Waste heat recovery heating system based on ejector heat pump

Abstract

The present application relates to the technical field of system control data processing, and especially relates to a waste heat recovery heating system based on an ejector heat pump, which comprises a multi-source data acquisition module, a genetic algorithm optimization module, a dynamic mode switching module, a thermal stress inhibition module and a self-learning updating module. The multi-source data acquisition module collects industrial waste heat temperature, power grid load and heat network backwater flow data through a temperature sensor array and a flow metering device, and extracts a characteristic vector by using a sliding window standardization and principal component analysis. The heat transfer end difference fluctuation is inhibited through a closed-loop feedback. The self-learning updating module constructs a multi-parameter correlation matrix to dynamically adjust the fitness function weight, and realizes parameter iterative optimization in combination with a historical optimal solution caching mechanism. The present application solves the problems of low industrial waste heat recovery efficiency, thermal coupling regulation hysteresis and equipment thermal stress accumulation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of system control data processing, and in particular to a waste heat recovery heating system based on an ejector heat pump. BACKGROUND

[0002] The waste heat recovery heating system based on an ejector heat pump is based on the principle of thermodynamic injection, mixes and pressurizes high-pressure driving working medium and low-pressure waste heat fluid through an ejector, and realizes the upgrading of low-temperature waste heat grade by using working medium phase change and pressure potential conversion. The existing waste heat recovery heating system based on an ejector heat pump adopts a heat-driven mode, uses industrial waste heat, flue gas or cooling water as energy input, and constructs a cycle through core components such as an evaporator, an ejector, a condenser and a throttle valve. The process of kinetic energy transfer and mixed pressurization of the driving fluid to the ejecting fluid is completed in the ejector, then the high-temperature mixed working medium releases heat to the heating end in the condenser, and finally the efficient recovery of waste heat resources and the adaptation of heating temperature are realized.

[0003] The technical scheme of the present application aims at the problems of low industrial waste heat utilization rate and energy waste caused by close thermal-electric coupling in the existing heating system, and the technical defects of large heat transfer temperature difference in traditional steam-water heat exchangers causing equipment thermal stress accumulation. The traditional process relies on steam turbine extraction heating, which restricts the flexibility of unit operation, and lacks multi-mode adjustment means when the heating load fluctuates. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a waste heat recovery heating system based on an ejector heat pump, which solves the problems of low energy utilization efficiency and equipment thermal stress accumulation caused by insufficient industrial waste heat recovery, strong thermal-electric coupling dependence and large heat transfer temperature difference in traditional steam-water heat exchangers in the existing heating system.

[0005] To solve the above technical problems, the specific technical scheme of the present application is as follows:

[0006] The present application provides a waste heat recovery heating system based on an ejector heat pump, which comprises:

[0007] A multi-source data acquisition module integrates a temperature sensor array, a load monitoring unit and a flow metering device, processes industrial waste heat temperature data through a sliding window standardization method, extracts an electric grid peak shaving demand feature vector using principal component analysis, and outputs a heat network load parameter to a genetic algorithm optimization module. The sliding window standardization method uses a Z-score algorithm to normalize the industrial waste heat temperature data, and the window width is dynamically adjusted according to the heat network load change rate;

[0008] The genetic algorithm optimization module receives the feature vector from the multi-source data acquisition module, encodes the ejector mixing ratio parameter into a binary gene fragment, uses decimal encoding for the valve opening parameter to form a mixed chromosome structure, constructs a sliding time window constraint parameter search space based on the real-time data stream, and generates an initial control parameter solution set including the condensate flow ratio parameter, which is then transmitted to the dynamic mode switching module.

[0009] The dynamic mode switching module parses the chromosome-encoded data of the initial control parameter solution set. When the waste heat-heat supply matching degree deviation exceeds the set threshold, it calls the operating condition data of the electric boiler in the historical power mode database, inputs the power grid peak shaving command into the electric boiler start-up and shutdown probability model to generate start-up and shutdown command candidate solutions, and generates the air extraction adjustment factor candidate solution by combining the trend of the return water flow of the heating network. After performing sliding time window weight allocation calculation on the two types of candidate solutions, it outputs the steam generation weight command to the actuator and sends the mixing ratio correction parameter to the thermal stress suppression module.

[0010] The thermal stress suppression module receives the heat transfer end difference time-series data collected by the distributed temperature field scanner, generates an optimized valve control sequence, and feeds the optimized valve control sequence back to the air extraction adjustment factor calculation unit of the dynamic mode switching module.

[0011] The self-learning update module receives pressure fluctuation suppression data from the thermal stress suppression module and weight allocation history data from the dynamic mode switching module. It then transmits the updated weight coefficients to the population initialization unit of the genetic algorithm optimization module and the candidate solution generation unit of the dynamic mode switching module through a bidirectional data channel.

[0012] Furthermore, the waste heat recovery heating system based on an ejector heat pump described in this invention also includes:

[0013] Specifically, the feature vector output of the multi-source data acquisition module is connected to the chromosome encoding unit of the genetic algorithm optimization module; the initial control parameter solution set output of the genetic algorithm optimization module is connected to the chromosome parsing unit of the dynamic mode switching module; the mixing ratio correction parameter output of the dynamic mode switching module is connected to the valve gradient calculation unit of the thermal stress suppression module; the valve control sequence feedback of the thermal stress suppression module is connected to the pumping volume adjustment factor calculation unit of the dynamic mode switching module; and the weight update signal output of the self-learning update module is connected to the fitness calculation unit of the genetic algorithm optimization module and the weight allocation calculation unit of the dynamic mode switching module, respectively.

[0014] Furthermore, the waste heat recovery heating system based on an ejector heat pump described in this invention includes a thermal stress suppression module, comprising:

[0015] Receive the heat transfer end difference time series data collected by the distributed temperature field scanner, identify the temperature abnormal fluctuation area through the K-means clustering algorithm, generate the valve opening gradient adjustment sequence based on the mixed ratio correction parameter sent by the dynamic mode switching module, calculate the mixed working medium pressure change rate in reverse according to the valve action sequence, select the control scheme that meets the thermal stress accumulation threshold through the elite reservation strategy, feed back the optimized valve control sequence to the steam extraction amount adjustment factor calculation unit of the dynamic mode switching module, the elite reservation strategy retains the top 10% of chromosome individuals in the genetic algorithm iteration process, directly enters the next generation population, and avoids losing high-quality solutions;

[0016] Further, the waste heat recovery heating system based on the ejector heat pump of the application, the self-learning update module comprises:

[0017] Receive the pressure fluctuation suppression data of the thermal stress suppression module and the weight distribution history data of the dynamic mode switching module, construct the correlation matrix of energy cascade utilization efficiency and equipment reliability index, adjust the parameter weight of the genetic algorithm fitness function based on the correlation matrix, and transmit the updated weight coefficient to the population initialization unit of the genetic algorithm optimization module and the candidate solution generation unit of the dynamic mode switching module through the bidirectional data channel.

[0018] Further, the waste heat recovery heating system based on the ejector heat pump of the application, the dynamic mode switching module comprises: an electric power surplus mode analysis unit, inputting the power fluctuation amplitude in the power grid peak shaving instruction into an electric boiler start-stop probability model, outputting a start-stop threshold parameter matched with the current waste heat recovery rate, the electric boiler start-stop probability model is constructed based on historical power peak shaving data, and the mapping relationship between the power grid power fluctuation amplitude and the electric boiler start-stop threshold is calculated through a logistic regression algorithm;

[0019] The waste heat gap compensation unit receives the condensate water diversion ratio parameter output by the genetic algorithm optimization module, combines the real-time monitoring data of the heat network return water flowmeter, and generates a set of steam extraction amount adjustment factor candidate solutions through a sliding average algorithm;

[0020] The start-stop threshold parameter of the electric power surplus mode analysis unit and the candidate solution set of the waste heat gap compensation unit are dynamically weighted in a sliding time window based on a heat network load prediction model, generate a steam generation weight instruction, transmit it to an electric boiler control unit, and send a mixed ratio correction parameter to a valve time sequence mapping unit of the thermal stress suppression module;

[0021] The thermal stress suppression module comprises:

[0022] The temperature field clustering analysis unit performs K-means clustering analysis on the heat transfer end difference time sequence data collected by the distributed temperature scanner, and identifies abnormal region coordinates exceeding a preset fluctuation threshold;

[0023] The valve action optimization unit receives the mixture ratio correction parameter sent by the dynamic mode switching module, generates a valve opening gradient adjustment sequence based on the abnormal region coordinates, and inputs the adjustment sequence to the pressure fluctuation suppression unit;

[0024] The pressure fluctuation suppression unit calculates the mixture working medium pressure change rate according to the valve gradient adjustment sequence, selects a control scheme that meets the preset heat stress accumulation rate through an elite reservation strategy, and feeds back the optimized valve control sequence to the extraction amount adjustment factor candidate solution generation unit of the dynamic mode switching module.

[0025] Further, the self-learning update module of the waste heat recovery heating system based on the ejector heat pump includes: a multi-objective evaluation unit that receives the working medium pressure change rate parameter in the pressure fluctuation suppression data of the heat stress suppression module and the steam generation weight historical value in the weight distribution data of the dynamic mode switching module, and constructs a correlation matrix including the energy cascade utilization efficiency coefficient and the equipment reliability attenuation coefficient;

[0026] The rule base update unit adjusts the weight proportion of the grid peak shaving response speed in the genetic algorithm fitness function based on the coefficient correlation analysis result in the correlation matrix, and transmits the updated weight coefficient to the chromosome coding weight register in the population initialization unit of the genetic algorithm optimization module through the data bus.

[0027] Further, the genetic algorithm optimization module of the waste heat recovery heating system based on the ejector heat pump includes: a chromosome coding unit that converts the ejector mixture ratio parameter into a binary gene fragment according to a preset bit width, and linearly maps the valve opening parameter to a decimal code according to 0-100%, forming a hybrid coding structure including a gene type identifier;

[0028] The fitness calculation unit receives the weight coefficient register data transmitted by the self-learning update module, calculates the waste heat utilization rate in the energy recovery efficiency index, the instruction execution delay time in the grid peak shaving response speed index, and the valve action frequency in the equipment loss rate index, and generates a multi-dimensional evaluation index vector;

[0029] The hybrid coding data stream of the chromosome coding unit and the evaluation index vector of the fitness calculation unit are jointly input into the crossover and mutation operation unit based on the roulette selection strategy to generate a control parameter solution set that meets the waste heat recovery rate threshold and the equipment loss rate threshold.

[0030] Further, the waste heat recovery heating system based on the ejector heat pump provided by the present application further comprises:

[0031] A historical optimal solution cache module stores the optimal chromosome code data of each generation generated by the genetic algorithm optimization module, and records the corresponding waste heat temperature fluctuation characteristic values;

[0032] A working condition matching unit receives the principal component analysis dimension reduction result of the real-time characteristic parameters output by the multi-source data acquisition module, retrieves similar working condition data in the historical optimal solution cache module through a cosine similarity algorithm, and generates an acceleration parameter set including an initial population size and a gene mutation rate;

[0033] The acceleration parameter set of the working condition matching unit is transmitted to the chromosome preloading cache area in the population initialization unit of the genetic algorithm optimization module through a DMA channel, so that the historical optimal chromosome fragments are matched and recombined with the real-time working condition characteristics.

[0034] Further, the waste heat recovery heating system based on the ejector heat pump provided by the present application further comprises:

[0035] A time series data fusion unit performs sliding window smoothing based on Z-score standardization processing on the original sampling data of the industrial waste heat temperature sensor array, and the window width is dynamically adjusted according to the heat network load change rate; a feature dimension reduction unit performs variance contribution rate screening on the standardized multi-dimensional data by using a principal component analysis method, extracts core feature vectors with an accumulated contribution rate exceeding 85%, and generates an initial input parameter set of the genetic algorithm optimization module;

[0036] The core feature vectors output by the feature dimension reduction unit are matched with the working condition data stored in the historical optimal solution cache module through a Mahalanobis distance similarity algorithm, and when the similarity exceeds a set threshold, the elite retention ratio in the population initialization parameter of the genetic algorithm optimization module is adjusted.

[0037] The present application has the following advantages:

[0038] The application obtains industrial waste heat temperature, power grid load and heat network flow data in real time through a multi-source data acquisition module, extracts a core feature vector through principal component analysis and inputs the genetic algorithm optimization module, dynamically optimizes the mixed ratio of an ejector, the opening of a valve and the split ratio of condensate water by using a mixed coding chromosome structure, improves waste heat recovery efficiency and reduces energy waste; a dynamic mode switching module triggers double-path adjustment based on a waste heat-heat supply deviation threshold, realizes accurate adaptation of power peak regulation demand and heat supply load by weight fusion of an electric boiler start-stop probability model and an exhaust gas amount adjustment factor candidate solution, and breaks through the rigid constraint of traditional system thermal-electric coupling; a thermal stress suppression module generates a valve gradient control sequence by combining temperature field clustering analysis and an elite reservation strategy, and realizes real-time correction of mixed working medium pressure fluctuation through a closed-loop feedback mechanism, effectively reducing thermal stress accumulation caused by abnormal fluctuation of heat transfer end difference; a self-learning update module dynamically adjusts the fitness function weight of the genetic algorithm by constructing a multi-parameter correlation matrix, and forms parameter iterative optimization across time dimensions by combining a historical optimal solution cache and a working condition matching mechanism, so that the system can adapt to industrial waste heat source fluctuation and seasonal load change, and comprehensively solve the problems of low energy utilization efficiency and poor equipment reliability. BRIEF DESCRIPTION OF DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. Obviously, other drawings can also be obtained by those skilled in the art without any creative labor under the premise of not deviating from the concept of the present application.

[0040] Figure 1 The non-heating season system operation schematic diagram of the waste heat recovery and heat supply system based on the ejector heat pump provided by the embodiment of the present application.

[0041] Figure 2 The closed cycle schematic diagram of the non-heating season operation mode of the waste heat recovery and heat supply system based on the ejector heat pump provided by the embodiment of the present application.

[0042] Figure 3 The diagram of the industrial waste heat and high back pressure condensate water of the waste heat recovery and heat supply system based on the ejector heat pump provided by the embodiment of the present application can meet the heat supply load

[0043] Figure 4 The diagram of the industrial waste heat and high back pressure condensate water of the waste heat recovery and heat supply system based on the ejector heat pump provided by the embodiment of the present application cannot meet the heat supply load, and there is surplus power in the power grid

[0044] Figure 5 The diagram of the industrial waste heat and high back pressure condensate water of the waste heat recovery and heat supply system based on the ejector heat pump provided by the embodiment of the present application cannot meet the heat supply load, and there is no surplus power in the power grid. DETAILED DESCRIPTION

[0045] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be described clearly and completely below in combination with specific embodiments of the present application and corresponding drawings. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application. The technical solutions provided by the embodiments of the present application will be described in detail below in combination with the drawings. In order to better understand the objects of the present application, the present application will be further described in detail below.

[0046] The present application provides a waste heat recovery heating system based on an ejector heat pump, comprising:

[0047] Please refer to Figures 1 to 5 The present application provides a waste heat recovery heating system based on an ejector heat pump, comprising:

[0048] The multi-source data acquisition module acquires industrial waste heat temperature data in real time through a distributed temperature sensor array deployed at the outlet of the waste heat steam generator, the condenser tube bundle area and the heat network return water pipeline, and obtains cooling water circulation flow parameters in combination with an electromagnetic flowmeter. The temperature and flow data are standardized by using a sliding time window algorithm, the window length is dynamically adjusted according to the heat network load change rate, the waste heat temperature fluctuation standard deviation and the heat network return water flow deviation rate are extracted as core feature vectors. The multi-dimensional data are processed by dimension reduction through a principal component analysis method, the principal components with a cumulative variance contribution rate exceeding 85% are selected as input parameters of a genetic algorithm optimization module, and the influence of redundant data on the optimization process is eliminated. The feature dimension reduction unit calculates the variance contribution rate of the standardized multi-dimensional data by using the principal component analysis method, selects the principal components with a cumulative contribution rate exceeding 85% as core feature vectors after descending arrangement, and eliminates the redundant data dimensions.

[0049] The genetic algorithm optimization module receives the feature vector, converts the ejector mixing ratio parameter into binary code according to the preset bit width, generates decimal code of the valve opening parameter by linear mapping, and forms a mixed chromosome structure including a gene type identifier. A sliding time window constraint parameter search space is constructed based on real-time data flow, cross variation operations are performed within the window, and a control parameter solution set that meets the waste heat recovery rate threshold and the equipment loss rate threshold is screened through a roulette selection strategy. The fitness function calculation unit integrates three indicators of waste heat utilization rate, power grid peak shaving response speed and valve action frequency, and the weight coefficients are dynamically adjusted by a self-learning update module to achieve multi-objective optimization balance. In the chromosome coding unit of the genetic algorithm optimization module, the ejector mixing ratio parameter is converted into a binary gene segment according to the preset bit width (for example, 4 bits), the valve opening parameter is linearly mapped to a decimal code within 0-100%, and a mixed coding structure containing a gene type identifier is formed to ensure the coding compatibility of different parameter types.

[0050] The dynamic mode switching module analyzes the control parameter solution set output by the genetic algorithm, and triggers a multi-path adjustment mechanism when the deviation rate of waste heat supply and heat network load exceeds 15%. In the power surplus mode, the historical database of electric boiler operating conditions is called to convert the power grid peak shaving instruction into a valve start-stop probability threshold, and an electric boiler cooperative control instruction is generated; when there is no power surplus, the exhaust amount compensation coefficient is calculated based on the PID algorithm, and the turbine extraction valve opening is adjusted. The two types of candidate solutions are dynamically fused through a sliding time window weight distribution algorithm, and a steam generation weight instruction is output to the actuator, and the mixed ratio correction parameter is sent to the thermal stress suppression module. The waste heat-heat supply matching degree deviation threshold is determined by analyzing historical operating conditions, and is dynamically adjusted based on the correlation between the heat network load deviation rate and the waste heat temperature fluctuation standard deviation; when the deviation rate exceeds the preset range (for example, 10%-20%), the multi-path adjustment mechanism is triggered.

[0051] The thermal stress suppression module receives distributed temperature field scanning data, identifies abnormal fluctuation areas of the heat transfer end difference by using the K-means clustering algorithm, and generates a valve opening gradient adjustment sequence according to the temperature gradient distribution. The pressure fluctuation suppression unit reversely calculates the influence of valve action on the pressure of the mixed working medium, and selects the optimal control scheme that meets the thermal stress accumulation rate standard through an elite reservation strategy. The optimized valve control sequence is fed back to the candidate solution generation unit of the dynamic mode switching module, forming a closed-loop adjustment mechanism to suppress the influence of abnormal fluctuations of the heat transfer end difference on the stability of the equipment. The temperature field clustering analysis unit performs K-means clustering analysis on the time series data collected by the distributed temperature scanner, sets the initial cluster center as the median of the temperature distribution, iteratively calculates the Euclidean distance between each temperature measurement point and the cluster center, and identifies the abnormal region coordinates that exceed the preset fluctuation threshold.

[0052] The self-learning updating module constructs the correlation matrix of energy cascade utilization efficiency and equipment reliability index, and analyzes the parameter correlation of waste heat utilization rate, valve loss rate and peak regulation response speed. The rule base updating operation is performed every 24 hours, and when it is detected that the ejector mixing ratio and the waste heat utilization rate present strong correlation, the weight proportion of the parameter in the fitness function of the genetic algorithm is improved. The updated weight coefficient is transmitted to the chromosome coding weight register through the data bus, and the population initialization rule is dynamically optimized, and the weight distribution calculation logic of the dynamic mode switching module is corrected in reverse.

[0053] The historical optimal solution cache module stores the optimal chromosome code and the corresponding waste heat temperature fluctuation characteristic value of each generation, and matches the real-time working condition with the historical data through the cosine similarity algorithm. When the core feature vector output by the feature reduction unit has a similarity greater than a set threshold with the cached data, the elite retention ratio adjustment mechanism is triggered to accelerate the convergence speed of the genetic algorithm. The initial population size and gene mutation rate parameters generated by the working condition matching unit are directly written into the chromosome preloading cache area through the DMA channel, so that the historical optimal solution and the real-time running characteristics are quickly recombined.

[0054] In the process of multi-module cooperation, the feature extraction accuracy of the data acquisition module directly affects the optimization effect of the genetic algorithm, and the control parameters output by the optimization module determine the adjustment strategy of the dynamic mode switching. The closed-loop feedback mechanism of the thermal stress suppression module provides data support for the pressure fluctuation of the self-learning updating, and the rule iteration of the self-learning module further optimizes the decision logic of the genetic algorithm and the mode switching module. The historical data cache and working condition matching mechanism form an experience reuse across time dimensions, so that the system can adapt to seasonal load fluctuations and waste heat source instability, and build a full-link intelligent regulation and control system from data acquisition, optimization decision to execution feedback.

[0055] In the technical scheme of the present application, the non-heating season system operation includes the following technical contents:

[0056] Please refer to Figure 1 , open valves 1, 3; close valves 2, 4, 5, 6, 7, 8. The back pressure of the steam turbine is switched to about 8kpa for operation. The cooling water absorbs heat in the condenser, flows to the cooling tower for heat dissipation, and returns to the condenser after cooling, forming a closed cycle. The specific operation mode is shown in please refer to Figure 2 .

[0057] Example 1;

[0058] Please refer to Figure 3 , in Figure 3In the operating mode shown, the multi-source data acquisition module collects steam temperature and return water flow data at a frequency of 1Hz through a temperature sensor (measuring point T1) deployed at the outlet of the waste heat steam generator and a flow meter (measuring point F1) in the heat network return water pipeline. A sliding time window algorithm is used to calculate the standard deviation of the T1 temperature data over 15 consecutive minutes. When the fluctuation exceeds 5℃, the genetic algorithm optimization module is triggered, encoding the ejector mixing ratio parameter into a 4-bit binary gene fragment (accuracy 0.1). The opening parameters of valves 2 / 4 / 7 / 8 are linearly mapped from 0-100% to decimal encoding, generating a mixed chromosome including the condensate flow ratio threshold. The fitness function is based on real-time calculated waste heat utilization rate (actual recovered heat / theoretical maximum value) and valve action frequency (number of opening changes per unit time) to construct evaluation indicators. An optimal solution set is selected using a roulette wheel selection strategy and output to the dynamic mode switching module to generate valve control commands.

[0059] Example 2;

[0060] Please see Figure 4 When the system switches to Figure 4 In the electric boiler supplemental heating mode, the dynamic mode switching module receives the power surplus signal transmitted by the power grid load monitoring unit (lasting ≥30 minutes). It then calls the electric boiler start-stop probability model to convert the power grid peak-shaving command into the opening probability threshold of valve 6 (the electric boiler starts when P≥0.8). Simultaneously, the condensate flow ratio parameter output by the genetic algorithm optimization module is weighted and calculated with the deviation rate of the heating network return water flow to generate coordinated control commands for valves 2 / 4 / 6 / 8. The thermal stress suppression module collects real-time temperature field data in the condenser tube bundle area (e.g., ...). Figure 2 The 5×5 temperature measurement matrix deployed in the dashed box area uses the K-means clustering algorithm to identify abnormal areas with temperature gradients > 2℃ / m, generates valve 7 opening gradient adjustment commands (adjustment amount = historical best value × 0.8 + real-time temperature gradient × 0.2), and feeds the optimized control sequence back to the dynamic mode switching module to form a closed loop.

[0061] Example 3;

[0062] Please refer to Figure 5, in Figure 5In the shown extraction regulation mode, the self-learning update module calls similar working condition data (heat supply network load deviation rate 15%-20% interval) in the historical database, and matches the optimal extraction amount regulation factor through the cosine similarity algorithm. The dynamic mode switching module calculates the compensation opening of valve 5 based on the PID control algorithm (K_p=0.8×ΔQ / rated load, ΔQ is the residual heat gap value), while adjusting the opening combination of valves 2 / 4 / 7 / 8. The pressure fluctuation suppression unit selects the control scheme that meets the heat stress accumulation rate threshold (<0.5 MPa / h) according to the mixed working medium pressure change rate after the action of valve 5, and reversely inputs the optimized parameters into the population initialization unit of the genetic algorithm optimization module to update the extraction amount weight coefficient in the chromosome coding rule.

[0063] Example 4;

[0064] Please refer to Figure 2 , for Figure 2 non-heating season operation mode, the multi-source data acquisition module monitors the closed cycle parameters through the condenser outlet temperature sensor (measurement point T2) and the cooling water flowmeter (measurement point F2). The characteristic dimension reduction unit extracts the main component of T2 temperature fluctuation (cumulative variance contribution rate ≥85%) by using the principal component analysis method. When an abnormal temperature rise is detected (more than 76±2℃ for 3 consecutive sampling points), the heat stress suppression module generates emergency regulation instructions for valves 1 / 3. The self-learning module records the cooling tower heat dissipation efficiency parameters in the non-heating season operation data, which is used to optimize the initial value setting of the condensate water diversion ratio in the heating season mode.

[0065] The present application monitors the industrial residual heat temperature fluctuation and heat supply network load parameters in real time through the multi-source data acquisition module, and dynamically encodes and multi-objectively optimizes the ejector mixing ratio, valve opening and condensate water diversion ratio through the genetic algorithm optimization module. The parameter solution set is constructed by using a mixed chromosome structure, and the optimal control scheme is selected through the roulette selection strategy, so that the residual heat recovery process can adaptively match the changes in heat supply network demand, improve the gradient utilization efficiency of low-grade residual heat, and solve the problem of energy waste caused by insufficient residual heat recovery in traditional systems.

[0066] For the problem of strong thermal-electric coupling dependence, the dynamic mode switching module triggers the double-path regulation mechanism based on the real-time power grid peak shaving instruction and the deviation of residual heat-heat supply matching degree. When there is power surplus, the heat supplement instruction is generated by calling the electric boiler start-stop probability model, and when there is no surplus, the extraction amount regulation factor is calculated through the PID algorithm, and the two types of candidate solutions are fused by combining the sliding time window weight distribution algorithm to dynamically allocate the steam generation weight. This design breaks the traditional single dependence on turbine extraction operation mode, realizes thermal-electric decoupling regulation and control, and enhances the response flexibility of the system in the power peak shaving scenario.

[0067] For the accumulation of thermal stress caused by heat transfer temperature difference, the thermal stress suppression module uses distributed temperature field scanning and clustering analysis technology to identify the abnormal fluctuation area of heat transfer temperature difference and generate valve gradient adjustment sequence. Through the elite retention strategy to screen the pressure fluctuation suppression scheme, the optimized control parameters are fed back to the dynamic mode switching module to form a closed-loop regulation. At the same time, the self-learning update module constructs a multi-parameter correlation matrix, dynamically corrects the fitness function weight of genetic algorithm and the control rule library, continuously optimizes the valve action timing and mixed working fluid pressure stability, and reduces the heat transfer temperature difference from the source and suppresses the growth of thermal stress.

Claims

1. A waste heat recovery heating system based on an ejector heat pump, characterized in that, Comprising: The multi-source data acquisition module integrates temperature sensor array, load monitoring unit and flow metering device, processes industrial waste heat temperature data through sliding window standardization method, extracts power grid peak shaving demand feature vector by principal component analysis, and outputs heat network load parameters to genetic algorithm optimization module; The genetic algorithm optimization module receives the feature vector of the multi-source data acquisition module, encodes the ejector mixing ratio parameter into binary gene fragments, and adopts decimal encoding for the valve opening parameter to form a mixed chromosome structure, constructs a sliding time window constraint parameter search space based on real-time data flow, generates an initial control parameter solution set including condensate water split ratio parameters, and transmits the initial control parameter solution set to the dynamic mode switching module; The dynamic mode switching module analyzes the chromosome encoding data of the initial control parameter solution set, and when the waste heat-heat supply matching degree deviation exceeds a set threshold, calls the electric boiler operating condition data in the historical power mode database, inputs the power grid peak shaving instruction into the electric boiler start-stop probability model to generate a start-stop instruction candidate solution, generates an air extraction amount adjustment factor candidate solution according to the heat network backwater flow change trend, and outputs a steam generation weight instruction to an actuator after sliding time window weight distribution calculation of the two types of candidate solutions, and sends a mixing ratio correction parameter to the thermal stress suppression module; The thermal stress suppression module receives the heat transfer end difference time series data collected by the distributed temperature field scanner, generates an optimized valve control sequence, and feeds back the optimized valve control sequence to the air extraction amount adjustment factor calculation unit of the dynamic mode switching module; The self-learning update module receives the pressure fluctuation suppression data of the thermal stress suppression module and the weight distribution history data of the dynamic mode switching module, and transmits the updated weight coefficients to the population initialization unit of the genetic algorithm optimization module and the candidate solution generation unit of the dynamic mode switching module through a bidirectional data channel; Also Comprising: The feature vector output end of the multi-source data acquisition module is connected to the chromosome coding unit of the genetic algorithm optimization module, the initial control parameter solution set output end of the genetic algorithm optimization module is connected to the chromosome analysis unit of the dynamic mode switching module, the mixing ratio correction parameter output end of the dynamic mode switching module is connected to the valve gradient calculation unit of the thermal stress suppression module, the valve control sequence feedback end of the thermal stress suppression module is connected to the air extraction amount adjustment factor calculation unit of the dynamic mode switching module, and the weight update signal output end of the self-learning update module is connected to the fitness calculation unit of the genetic algorithm optimization module and the weight distribution calculation unit of the dynamic mode switching module.

2. The heat-recovery heating system based on an ejector heat pump according to claim 1, characterized in that, The thermal stress suppression module comprises: Receive the heat transfer end difference time series data collected by the distributed temperature field scanner, identify the temperature abnormal fluctuation area by K-means clustering algorithm, generate the valve opening gradient adjustment sequence based on the mixing ratio correction parameter sent by the dynamic mode switching module, calculate the mixed working medium pressure change rate in reverse according to the valve action sequence, select the control scheme that meets the thermal stress accumulation threshold value through the elite reservation strategy, and feed back the optimized valve control sequence to the air extraction amount adjustment factor calculation unit of the dynamic mode switching module.

3. The heat-recovery heating system based on an ejector heat pump according to claim 1, characterized in that, The self-learning update module comprises: The pressure fluctuation suppression data of the thermal stress suppression module and the weight distribution history data of the dynamic mode switching module are received, an association matrix of energy cascade utilization efficiency and equipment reliability index is constructed, the parameter weight of the fitness function of the genetic algorithm is adjusted based on the association matrix, and the updated weight coefficient is transmitted to the population initialization unit of the genetic algorithm optimization module and the candidate solution generation unit of the dynamic mode switching module through a bidirectional data channel.

4. The heat-recovery heating system based on an ejector heat pump according to claim 1, characterized by, The dynamic mode switching module includes: a power surplus mode analysis unit, which inputs the power fluctuation amplitude in the grid peak shaving instruction into the electric boiler start-stop probability model, and outputs the start-stop threshold parameter matched with the current waste heat recovery rate; A waste heat gap compensation unit receives the condensate water diversion ratio parameter output by the genetic algorithm optimization module, combines the real-time monitoring data of the heat network return water flowmeter, and generates a set of candidate solutions of the extraction amount adjustment factor through a sliding average algorithm; The start-stop threshold parameter of the power surplus mode analysis unit and the candidate solution set of the waste heat gap compensation unit are dynamically weighted within a sliding time window based on the heat network load prediction model, to generate a steam generation weight instruction transmitted to the electric boiler control unit, and a mixing ratio correction parameter sent to the valve timing mapping unit of the thermal stress suppression module; The thermal stress suppression module includes: A temperature field clustering analysis unit performs K-means clustering analysis on the heat transfer terminal difference time series data collected by the distributed temperature scanner, and identifies the abnormal area coordinates exceeding the preset fluctuation threshold; A valve action optimization unit receives the mixing ratio correction parameter sent by the dynamic mode switching module, generates a valve opening gradient adjustment sequence based on the abnormal area coordinates, and inputs the adjustment sequence into the pressure fluctuation suppression unit; A pressure fluctuation suppression unit calculates the mixing working medium pressure change rate according to the valve gradient adjustment sequence, selects a control scheme that meets the preset thermal stress accumulation rate through an elite reservation strategy, and feeds back the optimized valve control sequence to the extraction amount adjustment factor candidate solution generation unit of the dynamic mode switching module.

5. The heat-recovery heating system based on an ejector heat pump according to claim 1, characterized in that, The self-learning update module includes: a multi-objective evaluation unit that receives the working medium pressure change rate parameter in the pressure fluctuation suppression data of the thermal stress suppression module and the steam generation weight history value in the weight distribution data of the dynamic mode switching module, and constructs an association matrix including the energy cascade utilization efficiency coefficient and the equipment reliability decay coefficient; A rule base update unit adjusts the weight proportion of the grid peak shaving response speed in the fitness function of the genetic algorithm based on the correlation analysis results of the coefficients in the association matrix, and transmits the updated weight coefficient to the chromosome coding weight register in the population initialization unit of the genetic algorithm optimization module through the data bus.

6. The heat-recovery heating system based on an ejector heat pump according to claim 1, characterized in that, The genetic algorithm optimization module includes: a chromosome coding unit that converts the ejector mixing ratio parameter into a binary gene fragment according to a preset bit width, and linearly maps the valve opening parameter to a decimal code according to 0-100%, to form a mixed coding structure including a gene type identifier; The fitness calculation unit receives the weight coefficient register data transmitted by the self-learning update module, calculates the waste heat utilization rate in the energy recovery efficiency index, the instruction execution delay time in the power grid peak regulation response speed index, and the valve action frequency in the equipment loss rate index, and generates a multi-dimensional evaluation index vector; The mixed encoding data stream of the chromosome encoding unit and the evaluation index vector of the fitness calculation unit are jointly input into the crossover and mutation operation unit based on the roulette selection strategy to generate a control parameter solution set that meets the waste heat recovery rate threshold and the equipment loss rate threshold.

7. The heat-recovery heating system based on an ejector heat pump according to claim 1, characterized in that, The system further comprises: A historical optimal solution cache module stores the historical optimal chromosome encoding data generated by the genetic algorithm optimization module and records the corresponding waste heat temperature fluctuation characteristic values; A working condition matching unit receives the principal component analysis dimension reduction result of the real-time characteristic parameters output by the multi-source data acquisition module, retrieves similar working condition data in the historical optimal solution cache module through the cosine similarity algorithm, and generates an acceleration parameter set including the initial population size and the gene mutation rate; The acceleration parameter set of the working condition matching unit is transmitted to the chromosome preloading cache area in the population initialization unit of the genetic algorithm optimization module through the DMA channel, so that the matching and recombination of the historical optimal chromosome fragment and the real-time working condition characteristics are realized.

8. The waste heat recovery heating system based on an ejector heat pump according to any one of claims 1-7, characterized in that, The multi-source data acquisition module comprises: A time series data fusion unit performs sliding window smoothing based on Z-score standardization processing on the original sampling data of the industrial waste heat temperature sensor array, and the window width is dynamically adjusted according to the heat network load change rate; a feature dimension reduction unit uses the principal component analysis method to perform variance contribution rate screening on the standardized multi-dimensional data, extracts the core feature vector with a cumulative contribution rate exceeding 85%, and generates an initial input parameter set for the genetic algorithm optimization module; The core feature vector output by the feature dimension reduction unit is matched with the working condition data stored in the historical optimal solution cache module through the Mahalanobis distance similarity matching algorithm, and when the similarity exceeds the set threshold, the elite retention ratio adjustment in the population initialization parameter of the genetic algorithm optimization module is triggered.

Images