatic control SIS system into the solidified screw compressor interlock risk prediction model. The model outputs the interlock risk prediction results within a future preset time window. The future preset time window is consistent with the aforementioned process disturbance prediction time window. The interlock risk prediction results include at least the interlock failure probability and the equipment damage risk level. The interlock failure probability ranges from [0,1], and the equipment damage risk level is divided into at least three levels according to the degree of risk, which correspond one-to-one with the subsequent interlock control levels.

[0037] Step S35: Verify the validity of the interlocking risk prediction results output by the model. The verification rules include: verification of the temporal continuity of the prediction results, verification of the correlation between the input features and the prediction results, and verification of the conformity with the safety operation boundary of the equipment. The prediction results that pass the verification are synchronously transmitted to the interlocking logic operation unit of the automatic control SIS system and the monitoring unit of the central control DCS system. The prediction results that fail the verification are discarded, and the model self-verification warning is triggered at the same time.

[0038] In this embodiment, by integrating multi-source strongly correlated features that have undergone redundancy verification and time-series synchronization, the validity and risk correlation of the input data are ensured; by constructing a full lifecycle, full-condition dataset, the model is adapted to the risk evolution patterns of the equipment throughout all operating stages; by adapting the model architecture and dual-branch output design to the coupling characteristics of two-level equipment, the core requirements of two-level compressor interlocking control are matched, and the output results can directly support subsequent hierarchical interlocking control; by verifying the validity of prediction results through multiple rules, the reliability of the output results is ensured, and pre-quantitative prediction of interlocking risks is achieved.

[0039] In a specific implementation, as one example, the interlocking risk prediction result output from the aforementioned steps is executed. The interlocking risk prediction result includes at least the interlocking failure probability and the equipment damage risk level. The equipment damage risk level is divided into three levels corresponding to the interlocking control level. The specific implementation is as follows: Step S41: Based on chemical production safety regulations, interlocking protection requirements for two-stage screw compressor equipment, and continuous operation requirements of the production process, construct a preset interlocking safety threshold system comprising three progressively increasing interlocking control levels: early warning level, control level, and emergency shutdown level. Each interlocking control level corresponds to an independent risk probability threshold, control action logic, and timing constraints. The risk probability threshold includes an interlocking failure probability threshold and an equipment damage risk level threshold. The interlocking failure probability thresholds for the early warning level, control level, and emergency shutdown level increase sequentially, as do the equipment damage risk levels. The timing constraints include the maximum permissible lag time for interlocking shutdown of the primary and secondary screw compressors and the response time window for abnormal operating condition protection actions. The maximum permissible lag time for interlocking shutdown corresponding to the three interlocking control levels decreases sequentially, and the response time window for abnormal operating condition protection actions narrows sequentially.

[0040] Step S42: Compare the interlock failure probability and equipment damage risk level output in the previous steps with the risk probability thresholds of each level in the preset interlock safety threshold system to determine whether the interlock risk prediction result exceeds the limit. If the interlock failure probability is lower than the early warning level interlock failure probability threshold and the equipment damage risk level is lower than the early warning level equipment damage risk level threshold, it is determined that it has not exceeded the limit. If the interlock failure probability reaches or exceeds the early warning level interlock failure probability threshold, or the equipment damage risk level reaches or exceeds the early warning level equipment damage risk level threshold, it is determined that it has exceeded the limit.

[0041] Step S43: If it is determined that the limit has not been exceeded, the original interlocking strategy is maintained. The original interlocking strategy is the fixed interlocking protection logic under the normal operating conditions of the two-stage screw compressor. That is, the automatic control SIS system executes the synchronous shutdown interlocking action of the two-stage compressor only when the operating status signal of the first-stage screw compressor triggers the shutdown condition; at the same time, it continuously receives the interlocking risk prediction results in real time and performs threshold limit exceedance judgment in a loop.

[0042] Step S44: If the boundary is determined to be exceeded, calculate the risk deviation based on the interlock failure probability after the boundary exceedance, the equipment damage risk level, and the risk probability threshold corresponding to the warning level. The risk deviation calculation formula is: D=α×(P-P0) / P0+β×(L-L0) / L0, where D is the risk deviation, P is the real-time interlock failure probability, P0 is the interlock failure probability threshold corresponding to the warning level, L is the real-time equipment damage risk level, L0 is the equipment damage risk level threshold corresponding to the warning level, α is the risk weight of the interlock failure probability, and β is the risk weight of the equipment damage risk level, α>β. The weight value is set according to the degree of influence of the parameter on the interlock safety. The interlock failure probability directly determines the reliability of the interlock action and has a higher impact on equipment safety, so α is greater than β. The calculated risk deviation corresponds one-to-one with the interlock control level. The higher the risk deviation, the higher the matched interlock control level.

[0043] Step S45: Based on the calculated risk deviation, match the corresponding interlock control level and synchronously retrieve the control action logic and timing constraints corresponding to the interlock control level. Among them, the early warning level corresponds to the non-stop early warning and pre-control action logic, and the timing constraints correspond to the longest maximum allowable delay time of interlock shutdown and the widest response time window; the control level corresponds to the load adjustment and pre-stop preparation action logic, and the timing constraints correspond to the maximum allowable delay time of interlock shutdown and the response time window in the middle interval; the emergency stop level corresponds to the synchronous interlock shutdown action logic of two-stage compressors, and the timing constraints correspond to the shortest maximum allowable delay time of interlock shutdown and the narrowest response time window. Moreover, the maximum allowable delay time of interlock shutdown corresponding to the emergency stop level meets the mandatory requirement of secondary synchronous shutdown after primary shutdown as specified in the production process.

[0044] This embodiment constructs a three-level interlocking safety threshold system that conforms to safety standards and equipment and process requirements, providing a clear and compliant judgment benchmark for interlocking risk assessment. By combining the out-of-bounds judgment of the pre-quantified risk prediction results, it realizes the linkage between interlocking control and risk evolution trends, replacing the traditional ex-post triggering mode with fixed thresholds. By calculating the risk deviation degree and matching the corresponding interlocking control level, it achieves differentiated control under different risk conditions, avoiding unnecessary production shutdowns and inadequate control actions under high-risk conditions. Through differentiated timing constraints bound to the control level, it adapts to the two-level compressor shutdown control requirements under different risk conditions.

[0045] In some embodiments of the present invention, for step S5, the risk deviation, matching interlocking control level, and timing constraints output from the aforementioned steps are executed, the input data source covers the timing synchronization equipment operation data obtained above, and the output is a two-level synchronous interlocking control strategy and an abnormal operating condition graded protection strategy; specifically, the method for constructing the interlocking control model includes the following steps: Step S51: Integrate the inherent parameters of the compressor equipment, real-time operating data, risk deviation, and process condition characteristics to construct the input parameter set for the interlocking control model. The inherent parameters of the compressor equipment include the rated power, rated pressure range, rotor rated speed, safe operating boundary parameters, and SIS system interlocking shutdown circuit configuration parameters. Real-time operating data includes the frequency conversion operating signal of the first-stage screw compressor after dual-channel redundant voting, the real-time operating current of the two-stage screw compressor after redundant channel comparison, real-time process operating parameters acquired synchronously, and operating status data of upstream and downstream equipment. The risk deviation is the real-time quantified value calculated in the preceding steps. Process condition characteristics include real-time production load, process pipeline pressure status, and production plan scheduling requirements. All input parameters undergo time-synchronized processing, and the timestamp accuracy is consistent with the acquisition cycle of the aforementioned integrated signal acquisition method.

[0046] Step S52: Based on the deep reinforcement learning model, a deep deterministic policy gradient model is used to construct an interlocking control model. The model sets up a state space, action space, reward function, and constraints. The dimension of the state space completely matches the dimension of the input parameter set constructed in step S51. The action space covers the adjustable parameters of the two-level synchronous interlocking control strategy and the executable actions of the abnormal condition graded protection strategy. The reward function takes the safety, response speed, and operational stability of interlocking control as joint optimization objectives, with safety having the highest weight. Positive rewards are given for no equipment safety boundary exceedances or no interlocking failures after the interlocking action is executed, and negative penalties are given for equipment safety boundary exceedance events. Response speed is measured by the response time window of the control action within the corresponding interlocking control level. Completion within the specified timeframe is rewarded positively, while actions lagging behind time constraints are penalized negatively. Operational stability is rewarded with minimizing unnecessary shutdowns and production condition fluctuations, while excessive shutdowns and significant fluctuations in operating conditions are penalized negatively. Constraints are based on the safe operating boundaries of the two-stage screw compressor and mandatory production process requirements, and all control strategies output by the model must not exceed these constraints. The model is trained using historical operating data of the two-stage screw compressor under all operating conditions, historical data of interlocking actions, and process operating conditions. Offline pre-training and online reinforcement training are performed on the model. Training stops when the control strategies output by the model meet the constraints and the cumulative value of the reward function reaches the preset threshold set according to the safety regulations for chemical production, resulting in a fixed interlocking control model.

[0047] Step S53: Input the real-time input parameter set constructed in step S51 into the solidified interlocking control model. The model outputs a two-level synchronous interlocking control strategy and an abnormal condition graded protection strategy that are completely matched with the current interlocking control level. The two-level synchronous interlocking control strategy includes the shutdown command sequence of the two-level compressor, the shutdown loop selection, and the emergency cut-off trigger condition: the shutdown command sequence strictly matches the maximum allowable lag time requirement of the interlocking shutdown for the corresponding interlocking control level; the shutdown loop selection is dynamically determined based on the real-time health status of the dual-channel redundant input loop of the automatic control SIS system; the emergency cut-off trigger condition matches the real-time risk deviation quantification value; the abnormal condition graded protection strategy includes graded execution logic corresponding to the interlocking control level, such as early warning release, load adjustment, intake valve pre-control, and interlocking shutdown. Specifically, the early warning level corresponds to early warning release and intake valve pre-control action, the control level corresponds to load adjustment and pre-shutdown preparation action, and the emergency shutdown level corresponds to the two-level synchronous interlocking shutdown action. The graded execution logic corresponds one-to-one with the grade division of the aforementioned preset interlocking safety threshold system.

[0048] Step S54: Perform safety compliance verification on the two-level synchronous interlocking control strategy and abnormal operating condition graded protection strategy output by the model. The verification includes: whether it meets the equipment safety operation boundary constraints, whether it meets the mandatory requirements of the production process, whether it matches the timing constraints of the corresponding interlocking control level, and whether it meets the current chemical production safety regulations. Control strategies that pass the verification are distributed to the automatic control SIS system and the central control DCS system according to the executing entity. Control strategies that fail the verification are removed, and the preset backup safety strategy of the corresponding interlocking control level is retrieved and distributed for execution.

[0049] Step S55: The automatic control SIS system executes the shutdown interlock and emergency cut-off related safety actions in the two-level synchronous interlock control strategy, strictly following the preset shutdown command sequence and triggering conditions; the central control DCS system executes the early warning release, load adjustment, and intake valve pre-control related pre-control actions in the abnormal operating condition graded protection strategy, and synchronously monitors the equipment operating status and process condition changes after the actions are executed; the action execution results and the real-time response data of the equipment and process are fed back to the interlock control model for online iterative optimization of the model.

[0050] In this embodiment, an interlocking control model based on deep reinforcement learning is designed to achieve multi-objective collaborative optimization of interlocking control safety, response speed, and operational stability under the hard constraints of equipment and process safety. Dynamic hierarchical control strategies matched to real-time risk levels are generated to adapt to control requirements under different risk conditions, avoiding the problems of delayed actions in high-risk conditions and excessive downtime in low-risk conditions. Safety compliance pre-verification and backup strategy settings ensure the compliance of control actions and the continuity of interlocking control. Continuous model optimization is achieved through feedback iteration of execution effects, adapting to changes in the operational characteristics of the equipment throughout its entire lifecycle.

[0051] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely exemplary illustrations of the application as defined herein, and are to be considered as covering any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Thus, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A screw compressor interlocking control method, wherein the screw compressor is a two-stage screw compressor, comprising a primary screw compressor and a secondary screw compressor connected in sequence, and is equipped with an automatic control SIS system for performing safety interlocking control, and a central control DCS system for monitoring operating status, characterized in that, The method includes: The variable frequency operation signal of the single-stage screw compressor is connected to the automatic control SIS system through a dual-channel redundancy method, and the real-time operating current signal of the two-stage screw compressor is connected to the central control DCS system through a redundant transmission circuit, so as to synchronously collect the real-time process operation parameters of the two-stage screw compressor. Based on the historical operating data and real-time process operating status of the two-stage screw compressor, process disturbance events within a preset time window are predicted, and the interlocking risk disturbance characteristics corresponding to the process disturbance events are extracted. The variable frequency operation signal, real-time operating current signal, real-time process operation parameters, and interlock risk disturbance characteristics of the first-stage screw compressor are input into the pre-constructed screw compressor interlock risk prediction model to obtain the interlock risk prediction results of the two-stage compressor within a future preset time window. Based on the preset interlocking safety threshold system, it is determined whether the interlocking risk prediction result exceeds the limit. If it does not exceed the limit, the original interlocking strategy is maintained. If it exceeds the limit, the risk deviation degree is calculated, and the matching interlocking control level and timing constraints are determined. The inherent parameters of the compressor equipment, real-time operating data, risk deviation and process condition characteristics are input into the pre-built interlocking control model, and a two-level synchronous interlocking control strategy and an abnormal condition graded protection strategy are output and executed.

2. A screw machine interlock control method according to claim 1, wherein The variable frequency drive (VFD) operation signal of the single-stage screw compressor is connected to the automatic control system (SIS) via dual-channel redundancy. The real-time operating current signal of the two-stage screw compressor is connected to the central control system (DCS) via redundant transmitter circuits. Real-time process operating parameters of the two-stage screw compressor are simultaneously acquired, including: An intermediate relay with contact health online monitoring is added to the inverter operation output terminal of the first-stage screw compressor. The two independent auxiliary contacts of the intermediate relay are respectively connected to the dual-channel redundant digital input module of the automatic control SIS system to complete the dual-channel redundant acquisition of the inverter operation signal. In the secondary circuit of the three-phase current transformer of the main motor of the two-stage screw compressor, an integrated current transmitter group with self-diagnostic function is configured. Each current transmitter group contains two independent transmission units, which connect the two independent standard signals to the analog input redundant channel of the central control DCS system to complete the redundant acquisition of real-time operating current signals.

3. A screw machine interlock control method according to claim 2, wherein The real-time process operating parameters include primary exhaust pressure, secondary intake pressure, rotor temperature, motor winding temperature, intake valve opening, and lubricating oil pressure.

4. A method of interlock control of a screw machine according to claim 1, wherein Predict process disturbance events within a preset time window and extract the interlocking risk disturbance features corresponding to the process disturbance events, including: Collect multi-dimensional source data that affects the interlock risk of two-stage screw compressors. The multi-dimensional source data includes historical process interlock records, equipment failure data, real-time production process parameters, upstream and downstream equipment operating status data, and production plan scheduling data. The multi-dimensional source data is preprocessed to complete missing value imputation, outlier removal, time-series alignment and normalization. Using the preprocessed multi-dimensional source data as the training set, a time-series machine learning model is constructed and trained to obtain a process disturbance prediction model; The real-time collected process operation data and production scheduling data are input into the process disturbance prediction model, which outputs the process disturbance events and corresponding interlocking risk disturbance characteristics within a preset time window.

5. A method of interlock control of a screw machine according to claim 1, wherein The preset interlocking safety threshold system includes three interlocking control levels that increase sequentially: early warning level, control level, and emergency stop level. Each interlocking control level corresponds to an independent risk probability threshold, control action logic, and timing constraints.

6. A screw machine interlock control method according to claim 5, wherein, The timing constraints include at least the maximum permissible delay time for the interlocking shutdown of the primary screw compressor and the secondary screw compressor, and the response time window for abnormal operating condition protection actions.

7. The screw compressor interlocking control method according to claim 1, characterized in that, The interlocking control model is based on a deep reinforcement learning model, with the safety, response speed and operation stability of the interlocking control as the reward function, and the equipment safety operation boundary and production process requirements of the two-stage screw compressor as the constraint conditions. The two-stage synchronous interlocking control strategy includes the shutdown command timing of the two-stage compressors and the selection of emergency cut-off trigger conditions for the shutdown circuit; The abnormal operating condition graded protection strategy includes graded execution logic corresponding to the interlock control level, such as early warning issuance, load adjustment, intake valve pre-control, and interlock shutdown.

8. The screw compressor interlocking control method according to claim 1, characterized in that, The screw compressor interlocking risk prediction model is a composite architecture of a temporal convolutional network with a two-level equipment coupling attention mechanism and a gated recurrent unit. The model output layer has two parallel branches. The first branch outputs the interlocking failure probability, and the second branch outputs the equipment damage risk level. The model is trained using the historical time-series operation data of the two-stage screw compressor throughout its entire life cycle as the training set, and a multi-task joint loss function is used for training.

9. A method of interlocked control of a screw machine according to claim 5, wherein, The risk deviation is calculated using the formula: D=α×(P-P0) / P0+β×(L-L0) / L0, where D is the risk deviation, P is the real-time interlock failure probability, P0 is the interlock failure probability threshold corresponding to the early warning level, L is the real-time equipment damage risk level, L0 is the equipment damage risk level threshold corresponding to the early warning level, α is the risk weight of the interlock failure probability, β is the risk weight of the equipment damage risk level, and α>β. The weight values ​​are set according to the degree of influence of the parameters on interlock safety.

10. A screw compressor interlocking control method according to claim 4, characterized in that, The interlocking risk disturbance characteristics include the pressure change gradient, load change magnitude, degree of parameter deviation from rated operating range, duration of disturbance event, matching degree of upstream and downstream operating conditions, and correlation degree of interlocking risk.

Images