Dynamic quantitative feeding control system and method based on real-time working condition feedback of reaction kettle

Abstract

The application discloses a dynamic quantitative feeding control system and method based on real-time working condition feedback of a reaction kettle, and comprises the following steps: collecting reaction kettle temperature, pressure, flow and other data, performing denoising and standardization, and generating working condition input data set; calculating heat release intensity, energy residual and material residual based on the working condition input data set, and constructing a physical characteristic vector; inputting the physical characteristics into an improved echo state network, and outputting temperature and pressure prediction; constructing temperature, pressure and rate boundary conditions to generate a safety constraint set according to the prediction results and equipment limitations; using self-evolution waveform optimization based on the safety constraint to generate a candidate waveform set and select the optimal feeding curve; inputting the optimal waveform to control feeding, establishing a trust propagation graph, calculating a score gate and triggering a safety interlock. The application realizes accurate prediction and safety control of reaction kettle feeding through the improved echo state network and self-evolution waveform optimization.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial automation control and process control, and particularly relates to a dynamic quantitative feeding control system and method based on real-time working condition feedback of a reaction kettle. BACKGROUND

[0002] With the rapid development of hydrogen energy industry, hydrogen production equipment, hydrogen fuel cell systems and hydrogen refueling stations are increasingly widely used in the fields of industry, transportation and energy storage. Hydrogen production and reaction processes generally involve high pressure, high temperature and complex catalytic reaction systems. One of the core equipment, the reaction kettle, plays a key role in the synthesis of hydrogen production materials, the preparation of hydrogen storage media and the production of electrode active materials. However, in the hydrogen production reaction process, the reaction heat release intensity is large, the heat conduction coupling is complex, and the feeding rate directly affects the reaction stability and hydrogen production efficiency. Traditional feeding control methods mostly use fixed rate or proportional-integral-derivative (PID) control, which cannot dynamically adjust according to the real-time working condition of the reaction kettle, leading to overshoot, lag or fluctuation of temperature and pressure, and thus affecting product purity and equipment safety.

[0003] At present, some enterprises have introduced intelligent sensing and data monitoring means in the process of hydrogen energy system integration and hydrogen production equipment manufacturing, but the working condition modeling still stays at the static control level, which is difficult to accurately capture the nonlinear characteristics of the reaction system under different loads and thermal states. Especially in the catalytic hydrogenation and high-temperature decomposition hydrogen production link, the raw material conversion rate and heat release characteristics change significantly over time, and traditional models cannot make forward-looking predictions and adaptive adjustments for such dynamic processes. Existing safety interlocking mechanisms are usually based on threshold judgment, with a lagging response and no intelligent discrimination ability, which is prone to imbalance between false positives and false negatives, affecting the continuity and reliability of hydrogen production equipment.

[0004] Therefore, how to provide a dynamic quantitative feeding control system and method based on real-time working condition feedback of a reaction kettle is a problem that needs to be solved by those skilled in the art. SUMMARY

[0005] One purpose of the present application is to provide a dynamic quantitative feeding control system and method based on real-time working condition feedback of a reaction kettle, which comprehensively uses intelligent control technologies such as improved echo state network modeling, dynamic fusion of physical characteristics, multi-objective self-evolution waveform optimization, and trust propagation type closed-loop control, and builds a dynamic control system that can self-learn, self-adapt, and be predictable, aiming at the problems of difficult to guarantee feeding accuracy, parameter lag, and slow response of safety interlocking of the reaction kettle under complex working conditions. The present application generates an optimal feeding waveform online by modeling the temperature, pressure, and energy characteristics of the reaction kettle in real time, and realizes safety gating and model self-calibration through the trust propagation mechanism during execution, so as to realize high-precision regulation and active safety control of the feeding process. The present application has the advantages of high prediction accuracy, fast control response, strong safety, and superior system self-learning ability, and improves the stability of the reaction kettle operation and the automation level of the production process.

[0006] The dynamic quantitative feeding control method based on real-time working condition feedback of a reaction kettle according to the embodiment of the present application comprises:

[0007] Collecting real-time working condition data of the reaction kettle, pre-processing the real-time working condition data to form a working condition input data set;

[0008] Based on the working condition input data set, the reaction heat release intensity, energy balance residual error, and material balance residual error are calculated according to the energy conservation and mass conservation relationship, added to the working condition input data set, and a physical characteristic vector containing the heat release intensity index, energy residual error, and material residual error is constructed;

[0009] The physical characteristic vector is taken as the input to establish an improved echo state network model, the working condition modeling of multiple time scales is realized through the state interaction of the fast reservoir and the slow reservoir, the memory weight is dynamically adjusted by using the energy balance error driven cross-layer migration matrix, and the temperature and pressure prediction results are output;

[0010] Based on the temperature and pressure prediction results, the safety constraint conditions are constructed by combining the upper and lower limits of the temperature, pressure, and change rate allowed by the equipment, and a safety constraint set is generated;

[0011] According to the safety constraint set and the temperature and pressure prediction results, a multi-objective self-evolution waveform generation mechanism is used to parameterize the feeding waveform, a candidate waveform set is generated, the candidate waveforms are evolved and optimized through the morphological variation operator, constraint correction operator, and smoothness reservation operator, the comprehensive cost function is calculated, and the optimal feeding waveform is selected;

[0012] The optimal feeding waveform is input into a control execution unit to control the variable frequency pump or electric valve to perform dynamic quantitative feeding operation, a trust propagation graph of the cognitive decision layer, sensor layer and actuator layer is constructed, a comprehensive trust score is calculated and gate control is performed, when the comprehensive trust score is lower than a threshold value, a safety interlock is triggered, and echo state network model parameters are updated according to feedback data.

[0013] Optionally, the real-time working condition data includes temperature, pressure, liquid level, stirring speed, feeding flow rate, jacket temperature and feed temperature inside the reaction kettle.

[0014] Optionally, the preprocessing of the real-time working condition data includes synchronous sampling, denoising filtering and standardization processing of the real-time working condition data, and abnormal data is removed and missing data is interpolated and compensated.

[0015] Optionally, the physical feature vector containing the heat release intensity index, energy residual and material residual is constructed, including:

[0016] The temperature, pressure, liquid level, feeding flow rate, jacket temperature and feed temperature are read from the working condition input data set according to the time stamp;

[0017] The temperature change rate and liquid level change rate are determined according to the difference between adjacent time samples and the sampling time interval, the total amount of material in the kettle is determined using the liquid level, the equivalent cross-sectional area of the kettle body and the material density, the mass flow rate of the feed is determined using the feeding flow rate and the feed density, and the cumulative rate of the material is determined from the change in the total amount of material at adjacent time and the sampling time interval;

[0018] According to the principle of heat conservation, the reaction heat release intensity is obtained by subtracting the enthalpy flow carried by the feed from the sum of the heat change of the kettle contents and the heat exchange between the kettle and the jacket, wherein the heat change is determined by the total amount of material in the kettle, the constant-pressure specific heat and the temperature change rate, the heat exchange is determined by the heat transfer coefficient, the heat transfer area and the difference between the kettle temperature and the jacket temperature, and the enthalpy flow is determined by the mass flow rate of the feed, the constant-pressure specific heat of the feed and the difference between the feed temperature and the kettle temperature;

[0019] According to the principle of mass conservation, the material balance residual is obtained from the difference between the cumulative rate of the material and the mass flow rate of the feed, and according to the principle of heat conservation, the energy balance residual is determined from the difference between the sum of the heat change of the kettle contents and the heat transfer and the heat carried by the feed and the reaction heat release intensity;

[0020] The reaction heat release intensity, energy balance residual and material balance residual are aligned with the corresponding time stamp to form a physical feature vector containing the heat release intensity index, energy residual and material residual.

[0021] Optionally, the output temperature and pressure prediction results include:

[0022] The working condition input data set and the physical feature vector are read according to the timestamp, and the two are spliced to form a comprehensive feature vector;

[0023] An improved echo state network model is established, which is composed of a fast dynamic expression layer, a slow time domain memory layer and a cross-layer memory migration-reading fusion layer, and the initial values of the state dimensions, leakage rates, spectral radii and interlayer coupling weights of each layer are set;

[0024] The comprehensive feature vector is input into the fast dynamic expression layer, an impact suppression buffer is set to limit the amplitude and slope of high-frequency mutation variables, a residual sensitive gate is set, and the fast dynamic expression layer state update is triggered or suppressed according to the energy balance residual, and the fast dynamic state is output;

[0025] The comprehensive feature vector and the fast dynamic state are input into the slow time domain memory layer, multi-resolution sliding aggregation is performed to form trend features, an adaptive forgetting factor is scheduled according to the material balance residual to control the memory retention time, a change point detection is set to refresh part of the memory window when a working condition mutation is detected, and the slow time domain memory state is output;

[0026] In the cross-layer memory migration-reading fusion layer, the cross-layer migration weight is calculated based on the energy balance residual and the material balance residual, the memory migration is performed between the fast dynamic state and the slow time domain memory state to form a fusion state, the cross-layer consistency check and rollback mechanism are performed, and the temperature prediction value and the pressure prediction value of the next time are output;

[0027] Multi-step rolling prediction is performed starting from the temperature prediction value and the pressure prediction value, and the temperature and pressure prediction results of multiple future time steps are gradually generated according to the prediction step length.

[0028] Optionally, the generating the safety constraint set comprises:

[0029] The temperature and pressure prediction results and the uncertainty range are obtained, and the sampling period and the prediction step length are read;

[0030] The temperature lower limit, the temperature upper limit, the pressure lower limit and the pressure upper limit are set according to the device operation safety range, and the temperature and pressure values at each prediction time are checked point by point, if the prediction value exceeds the set upper and lower limits, the time is recorded as a boundary risk point, and the corresponding temperature and pressure boundary constraint condition is generated;

[0031] According to the allowed temperature change rate upper limit and the pressure change rate upper limit, the temperature change rate and the pressure change rate between adjacent prediction times are calculated, and it is judged whether they are within the allowed range, if they exceed the allowed rate, the time is recorded as a rate risk point, and the corresponding rate constraint condition is generated;

[0032] According to the uncertainty information of the prediction result, a temperature uncertainty amplification coefficient and a pressure uncertainty amplification coefficient are set, the temperature and pressure prediction values are added to the corresponding safety margin, and boundary checking is performed again, a threshold is set, and the energy balance residual and the material balance residual are checked at each time, if the two residuals exceed the threshold, they are marked as physical abnormal points, and robustness constraint conditions and physical constraint conditions are generated;

[0033] The temperature and pressure boundary constraint conditions, the rate constraint conditions, the robustness constraint conditions and the physical constraint conditions are integrated to form a safety constraint set.

[0034] Optionally, the calculating the comprehensive cost function and selecting the optimal feeding waveform comprises:

[0035] The temperature and pressure prediction results, the uncertainty information and the safety constraint set are obtained as input conditions for waveform generation and checking;

[0036] A waveform representation mode of a multi-objective self-evolution waveform generation mechanism is established, a morphology grammar bank composed of segmented rising sections, platform sections, descending sections, S-shaped sections and micro-pulse sections is used to parameterize the description of the feeding waveform, and a feeding quantity quota scheduler is set to allocate and recover the feeding quotas of each section, so that the total feeding quantity meets the allowable deviation of the target feeding quantity;

[0037] The double-channel evolution process of the self-evolution operator set is performed on the candidate waveform set:

[0038] The morphology channel is used for adding, deleting and rearranging the morphology sections and switching the section types, and the parameter channel is used for adjusting the amplitude, duration and slope of each section;

[0039] The morphology mutation operator, the constraint correction operator and the smoothness reservation operator are applied in sequence, and the safety margin embedding operator is introduced to automatically reduce the local slope or insert a buffer section when the prediction approaches the boundary;

[0040] The falsification rollback operator is introduced to exclude the candidate waveforms that do not meet the safety constraint set in advance, and the intra-section micro-pulse shaping operator is introduced to refine the duty cycle and interval of the micro-pulse;

[0041] For each candidate waveform after evolution, an improved echo state network model is called for forward-looking evaluation to obtain the temperature and pressure response trajectory within the prediction step, and the safety constraint set is checked at each time, the candidate waveforms that violate the constraint are directly excluded, and the candidate waveforms that meet the constraint are calculated for comprehensive cost;

[0042] The candidate waveform with the minimum comprehensive cost is selected as the optimal feeding waveform when the safety constraint set is met, when there are parallel optimals, the one with lower peak temperature and more gentle flow change is selected preferentially, and the optimal feeding waveform is output.

[0043] Optionally, the trust propagation graph of the cognitive decision layer, the sensor layer and the actuator layer is constructed, the comprehensive trust score is calculated and the gate control is performed, comprising:

[0044] The optimal feeding waveform is received and issued to the control execution unit, and the trust propagation graph is established, which is composed of the cognitive decision layer nodes, the sensor layer nodes and the actuator layer nodes, and the directed relationship edges between the nodes are set for transmitting and attenuating trust degree;

[0045] During the execution of the optimal feeding waveform, real-time feedback data is collected, the temperature prediction value and the pressure prediction value at the corresponding time, the energy balance residual and the material balance residual at the corresponding time are read, and the observation consistency index of each node is generated and written into the trust propagation graph according to the consistency of the actual value and the prediction value and the closeness of the residual and the threshold value;

[0046] The relationship edge weight and the node trust degree in the trust propagation graph are updated according to the observation consistency index, the current comprehensive trust score is calculated, the first gate threshold and the second interlocking threshold are set, and the second interlocking threshold is smaller than the first gate threshold;

[0047] The gate control is performed according to the comprehensive trust score, the first gate threshold and the second interlocking threshold:

[0048] When the comprehensive trust score is lower than the first gate threshold and not lower than the second interlocking threshold, the conservative feeding mode is switched, and the upper limit of the feeding flow and the flow change rate are limited;

[0049] When the comprehensive trust score is lower than the second interlocking threshold, the safety interlocking is triggered, and the feeding is immediately stopped and the cooling device and the pressure relief device are linked in a preset order;

[0050] The control instructions after the gate control or interlocking are issued to the frequency conversion pump and the electric valve, and the corresponding node and constraint trigger information are recorded;

[0051] According to the real-time feedback data and the gate execution result, the improved echo state network model is updated online, including the adjustment of the recursive correction and the cross-layer memory migration configuration of the read-out parameters.

[0052] The dynamic quantitative feeding control system based on the real-time working condition feedback of the reaction kettle according to the embodiments of the present application comprises the following modules:

[0053] The data acquisition module is used for acquiring the real-time working condition data of the reaction kettle and performing preprocessing to generate the working condition input data set;

[0054] The physical feature construction module is used for calculating the reaction heat release intensity, the energy residual and the material residual based on the working condition input data set to form the physical feature vector;

[0055] A modeling prediction module is configured to establish an improved echo state network model, implement multi-time scale modeling through state interaction, and output temperature and pressure prediction results.

[0056] A safety constraint construction module is configured to set upper and lower limits of temperature, pressure and change rate according to the temperature and pressure prediction results and the allowable range of the equipment, and generate a safety constraint set.

[0057] A waveform optimization module is configured to generate a candidate feeding waveform, and select an optimal feeding waveform through morphological variation, constraint correction and smoothness reservation optimization.

[0058] A trust control module is configured to issue the optimal feeding waveform, construct a trust propagation graph, calculate a comprehensive trust score and perform gated control, and trigger a safety interlock when the score is lower than a threshold.

[0059] The present application has the following advantages:

[0060] The present application introduces an improved echo state network model to achieve multi-time scale dynamic modeling of complex conditions of a reaction kettle. The synergistic effect of fast and slow reserve pools and cross-layer migration matrix can not only capture the rapid fluctuation characteristics of temperature, pressure and other parameters in the reaction process, but also effectively represent the long-term thermodynamic trend of the system, thereby achieving high-precision prediction of the reaction dynamics. Compared with traditional static models or single-layer neural network structures, the model of the present application has stronger time sequence memory ability and generalization performance, improving the predictability and stability of the model.

[0061] The present application uses a multi-objective self-evolution waveform generation mechanism to enable adaptive optimization of the feeding strategy under safety constraints. This mechanism considers multiple factors such as temperature, pressure change rate and actuator constraints, and dynamically optimizes the feeding waveform through self-evolution operators such as morphological variation, constraint correction and smoothness reservation, thereby achieving smooth feeding and energy release balance in the reaction process while meeting the total quantitative requirements. Compared with traditional fixed waveforms or manually set strategies, the present application can automatically generate an optimal feeding curve under complex reaction conditions, significantly reducing the risk of temperature and pressure overshoot, and improving the controllability and consistency of the reaction.

[0062] The present application introduces a trust propagation type closed-loop control mechanism in the control execution phase, constructs a trust propagation graph of the cognitive decision layer, sensor layer and actuator layer, calculates a comprehensive trust score in real time, and performs gated and interlocked control. This mechanism can automatically switch to a conservative feeding mode or trigger a safety interlock when the system trust level decreases or an abnormal trend is detected, thereby achieving active safety protection of the feeding process. The echo state network model parameters are dynamically updated based on real-time feedback data, enabling the system to have self-learning and self-correction capabilities, thereby achieving high precision, strong robustness and high safety of the reaction kettle feeding control. BRIEF DESCRIPTION OF DRAWINGS

[0063] The accompanying drawings are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and are meant to explain the present application and are not intended to limit the application. In the drawings:

[0064] Figure 1 A flow chart of the dynamic quantitative feeding control method based on real-time working condition feedback of a reaction kettle according to the present application;

[0065] Figure 2 A structural schematic diagram of the dynamic quantitative feeding control system based on real-time working condition feedback of a reaction kettle according to the present application. DETAILED DESCRIPTION

[0066] The present application will now be further described in detail with reference to the drawings. These drawings are simplified schematic diagrams and only schematically show the basic structure of the present application, and thus only show the components related to the present application.

[0067] Reference Figure 1 The dynamic quantitative feeding control method based on real-time working condition feedback of a reaction kettle comprises:

[0068] Collecting real-time working condition data of the reaction kettle, pre-processing the real-time working condition data to form a working condition input data set;

[0069] Based on the working condition input data set, the reaction heat release intensity, energy balance residual error and material balance residual error are calculated according to the energy conservation and mass conservation relationship, added to the working condition input data set, and a physical feature vector containing the heat release intensity index, energy residual error and material residual error is constructed;

[0070] The physical feature vector is taken as the input, an improved echo state network model is established, the working condition modeling of multiple time scales is realized through the state interaction of the fast reservoir pool and the slow reservoir pool, the memory weight is dynamically adjusted by using the energy balance error driven cross-layer migration matrix, and the temperature and pressure prediction results are output;

[0071] Based on the temperature and pressure prediction results, the safety constraint conditions are constructed by combining the upper and lower limits of the temperature, pressure and change rate allowed by the equipment, and a safety constraint set is generated;

[0072] According to the safety constraint set and the temperature and pressure prediction results, a multi-objective self-evolution waveform generation mechanism is used to parameterize the feeding waveform, a candidate waveform set is generated, the candidate waveforms are evolved and optimized through the shape variation operator, constraint correction operator and smoothness reservation operator, the comprehensive cost function is calculated, and the optimal feeding waveform is selected;

[0073] The optimal feeding waveform is input into a control execution unit to control the variable frequency pump or electric valve to perform dynamic quantitative feeding operation, a trust propagation graph of the cognitive decision layer, sensor layer and actuator layer is constructed, a comprehensive trust score is calculated and gate control is performed, when the comprehensive trust score is lower than a threshold value, a safety interlock is triggered, and echo state network model parameters are updated according to feedback data.

[0074] In the embodiment, the real-time working condition data includes temperature, pressure, liquid level, stirring speed, feeding flow, jacket temperature and feed temperature inside the reaction kettle.

[0075] In the embodiment, the preprocessing of the real-time working condition data includes synchronous sampling, denoising filtering and standardization processing of the real-time working condition data, and abnormal data is removed and missing data is interpolated and compensated.

[0076] In the embodiment, the physical feature vector containing the heat release intensity index, energy residual and material residual is constructed, including:

[0077] The temperature, pressure, liquid level, feeding flow, jacket temperature and feed temperature are read from the working condition input data set according to the time stamp;

[0078] The temperature change rate and liquid level change rate are determined according to the difference between adjacent time samples and the sampling time interval, the total amount of material in the kettle is determined using the liquid level, the equivalent cross-sectional area of the kettle body and the material density, the mass flow of the feed is determined using the feeding flow and the feed density, and the cumulative rate of the material is determined from the change of the total amount of material at adjacent time and the sampling time interval;

[0079] According to the law of conservation of heat, the reaction heat release intensity is obtained by subtracting the enthalpy flow carried by the feed from the sum of the heat change of the kettle content and the heat exchange between the kettle and the jacket, wherein the heat change is determined by the total amount of material in the kettle, the constant-pressure specific heat and the temperature change rate, the heat exchange is determined by the heat transfer coefficient, the heat transfer area and the difference between the kettle temperature and the jacket temperature, and the enthalpy flow is determined by the mass flow of the feed, the constant-pressure specific heat of the feed and the difference between the feed temperature and the kettle temperature;

[0080] According to the law of conservation of mass, the material balance residual is obtained from the difference between the cumulative rate of the material and the mass flow of the feed, and according to the law of conservation of heat, the energy balance residual is determined from the difference between the sum of the heat change of the kettle content and the heat transfer and the heat carried by the feed and the reaction heat release intensity;

[0081] The reaction heat release intensity, energy balance residual and material balance residual are aligned with the corresponding time stamp to form a physical feature vector containing the heat release intensity index, energy residual and material residual.

[0082] In the embodiment, the output temperature and pressure prediction results include:

[0083] The working condition input data set and the physical feature vector are read according to the timestamp, and the two are spliced to form a comprehensive feature vector;

[0084] An improved echo state network model is established, which is composed of a fast dynamic expression layer, a slow time domain memory layer and a cross-layer memory migration-reading fusion layer, and the initial values of the state dimensions, leakage rates, spectral radii and interlayer coupling weights of each layer are set;

[0085] The comprehensive feature vector is input into the fast dynamic expression layer, a shock suppression buffer is set to limit the amplitude and slope of high-frequency mutation variables, a residual sensitive gate is set, and the fast dynamic expression layer state update is triggered or suppressed according to the energy balance residual, and the fast dynamic state is output;

[0086] The comprehensive feature vector and the fast dynamic state are input into the slow time domain memory layer, multi-resolution sliding aggregation is performed to form trend features, an adaptive forgetting factor is scheduled according to the material balance residual to control the memory retention time, a change point detection is set to refresh part of the memory window when a working condition mutation is detected, and the slow time domain memory state is output. The multi-resolution sliding aggregation is specifically:

[0087] The input fast dynamic state sequence is segmented and slidingly aggregated on different time scales, and statistical features in short-term, medium-term and long-term windows are calculated respectively to capture change trends at different time levels;

[0088] In the sliding window of each time scale, local fluctuation features and global change features are fused by weighted average and difference accumulation to generate a multi-resolution trend description vector;

[0089] The aggregation results under different time scales are dynamically weighted and combined according to time weights to form a multi-resolution aggregation output with time level correlation, which is used as the input feature of the slow time domain memory layer;

[0090] In the cross-layer memory migration-reading fusion layer, the cross-layer migration weight is calculated based on the energy balance residual and the material balance residual, the memory migration is performed between the fast dynamic state and the slow time domain memory state to form a fusion state, the cross-layer consistency check and rollback mechanism are executed, and the temperature prediction value and the pressure prediction value of the next moment are output;

[0091] Multi-step rolling prediction is performed starting from the temperature prediction value and the pressure prediction value, and temperature and pressure prediction results of future time steps are gradually generated according to the prediction step length.

[0092] In the embodiment, the generation of the safety constraint set includes:

[0093] The temperature and pressure prediction results and the uncertainty range are obtained, and the sampling period and the prediction step length are read;

[0094] According to the device operation safety range setting temperature lower limit, temperature upper limit, pressure lower limit and pressure upper limit, the temperature and pressure values at each prediction time are checked point by point, if the prediction value exceeds the set upper and lower limits, the time is recorded as a boundary risk point, and the corresponding temperature and pressure boundary constraint conditions are generated;

[0095] According to the allowed temperature change rate upper limit and pressure change rate upper limit, the temperature change rate and pressure change rate between adjacent prediction times are calculated, and it is judged whether it is within the allowed range, if it exceeds the allowed rate, the time is recorded as a rate risk point, and the corresponding rate constraint condition is generated;

[0096] According to the uncertainty information of the prediction result, set the temperature uncertainty amplification coefficient and the pressure uncertainty amplification coefficient, add the corresponding safety margin to the temperature and pressure prediction value, and then perform boundary check again, set the threshold, check the energy balance residual and material balance residual at each time, if the two residuals exceed the threshold, mark it as a physical abnormal point, and generate robustness constraint condition and physical constraint condition;

[0097] Integrate the temperature and pressure boundary constraint conditions, rate constraint conditions, robustness constraint conditions and physical constraint conditions to form a safety constraint set.

[0098] In the embodiment, the calculation of the comprehensive cost function includes:

[0099] Obtain the temperature and pressure prediction results, uncertainty information and safety constraint set as the input conditions for waveform generation and checking;

[0100] Establish a waveform representation method of multi-objective self-evolution waveform generation mechanism, use a morphology grammar bank composed of segmented rising section, platform section, descending section, S-shaped section and micro-pulse section to parameterize the description of the feeding waveform, set the feeding amount quota scheduler to allocate and recover the feeding quota of each section, so that the total feeding amount meets the allowed deviation of the target feeding amount;

[0101] Perform a double-channel evolution process of a set of self-evolution operators on the candidate waveform set:

[0102] The morphology channel is used for adding, deleting and rearranging the morphology section, switching the section type, the parameter channel is used for adjusting the amplitude, duration and slope of each section;

[0103] Apply the morphology mutation operator, constraint correction operator and smoothness preservation operator in turn, and introduce a safety margin embedding operator to automatically reduce the local slope or insert a buffer section when the prediction approaches the boundary;

[0104] The proof-by-disproof rollback operator is introduced to eliminate the candidate waveform that does not meet the safety constraint set in advance, and the intra-segment micro-pulse shaping operator is introduced to refine the duty cycle and interval of the micro-pulse;

[0105] For each evolved candidate waveform, the improved echo state network model is called for forward-looking evaluation to obtain the temperature and pressure response trajectory within the prediction step, and the safety constraint set is checked at each time. The candidate waveform that violates the constraint is directly eliminated, and the comprehensive cost of the candidate waveform that meets the constraint is calculated. The comprehensive cost of the candidate waveform that meets the constraint is calculated, specifically:

[0106] According to the sum of squares of deviations of the temperature and pressure prediction trajectory, the stability cost of the working condition is calculated, which is used to measure the steady-state control performance of the waveform in the dynamic response process;

[0107] According to the change amplitude of the feeding rate and the smoothness of the waveform, the execution smoothing cost is calculated, which is used to evaluate the controllability and equipment wear risk of the waveform in the execution phase;

[0108] The comprehensive economic cost is calculated by combining the reaction completion time, energy utilization rate and safety margin factor, and the three cost items are normalized and weighted according to the weight to obtain the comprehensive cost value of the candidate waveform, which is used to guide the selection of the optimal waveform;

[0109] The candidate waveform with the minimum comprehensive cost is selected as the optimal feeding waveform when the safety constraint set is met. When there are parallel optimals, the one with lower peak temperature and more gentle flow change is preferred. The optimal feeding waveform is output.

[0110] In the embodiment, the trust propagation graph of the cognitive decision layer, the sensor layer and the actuator layer is constructed, the comprehensive trust score is calculated and the gate control is performed, including:

[0111] The optimal feeding waveform is received and issued to the control execution unit, and the trust propagation graph is established. The trust propagation graph is composed of cognitive decision layer nodes, sensor layer nodes and actuator layer nodes. The nodes are connected by directed relationship edges for transmitting and attenuating trust degree;

[0112] During the execution of the optimal feeding waveform, real-time feedback data is collected, temperature and pressure prediction values at the corresponding time, energy balance residual and material balance residual at the corresponding time are read, and observation consistency indexes of each node are generated and written into the trust propagation graph according to the consistency of actual value and prediction value and the closeness of residual and threshold. The real-time feedback data includes temperature, pressure, feeding flow and valve opening degree;

[0113] According to the observation consistency index, the relationship edge weight and node trust degree in the trust propagation graph are updated, the current comprehensive trust score is calculated, the first gating threshold and the second interlocking threshold are set, the second interlocking threshold is less than the first gating threshold, and the current comprehensive trust score is calculated.

[0114] The real-time state confidence of the nodes of the comprehensive sensor layer, the actuator layer and the cognitive decision layer is integrated, and the local trust score is calculated according to the signal consistency and delay deviation between nodes;

[0115] In the trust propagation graph, the trust scores of the nodes are weighted, accumulated and normalized along the information transmission path to obtain a global trust vector after interlayer propagation;

[0116] The weighted average of each node in the global trust vector is taken as the comprehensive trust score output, which is used to reflect the overall consistency and operation reliability of the system;

[0117] According to the comprehensive trust score, the first gating threshold and the second interlocking threshold, the gating control is performed:

[0118] When the comprehensive trust score is lower than the first gating threshold and not lower than the second interlocking threshold, the conservative feeding mode is switched, and the upper limit of the feeding flow and the flow change rate are limited;

[0119] When the comprehensive trust score is lower than the second interlocking threshold, the safety interlocking is triggered, and the feeding is immediately stopped and the cooling device and the pressure relief device are linked in a preset order;

[0120] The control instructions after gating or interlocking are sent to the frequency conversion pump and the electric valve, and the corresponding node and constraint trigger information are recorded;

[0121] According to the real-time feedback data and the gating execution result, the improved echo state network model is updated online, including recursive correction of readout parameters and adjustment of cross-layer memory migration configuration.

[0122] Reference Figure 2 , a dynamic quantitative feeding control system based on real-time working condition feedback of a reaction kettle, comprising the following modules:

[0123] A data acquisition module is used to acquire real-time working condition data of the reaction kettle and perform preprocessing to generate working condition input data set;

[0124] A physical feature construction module is used to calculate reaction heat release intensity, energy residual and material residual based on the working condition input data set to form a physical feature vector;

[0125] A modeling and prediction module is used to establish an improved echo state network model, realize multi-time scale modeling through state interaction, and output temperature and pressure prediction results;

[0126] a safety constraint construction module configured to set upper and lower limits of temperature, pressure and change rate according to the temperature and pressure prediction results and the device allowable range, and generate a safety constraint set;

[0127] a waveform optimization module configured to generate a candidate feeding waveform, and select an optimal feeding waveform through shape variation, constraint correction and smoothness reservation optimization;

[0128] a trust control module configured to issue the optimal feeding waveform, construct a trust propagation graph, calculate a comprehensive trust score and perform gate control, and trigger a safety interlock when the score is lower than a threshold.

[0129] Example 1: In order to verify the feasibility of the present application in implementation, the present application is applied to a liquid organic hydrogen storage material preparation system of a certain hydrogen energy equipment manufacturing enterprise. The effective volume of the reaction kettle of the system is 2.5 m³, which is used for catalytic hydrogenation reaction, and mainly generates hydrogen energy carrier material (LOHC). The process has typical strong exothermic characteristics, and the feeding rate and heat management requirements are extremely high. The traditional PID control system often leads to temperature overshoot, pressure lag and uneven heat release due to the inability to respond to the change in reaction rate in time, which reduces the hydrogen production efficiency, and even triggers the safety interlock shutdown, affecting the continuous operation of the production line.

[0130] In the hydrogen production system, the dynamic quantitative feeding control method proposed by the present application is adopted, and the system configuration includes a data acquisition module, a physical characteristic construction module, a modeling prediction module, a safety constraint construction module, a waveform optimization module and a trust control module. The reaction kettle is collected in real time through six types of sensors such as temperature, pressure, liquid level, flow, stirring speed, etc., and the sampling frequency is 1 Hz. The multi-dimensional signals collected are subjected to data denoising and time synchronization by the pre-processing module to form a standardized operating condition input data set. Then, the physical characteristic construction module calculates the reaction heat release intensity, energy balance residual and material balance residual according to the energy conservation and mass conservation relationship, adds them to the input data set, and constructs the physical characteristic vector of the reaction kettle.

[0131] In the modeling stage, the system adopts an improved echo state network model to realize multi-time scale prediction through the collaborative modeling of fast reservoir pool, slow reservoir pool and cross-layer dynamic memory migration unit. The model can dynamically adjust the memory weight according to the energy balance error, thereby enhancing the prediction accuracy under nonlinear time-varying conditions. Experimental results show that the prediction error of the model for temperature and pressure in the next 30 seconds is controlled within ±1.2℃ and ±0.013 MPa, respectively, which is about 65% higher in accuracy than the traditional neural network.

[0132] Subsequently, the safety constraint modeling module automatically sets the upper limit of temperature 320℃, the upper limit of pressure 0.65MPa and the threshold of change rate according to the prediction results and the equipment design limits, and generates a safety constraint set. The waveform optimization module adopts a multi-objective self-evolution waveform generation mechanism to generate multiple candidate feeding waveform schemes, and performs evolution optimization through shape variation, constraint correction and smoothness reservation operators. The system performs forward-looking evaluation and cost calculation on each candidate waveform, and finally selects the feeding waveform with the minimum comprehensive cost and meeting the safety constraints as the control instruction.

[0133] In the execution phase, the control system transmits the optimal waveform signal to the variable frequency pump to realize continuous and gradual dynamic quantitative feeding. At the same time, the trust control module establishes a trust propagation graph of the cognitive decision layer, the sensor layer and the actuator layer, monitors the node signal consistency and delay, and calculates the comprehensive trust score. When the score drops below the threshold due to signal drift or sudden changes in working conditions, the system immediately performs gate control to reduce the feeding rate or temporarily stop operation, and starts the safety interlocking mechanism. Through feedback data, the model automatically updates parameters to realize self-learning and closed-loop optimization.

[0134] In a 72-hour continuous running pilot test, the system completed 6 complete reaction batches. Compared with the traditional PID control results, the method significantly improves the stability of temperature and pressure control.

[0135] Table 1 Comparison of dynamic quantitative feeding control performance of reaction kettle

[0136]

[0137] As can be seen from Table 1, the dynamic quantitative feeding control performance of the method in the hydrogen energy preparation reaction system is significantly better than that of the traditional PID control method. In terms of temperature and pressure control, the method achieves higher process stability. The average reaction temperature is 316.5℃, which is basically the same as 316.2℃ under traditional PID control, but the temperature fluctuation amplitude is significantly reduced from ±6.1℃ to ±2.5℃, a decrease of 59%. At the same time, the pressure fluctuation amplitude is reduced from ±0.10MPa to ±0.05MPa, a decrease of 50%. This shows that the improved echo state network model of the application can effectively predict the reaction thermodynamic changes and adjust the feeding strategy in advance, thereby avoiding the overshoot phenomenon of temperature and pressure, and improving the thermal stability and safety of the reaction system.

[0138] From the production efficiency and energy efficiency performance, the method of the present application realizes higher reaction rate and energy utilization rate without increasing energy consumption. The average feeding rate is increased from 4.2 kg / min to 4.3 kg / min, the total feeding time is shortened from 51 minutes to 46 minutes, and the overall process efficiency is increased by about 10%. The energy utilization rate is increased from 83.4% to 87.8%, indicating that the system can maintain high heat recovery and transfer efficiency during dynamic feeding adjustment. In terms of model prediction accuracy, the temperature prediction error is reduced from ±3.4℃ to ±1.2℃, and the pressure prediction error is reduced from ±0.038MPa to ±0.013MPa, which is increased by 65% and 66% respectively, verifying the high-precision modeling capability of the improved echo state network under multi-time scale dynamic working conditions.

[0139] Finally, in terms of safety and intelligent control, the method of the present application shows significant robustness and adaptability. The traditional PID control system triggers the safety interlock 2.1 times on average in batch operation, while the method of the present application does not trigger false alarm in all batches, fully demonstrating that the trust propagation type safety control mechanism can dynamically adjust the system response according to the node signal consistency and trust score, fundamentally reducing false positives and delays. The average trust score of the system is 93.1%, showing high level of signal consistency and decision reliability. In summary, the present application not only improves the feeding accuracy and thermal control stability of the reaction kettle in the process of hydrogen energy preparation, but also takes into account the energy efficiency, safety and intelligence level, and has wide industrial promotion and application value.

[0140] The above is only the preferred embodiment of the present application, but the protection scope of the present application is not limited thereto, any skilled person in the art can make equivalent replacement or change according to the technical solution and inventive concept of the present application within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A dynamic quantitative feeding control method based on real-time operating condition feedback of a reactor, characterized in that, include: Collect real-time operating data of the reactor, preprocess the real-time operating data, and form an operating condition input dataset; Based on the operating condition input dataset, according to the relationship between energy conservation and mass conservation, the reaction exothermic intensity, energy balance residual and material balance residual are calculated and added to the operating condition input dataset to construct a physical feature vector containing the exothermic intensity index, energy residual and material residual. Using physical feature vectors as input, an improved echo state network model is established. Multi-timescale operating condition modeling is achieved through the state interaction between the fast and slow reservoirs. The memory weights are dynamically adjusted using the cross-layer migration matrix driven by energy balance error, and the temperature and pressure prediction results are output. Based on the temperature and pressure prediction results, and combined with the upper and lower limits of the allowable temperature, pressure and rate of change of the equipment, safety constraints are constructed and a set of safety constraints is generated. Based on the safety constraint set and temperature and pressure prediction results, a multi-objective self-evolution waveform generation mechanism is adopted to parameterize the feeding waveform, generate a candidate waveform set, and optimize the candidate waveforms through morphological variation operator, constraint correction operator and smoothness preservation operator, calculate the comprehensive cost function, and select the optimal feeding waveform. The optimal feeding waveform is input into the control execution unit to control the variable frequency pump or electric valve to perform dynamic quantitative feeding operation. A trust propagation diagram of the cognitive decision layer, sensor layer and actuator layer is constructed. The comprehensive trust score is calculated and gating control is performed. When the comprehensive trust score is lower than the threshold, the safety interlock is triggered and the echo state network model parameters are updated based on the feedback data.

2. The dynamic quantitative feeding control method based on real-time operating condition feedback of the reactor according to claim 1, characterized in that, The real-time operating data includes the temperature, pressure, liquid level, stirring speed, feed flow rate, jacket temperature, and feed temperature inside the reactor.

3. The dynamic quantitative feeding control method based on real-time operating condition feedback of the reactor according to claim 1, characterized in that, The preprocessing of real-time operating data includes synchronous sampling, noise reduction filtering and standardization of real-time operating data, removal of abnormal data and interpolation compensation for missing data.

4. The dynamic quantitative feeding control method based on real-time operating condition feedback of the reactor according to claim 1, characterized in that, The construction of the physical feature vector, which includes the exothermic intensity index, energy residual, and material residual, includes: Read temperature, pressure, liquid level, feed flow rate, jacket temperature and feed temperature from the working condition input dataset by timestamp; The temperature change rate and liquid level change rate are determined based on the difference between samples at adjacent time points and the sampling time interval. The total amount of material in the vessel is determined using the liquid level, the equivalent cross-sectional area of ​​the vessel body, and the material density. The feed mass flow rate is determined using the feed flow rate and the feed density. The material accumulation rate is determined by the change in the total amount of material at adjacent time points and the sampling time interval. Based on the principle of heat conservation, the reaction exothermic intensity is obtained by subtracting the enthalpy flow carried by the feed from the sum of the heat change of the contents of the vessel and the heat exchange between the vessel and the jacket. The heat change is determined by the total amount of material in the vessel, the specific heat at constant pressure, and the rate of temperature change. The heat exchange is determined by the heat transfer coefficient, the heat transfer area, and the difference between the temperature inside the vessel and the temperature in the jacket. The enthalpy flow is determined by the feed mass flow rate, the specific heat at constant pressure of the feed, and the difference between the feed temperature and the temperature inside the vessel. Based on the principle of mass conservation, the material balance residual is obtained by the difference between the material accumulation rate and the feed mass flow rate. Based on the principle of heat conservation, the energy balance residual is determined by the difference between the sum of the heat change of the contents of the vessel and the heat transfer, minus the heat carried by the feed, and the intensity of the exothermic reaction. Align the reaction exothermic intensity, energy balance residual, and material balance residual with their corresponding timestamps to form a physical feature vector that includes the exothermic intensity index, energy residual, and material residual.

5. The dynamic quantitative feeding control method based on real-time operating condition feedback of the reactor according to claim 1, characterized in that, The output temperature and pressure prediction results include: Read the working condition input dataset and physical feature vector according to the timestamp, and concatenate the two to form a comprehensive feature vector; An improved echo state network model is established, which consists of a fast dynamic expression layer, a slow time-domain memory layer, and a cross-layer memory transfer-readout fusion layer. Initial values ​​are set for the state dimension, leakage rate, spectral radius, and inter-layer coupling weight of each layer. The comprehensive feature vector is input into the fast dynamic expression layer. An impact suppression buffer is set to limit the amplitude and slope of high-frequency mutations. A residual sensitive gate is set to trigger or suppress the state update of the fast dynamic expression layer according to the energy balance residual, and the fast dynamic state is output. The comprehensive feature vector and the fast dynamic state are input into the slow time domain memory layer, multi-resolution sliding aggregation is performed to form trend features, and the memory retention time is controlled based on the material balance residual scheduling adaptive forgetting factor. Change point detection is set to refresh part of the memory window when a sudden change in the working condition is detected, and the slow time domain memory state is output. Within the cross-layer memory migration-readout fusion layer, cross-layer migration weights are calculated based on energy balance residuals and material balance residuals. Memory migration is performed between the fast dynamic state and the slow time-domain memory state to form a fusion state. Cross-layer consistency verification and rollback mechanisms are executed, and the temperature and pressure prediction values ​​for the next moment are output. Starting with the predicted temperature and pressure values, multi-step rolling predictions are performed, and the predicted temperature and pressure results for multiple future time steps are generated step by step according to the prediction step size.

6. The dynamic quantitative feeding control method based on real-time operating condition feedback of the reactor according to claim 1, characterized in that, The generated set of security constraints includes: Obtain the temperature and pressure prediction results and uncertainty range, and read the sampling period and prediction step size; Based on the safe operating range of the equipment, set the lower temperature limit, upper temperature limit, lower pressure limit, and upper pressure limit. Check the temperature and pressure values ​​at each predicted time point by point. If the predicted value exceeds the set upper and lower limits, record the time as the boundary risk point and generate the corresponding temperature and pressure boundary constraints. Based on the upper limits of the allowable temperature change rate and pressure change rate, calculate the temperature change rate and pressure change rate between adjacent prediction times, determine whether they are within the allowable range, and if they exceed the allowable rate, record the time as the rate risk point and generate the corresponding rate constraint conditions. Based on the uncertainty information of the prediction results, set the temperature uncertainty amplification factor and the pressure uncertainty amplification factor, add the corresponding safety margin to the predicted temperature and pressure values ​​and then perform boundary checks again. Set thresholds and check the energy balance residuals and material balance residuals at each time step. If the two residuals exceed the thresholds, they are marked as physical anomalies, and robustness constraints and physical constraints are generated. The temperature and pressure boundary constraints, rate constraints, robustness constraints, and physical constraints are integrated to form a set of safety constraints.

7. The dynamic quantitative feeding control method based on real-time operating condition feedback of the reactor according to claim 1, characterized in that, The calculation of the comprehensive cost function and the selection of the optimal feeding waveform include: The temperature and pressure prediction results, uncertainty information, and safety constraint set are obtained as input conditions for waveform generation and verification. A waveform representation method for a multi-objective self-evolutionary waveform generation mechanism is established. A morphological syntax library consisting of segmented rising segments, plateau segments, falling segments, S-shaped segments, and micro-pulse segments is used to parameterize the feeding waveform. A feeding quota scheduler is set up to allocate and recover the feeding quota of each segment so that the total feeding amount meets the allowable deviation of the target feeding amount. Perform a two-channel evolution process of self-evolution operators on the candidate waveform set: The shape channel is used to add, delete, and rearrange shape segments, and switch segment types; the parameter channel is used to adjust the amplitude, duration, and slope of each segment. The morphological mutation operator, constraint correction operator, and smoothness preservation operator are applied in sequence, and a safety margin embedding operator is introduced to automatically reduce the local slope or insert a buffer segment when the prediction is close to the boundary. A falsification backoff operator is introduced to eliminate candidate waveforms that do not meet the set of safety constraints in advance, and an intra-segment micro-pulse shaping operator is introduced to refine the micro-pulse duty cycle and interval; For each candidate waveform after evolution, the improved echo state network model is called to perform look-ahead evaluation to obtain the temperature and pressure response trajectory within the prediction step. The model is checked time by time according to the set of safety constraints. Candidate waveforms that violate the constraints are directly eliminated, and the comprehensive cost is calculated for candidate waveforms that meet the constraints. Among the candidate waveforms that meet the safety constraints, the one with the lowest overall cost is selected as the optimal feeding waveform. When there are ties for the best, the one with the lower peak temperature and the smoother flow rate change is selected first, and the optimal feeding waveform is output.

8. The dynamic quantitative feeding control method based on real-time operating condition feedback of the reactor according to claim 1, characterized in that, The construction of a trust propagation graph comprising the cognitive decision-making layer, sensor layer, and actuator layer, the calculation of a comprehensive trust score, and the implementation of gating control include: The optimal feeding waveform is received and sent to the control execution unit to establish a trust propagation graph. The trust propagation graph consists of nodes of the cognitive decision layer, each measuring point node of the sensor layer, and each execution mechanism node of the actuator layer. Directed relation edges are set between the nodes to transmit and attenuate the trust degree. During the execution of the optimal feeding waveform, real-time feedback data is collected, and the predicted temperature and pressure values, energy balance residuals and material balance residuals at the corresponding times are read. Based on the consistency between the actual values ​​and the predicted values, and the closeness between the residuals and the thresholds, the observation consistency index of each node is generated and written into the trust propagation graph. Update the relation edge weights and node trust levels in the trust propagation graph based on the observation consistency index, calculate the current comprehensive trust score, and set a first gate threshold and a second interlocking threshold, with the second interlocking threshold being less than the first gate threshold; Gating control is executed based on the comprehensive trust score, the first gating threshold, and the second interlocking threshold. When the comprehensive trust score is lower than the first gate threshold and not lower than the second interlocking threshold, it switches to conservative feeding mode to limit the upper limit of feeding flow and the rate of flow change. When the overall trust score is lower than the second interlock threshold, the safety interlock is triggered, the feeding is immediately stopped, and the cooling device and the pressure relief device are activated in a preset sequence. The control commands after gate control or interlocking are sent to the variable frequency pump and electric valve, and the corresponding node and constraint trigger information are recorded. Based on real-time feedback data and gating execution results, the improved echo state network model is updated online, including recursive correction of readout parameters and adjustment of cross-layer memory migration configuration.

9. A dynamic quantitative feeding control system based on real-time operating condition feedback of a reactor, executing the dynamic quantitative feeding control method based on real-time operating condition feedback of a reactor as described in any one of claims 1 to 8, characterized in that, Includes the following modules: The data acquisition module is used to collect real-time operating data of the reactor and preprocess it to generate an operating condition input dataset. The physical feature construction module is used to calculate the reaction exothermic intensity, energy residual, and material residual based on the operating condition input dataset, and form a physical feature vector. The modeling and prediction module is used to build an improved echo state network model, realize multi-timescale modeling through state interaction, and output temperature and pressure prediction results. The safety constraint construction module is used to set upper and lower limits for temperature, pressure and rate of change based on temperature and pressure prediction results and the equipment's allowable range, and generate a set of safety constraints. The waveform optimization module is used to generate candidate feeding waveforms and optimize them through morphological variation, constraint correction and smoothness preservation to select the optimal feeding waveform; The trust control module is used to send out the optimal feeding waveform, construct a trust propagation graph, calculate the comprehensive trust score and execute gating control. When the score is lower than the threshold, a safety interlock is triggered.

Images