Water quality sampling control method and control system for environmental detection

Abstract

The application discloses a water quality sampling control method and control system for environmental detection, and relates to the technical field of water quality sampling control; the control method comprises the following steps: constructing a system state vector, and adopting a cooperative operation model cluster and loading; wherein the cooperative operation model cluster at least comprises a state recognition classification model, a sampling time prediction model, an improved resource optimization scheduling model and a self-adaptive feedback control model; the technical key points are as follows: dynamic adaptation is realized through multi-model cooperation, the problem that fixed logic cannot cope with dynamic environment is solved, intelligentization and self-adaptive operation of water quality sampling control are realized, and through linkage and cooperation of each model in the cooperative operation model cluster, water quality sampling work can independently adjust an operation strategy under complex working conditions such as power failure, low power and water pressure fluctuation, and task integrity and data representativeness are ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of water quality sampling control, in particular to a water quality sampling control method and system for environmental detection. BACKGROUND

[0002] For the water quality sampling control of underground wells, strict specifications must be followed to ensure the representativeness of the samples; before sampling, the underground water level must be measured, and the water in the well must be fully pumped out, the sampling depth should be below 0.5 meters below the underground water surface to exclude the influence of stagnant water; for closed production wells, sampling should be done at the water outlet valve of the pump house, and the water storage should be emptied; the sampling container should be pretreated according to the detection items, washed with detergent and distilled water to avoid affecting the subsequent sampling results; the sampler used during sampling can work on sewage inspection wells, equipped with a stainless steel shell, powered by lithium batteries, installed with a sling, and remotely controlled at close range, suitable for long-term unattended automatic sampling in wells, and can be used as an ideal tool for third-party detection agencies to inspect well samples; after sampling, a sealed label can be set to save information such as monitoring well number, time, items and sampler, and finally the sampling plan and records are checked to ensure that nothing is missed.

[0003] The prior art has many technical bottlenecks in the energy saving and data quality control of underground water sampling:

[0004] For example, the traditional sampling scheme lacks quantitative evaluation means, such as intermittent closing of the water pump or reducing the number of well flushing, which is often operated by experience, resulting in damaged data representativeness, usually manifested as: insufficient well flushing causing VOCs samples to mix with stagnant water, resulting in a large deviation in the detection results; secondly, the static threshold setting cannot adapt to the dynamic environment, such as using fixed parameters for sampling in low-permeability formations in well water, which easily causes sediment disturbance, while blind parameter adjustment may lead to substandard sample sampling; in addition, a single control logic is difficult to cope with complex underground conditions, if the related water quality sampling equipment is not switched to the standby pump in time after power failure, it will result in long-term loss of sampling window, and then affect the subsequent data re-measurement cost problem; therefore, how to improve and optimize the adjustment of the underground water quality sampling control is the key problem to be solved at present. SUMMARY

[0005] To achieve the above purpose, the present application is implemented by the following technical solutions:

[0006] A water quality sampling control method for environmental detection, comprising the steps of: constructing a system state vector and loading a collaborative work model cluster; wherein the collaborative work model cluster at least includes: a state recognition classification model, a sampling time prediction model, an improved resource optimization scheduling model, and a self-adaptive feedback control model;

[0007] According to the system state vector, the state recognition classification model runs the classification rule engine, and outputs the mode judgment result;

[0008] The sampling time prediction model is run, the predetermined sampling time sequence is constructed, and the sliding window matching action is completed;

[0009] The improved resource optimization scheduling model is run, the product of the residual task energy consumption and the safety margin coefficient is compared with the current available power, and according to the comparison result, whether to trigger the hierarchical adjustment model is selected, and the hierarchical adjustment model adopts a multi-level simulation adjustment strategy, and according to different adjustment levels, simulation is carried out, and the corresponding adjustment level that satisfies: the current available power≥the product of the residual task energy consumption and the safety margin coefficient is selected, and a first adjustment set is formed; The task integrity estimate value obtained by introducing the task integrity weight function is compared with the preset standard threshold value, and the comparison result is used as a constraint condition, and according to the constraint condition, the corresponding adjustment level is selected from the first adjustment set, and a second adjustment set is formed, and the adjustment level corresponding to the maximum value of the task integrity estimate value is selected from the second adjustment set;

[0010] The historical sampling feedback data set is counted, and the parameter correction model is constructed to correct the multi-level simulation adjustment strategy;

[0011] An adaptive feedback control model is used to realize the flow control of the washing process and the adaptive adjustment of the emptying backflushing time.

[0012] Further, a system state vector is constructed, which includes a multi-dimensional vector including: battery voltage, environmental temperature, last task state, last sampling timestamp, and current sampling mode code; In the last task state: 0=not completed, 1=completed; The content loaded after the cluster of the collaborative work model includes at least: the preset time period of the sampling time prediction model; The energy consumption coefficient of the improved resource optimization scheduling model, which is used as a reference for single standard sampling power consumption; The initial control parameters of the adaptive feedback control model: proportional gain Kp, integral gain Ki, and differential gain Kd.

[0013] Further, the process of running the classification rule engine is: if the last task is not completed and the last sampling timestamp is earlier than the current time, it is determined that: the interruption recovery mode; If the last task has been completed or has never been executed, it is determined that: standby configuration mode; If the current time is equal to the next preset sampling time point, it is determined that: immediate sampling mode.

[0014] Further, when constructing the predetermined sampling time sequence, the predetermined sampling time sequence is generated based on a preset time period Δt; the execution process of the sliding window matching action is: defining a time tolerance window δ, and checking whether the current time falls within the tolerance interval of any unfinished predetermined sampling time Tn; wherein the tolerance interval is: Tn- δ≤ current time≤ Tn+ δ; if it is satisfied, the supplement sampling action is triggered immediately; otherwise, it jumps to the next Tn+1.

[0015] Further, the multi-level simulation adjustment strategy includes: adjustment level L1: reducing the sampling volume according to the preset gradient: from the original sampling volume V_set to the present sampling volume V_set×b1; wherein b1 is the first adjustment coefficient corresponding to the preset gradient, and b1∈(0, 1); adjustment level L2: reducing the number of rinsing according to the preset interval: from the original rinsing number R_wash to the present rinsing number R_wash-1; wherein the minimum value of the original rinsing number R_wash is 2, and the original rinsing number R_wash is a positive integer; adjustment level L3: shortening the emptying and back flushing time according to the preset proportion: from the original emptying and back flushing time t_blow to the present emptying and back flushing time t_blow×b2; wherein b2 is the second adjustment coefficient corresponding to the preset proportion, and b2∈(0, 1).

[0016] Further, the way of introducing the task integrity weight function is:

[0017] W_iy=w1×(V_ac / V_ta)+w2×(R_ac / R_ta)+w3×(t_ac / t_ta); wherein, W_iy is the task integrity estimate, and the value is in [0, 1]; w1, w2 and w3 are weight coefficients, and 1> w1> w2> w3> 0; V_ac and V_ta are the actual executed sampling volume and the target sampling volume respectively, R_ac and R_ta are the actual executed rinsing number and the target rinsing number respectively, t_ac and t_ta are the actual executed emptying and back flushing time and the target emptying and back flushing time respectively; the comparison result of the task integrity estimate W_iy and the preset standard threshold value is as follows: when the task integrity estimate W_iy exceeds the preset standard threshold value, it is retained; otherwise, it is rejected.

[0018] Further, the historical sampling feedback data set at least includes: the number of effective samples and the total number of samples under the condition of using the first adjustment coefficient, the number of blockage alarms and the total number of emptying and back flushing under the condition of using the second adjustment coefficient;

[0019] The process of constructing and running the parameter correction model is:

[0020] When the first adjustment coefficient is dealt with, an effectiveness function is established: η_V(b1)=N_vd(b1) / N_tl(b1); in the formula, η_V(b1) is the sample effectiveness rate under the b1, N_vd(b1) and N_tl(b1) are respectively the number of samples detected effectively and the total number of samples when the b1 is used; the target threshold is set: η_V(b1)≥ηmin must be met, wherein ηmin represents the target threshold; the setting is corrected: when the set target threshold is not met in the continuous S times of tasks, the up-regulation action is performed; otherwise, the down-regulation action is performed;

[0021] When the second adjustment coefficient is dealt with, a clogging risk function is established: ρ_cl(b2)=C_cl(b2) / N_bl(b2); in the formula, ρ_cl(b2) is the clogging occurrence rate per emptying backflush operation, C_cl(b2) is the number of clogging alarms under the b2 condition, and N_bl(b2) is the total emptying backflush number; the safety threshold is set: ρ_cl(b2)≤ρmax must be met, wherein ρmax represents the safety threshold; the setting is corrected: when the set safety threshold is not met, the up-regulation action is performed; otherwise, the down-regulation action is performed.

[0022] Further, when the flow control of the rinsing process is implemented: the target is to ensure that a set volume of water sample is extracted each time; variables are defined at least including: the target rinsing volume, the actual volume extracted at time t, the volume error, the control output, and the PID controller parameter set; wherein the target rinsing volume is denoted as Q _ta, the actual volume extracted at time t is denoted as Q _ac(t), which is obtained by integrating the pump speed wz(t) and time, the volume error e(t) is: e(t)=Q _ta-Q _ac(t); the control output u(t) is used to adjust the pump driving power; the PID controller parameter set is the initial control parameter of the adaptive feedback control model; and the control output u(t) is obtained according to the PID controller parameter set.

[0023] Further, when the emptying backflush time is adaptively adjusted: the target is to dynamically adjust the emptying backflush time according to the environment temperature; variables are defined at least including: the final executed emptying backflush duration t_ac', the actually executed emptying backflush duration t_ac, the current environment temperature T_e, the reference environment temperature T_ref, and the temperature compensation coefficient d1; the way according to which the emptying backflush time is adaptively adjusted is: t_ac'=t_ac×[1+d1×(T_e-T_ref)]; in the formula, the temperature compensation coefficient d1 takes a value in the range of (0, 0.1).

[0024] A water quality sampling control system for environmental detection, comprising: an initialization module: constructing a system state vector, adopting a collaborative work model cluster and loading; wherein the collaborative work model cluster at least includes: a state recognition classification model, a sampling time prediction model, an improved resource optimization scheduling model and a self-adaptive feedback control model;

[0025] A state recognition module: according to the system state vector, the state recognition classification model runs a classification rule engine, and outputs a mode judgment result;

[0026] A sampling prediction module: running the sampling time prediction model, constructing a predetermined sampling time sequence, and completing a sliding window matching action;

[0027] An energy consumption and task coordination module: running the improved resource optimization scheduling model, comparing the product of the residual task energy consumption and the safety margin coefficient with the current available power, selecting whether to trigger the hierarchical adjustment model according to the comparison result, and the hierarchical adjustment model adopts a multi-level simulation adjustment strategy, simulates according to different adjustment levels, selects the corresponding adjustment level that meets: the current available power is greater than or equal to the product of the residual task energy consumption and the safety margin coefficient, and forms a first adjustment set; introducing the task integrity evaluation value obtained by the task integrity weight function, taking the comparison result between the task integrity evaluation value and the preset standard threshold as a constraint condition, and according to the constraint condition, the corresponding adjustment level with the maximum task integrity evaluation value is selected from the second adjustment set to form a second adjustment set;

[0028] A feedback correction submodule: statistics historical sampling feedback data set, and construct a parameter correction model, realize the correction of the multi-level simulation adjustment strategy;

[0029] A self-adaptive control module: adopting the self-adaptive feedback control model, realizing the flow control and the adaptive adjustment of the emptying and back blowing time in the washing process.

[0030] The present application provides a water quality sampling control method and control system for environmental detection, which has the following beneficial effects:

[0031] The present application introduces a task integrity weight function, realizes quantitative evaluation, converts the fuzzy influence on sampling quality into a calculable value, selects the scheme with the smallest data influence in multiple energy saving paths, realizes intelligent decision-making, wherein the weight coefficient involved can be preset by the user or the scene, and is suitable for different monitoring requirements, in addition, the integrity score is recorded every time the adjustment is made, which is convenient for later auditing and has traceability;

[0032] The scheme introduces safety margin judgment and multi-level adjustment strategy, effectively saves energy consumption on the premise of ensuring sample quality, can significantly improve the energy efficiency ratio of single task, prolongs the endurance time of related water quality sampling equipment in the field work, on the other hand, through learning and parameter correction of historical data, further optimizes the application scene and threshold setting of each adjustment level, so that the whole system can continuously self-improve in long-term operation, finally realizes the balance adjustment operation between task completion rate and energy consumption, ensures the effectiveness of the whole control work;

[0033] The scheme can accurately control the sampling time and process variable through the precise time sequence sliding window algorithm and the closed-loop adjustment mechanism based on the extended PID, thereby greatly improving the accuracy of data acquisition, which not only helps to eliminate the error caused by external interference factors, ensures the consistency and repeatability of data, but also can make intelligent decision combined with the task integrity weight function, maximally reduces the influence of various energy-saving measures on the final analysis result, and protects the authenticity and reliability of monitoring data;

[0034] The scheme realizes dynamic adaptation through multi-model cooperation, solves the problem that fixed logic cannot cope with dynamic environment, realizes the intelligentization and adaptive operation of water quality sampling control, and through the linkage of each model in the cooperative operation model cluster, the water quality sampling work can independently adjust the operation strategy under complex working conditions such as power failure, low power and water pressure fluctuation, and ensure the task integrity and data representativeness. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 It is a flowchart of the water quality sampling control method in the application.

[0036] Figure 2 It is a flowchart of the improved resource optimization scheduling model operation in the application. DETAILED DESCRIPTION

[0037] The technical solutions in the embodiments of the application will be described clearly and completely in the embodiments of the application combined with the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the application.

[0038] Embodiment 1:

[0039] Please refer to Figure 1 and Figure 2The embodiment provides a water quality sampling control method for environmental detection, and mainly aims at underground well water sampling. A system overall architecture and a model integration framework are designed. The water quality sampling control system carried by the framework adopts a four-layer closed loop architecture from perception to decision to execution to feedback. The first core point lies in introducing the following four algorithms / models and constructing a collaborative operation mechanism. Specifically, a state identification classification model based on a rule engine and a lightweight decision tree logic is used for identifying the running phase of the system; a sampling time prediction model based on a time series sliding window algorithm is used for dynamically predicting and compensating the next sampling time; a resource optimization scheduling model adopting a greedy strategy and a battery energy consumption model is used for optimizing task execution and energy distribution; and an adaptive feedback control model based on a closed loop adjustment mechanism of an extended PID (Proportion Integration Differentiation) thought is used for dynamically adjusting the execution process such as rinsing, sampling and emptying. The above basic models are improved to realize the second core point, that is, a multi-improved model collaborative decision engine is formed by integrating the central control unit to drive the intelligent operation of the overall control scheme.

[0040] Specifically, the corresponding control system used in the embodiment is carried in a water quality automatic sampler.

[0041] The function features of the control system include the following: multiple triggering modes, including timing, time proportioning and manual sampling; bottle sampling function, including single bottle sampling and automatic multi-bottle sampling; mixed sampling, in which water samples collected multiple times can be placed in the same bottle; recording function, including sampling recording and power failure recording; power failure protection function, in which the instrument can automatically restore the running state after power failure and power recovery, and the instrument parameters are not lost after power failure; automatic emptying function, in which the sampling pipeline is automatically emptied and the sampling head is back-flushed after each sampling to ensure that the sampling pipeline is not blocked; automatic rinsing function, in which the pipeline is rinsed with the water sample to be tested before each sampling to ensure the representativeness of the sample; water sample preservation, normal temperature preservation; delay start function, in which the sampling is started after the delay time arrives; battery power supply, in which the sampler is powered by a lithium battery to ensure that the instrument can work normally for two days; appearance structure, integrated design, convenient to carry; sampling program, in which the user can set a commonly used sampling program and save it; protection level, IP65, 304 stainless steel shell, can be used in the rain; short-range wireless control function, in which a wireless handheld controller is used to remotely control the sampler to set, control and query the state; and a retractable support with a manual winch can be selected to facilitate the taking and placing of the sampler in the well.

[0042] The specific steps of the control method are as follows:

[0043] S1, start and state space initialization:

[0044] S1.1, construct a system state vector:

[0045] A multi-dimensional vector corresponding to the system state is constructed, and the multi-dimensional vector includes: battery voltage, ambient temperature, last task state, last sampling timestamp, and current sampling mode code, etc. Taking these multi-dimensional vectors as an example: a five-dimensional vector S=(Vb, Te, Ps, Ls, Mp); wherein Vb represents the battery voltage, unit: V, used to reflect the current power supply capacity; Te represents the ambient temperature, unit: ℃, used for temperature compensation control; Ps represents the last task state, wherein 0=not completed, 1=completed; Ls represents the last sampling timestamp, which can be selected as YYYY-MM-DD HH:MM:SS; for example: 2020-08-13 08:00:00; Mp represents the current sampling mode code; wherein 0x01=timing, 0x02=proportional to time, etc.

[0046] S1.2, model parameter loading:

[0047] A pre-constructed collaborative job model cluster is loaded from the memory matched with the system; wherein the collaborative job model cluster includes: a state recognition classification model, a sampling time prediction model, an improved resource optimization scheduling model, and a self-adaptive feedback control model; and the corresponding loading content is: the preset time period of the sampling time prediction model, for example: the period Δt=4h; the energy consumption coefficient Es=0.5Wh of the resource optimization scheduling model, which is used as a standard for single standard sampling power consumption; the initial control parameters of the self-adaptive feedback control model: proportional gain Kp, integral gain Ki, and differential gain Kd; the initial control parameter setting values can be set to 1.1, 0.05, and 0.1 in turn, which are only examples, and the initial setting can be made according to the demand in actual operation.

[0048] S2, state recognition and classification:

[0049] S2.1, running a classification rule engine:

[0050] The state recognition classification model is run according to the classification rule engine built, which is based on the following logical judgment of the current state of the system: if the last task is not completed, i.e. Ps=0, and the last sampling timestamp is earlier than the current time, it is determined that the interruption recovery mode is required to check whether the sampling point is missed; if the last task is completed, i.e. Ps=1, or the sampling has never been executed, i.e. Ls shows empty, it is determined that the standby configuration mode is required to wait for the user to issue a new task; if the current time is exactly equal to the next preset sampling time point, it is determined that the immediate sampling mode is required to immediately start the sampling process; the logic given in the classification rule engine is equivalent to a lightweight decision tree, which does not require complex training and is suitable for embedded low-power systems.

[0051] S2.2, output model category:

[0052] The output result of the state recognition classification model is used as the control entry of the next stage; for example: the state recognition classification model inputs the current multi-dimensional vector, determines that Ps=0 and Ls=2020-08-13 08:00:00, and the current time is 12:00, and determines that the interruption recovery mode is entered, and then enters the subsequent S3;

[0053] The above S2 is used to determine which running branch the system should enter by running the state recognition classification model.

[0054] S3, sampling time prediction and compensation:

[0055] S3.1, constructing a predetermined sampling time sequence:

[0056] Based on the preset time period Δt, a predetermined sampling time sequence Tn=T0+n×Δt (n=0, 1, 2,..., N-1) is generated; wherein Tn represents the nth predetermined sampling time, T0 represents the starting time of the first sampling, Δt represents the preset time period, unit: hour, n represents the sampling sequence number, starting from 0, and N represents the total sampling number;

[0057] S3.2, sliding window matching:

[0058] A time tolerance window δ is defined: δ=15min, and it is determined whether the current time falls within the tolerance interval of any unfinished predetermined sampling time Tn; wherein the tolerance interval is: Tn−δ≤current time≤Tn+δ; if it is satisfied, a compensation action is triggered immediately; otherwise, it jumps to the next Tn+1; a specific example is: when T1=12:00, the current time is 12:07, which satisfies the tolerance interval: [11:45, 12:15], so it is determined that T1 should be executed but not completed, and the second sampling (i.e. compensation T1) is triggered;

[0059] The above S3 uses a sampling time prediction model to process the time deviation caused by power failure or delay.

[0060] S4, energy consumption and task cooperative scheduling:

[0061] S4.1, remaining task energy consumption calculation:

[0062] The total power consumption E_total (i.e. remaining task energy consumption) is generated according to the remaining sampling number N_r and the single standard energy consumption E_s: E_total=N_r×E_s; in the formula, the single standard energy consumption E_s can be obtained by historical data, the power consumption required for each sampling is counted, and the average value is taken as the final standard, so that the single standard energy consumption E_s is obtained;

[0063] S4.2, safety margin determination and multi-level adjustment:

[0064] A safety margin coefficient a1 is set, a1 e [1.1, 1.3], in this embodiment, the safety margin coefficient a1 = 1.2, that is, 20% redundancy is reserved, if the current available power E_curr < a1 x E_total, the hierarchical adjustment model is triggered, the hierarchical adjustment model adopts a multi-level simulation adjustment strategy, and the corresponding adjustment level which meets the following conditions is simulated and screened:

[0065] The current available power E_curr is greater than or equal to a1 x E_total, and the corresponding adjustment level is collected to form a first adjustment set; if the first adjustment set cannot be obtained, an abort signal is sent, and it is suggested to abort the water quality sampling task;

[0066] In actual use, a task integrity weight function is introduced, the result of the function, that is, the comparison result of the task integrity evaluation and the preset standard threshold value is used as a constraint condition, the corresponding adjustment level is selected from the first adjustment set according to the constraint condition, a second adjustment set is formed, if the second adjustment set cannot be obtained, the adjustment level with the highest priority is selected from the first adjustment set to execute the target adjustment action; finally, the corresponding adjustment level of the maximum value in the task integrity evaluation is selected from the second adjustment set as the adjustment action to be actually executed next;

[0067] The multi-level simulation adjustment strategy includes:

[0068] Adjustment level L1: the sampling volume is reduced according to a preset gradient: from the original sampling volume V_set to the present sampling volume V_set x b1; wherein b1 is a first adjustment coefficient corresponding to the preset gradient, and b1 e (0, 1), in this embodiment, b1 = 0.8; for example, if the original sampling volume V_set = 500 ml, the reduced present sampling volume is: 500 ml x 0.8 = 400 ml; adjustment level L2: the number of rinsing is reduced according to a preset interval: from the original number of rinsing R_wash to the present number of rinsing R_wash-1; wherein the minimum value of the original number of rinsing R_wash is 2, and the original number of rinsing R_wash is a positive integer; in this embodiment, the original number of rinsing R_wash = 2; for example, the original number of rinsing R_wash = 2 times, and the reduced present number of rinsing is: 2-1 = 1 time; adjustment level L3: the emptying back flushing time is shortened according to a preset proportion: from the original emptying back flushing time t_blow to the present emptying back flushing time t_blow x b2; wherein b2 is a second adjustment coefficient corresponding to the preset proportion, and b2 e (0, 1), in this embodiment, b2 = 0.75; for example, if the original emptying back flushing time t_blow = 8 s, the shortened present emptying back flushing time is 8 s x 0.75 = 6 s; it should be noted that the adjustment level L1 is the adjustment level with the highest priority, and the above multi-level simulation adjustment strategy is simulated, not actually obtained;

[0069] For the convenience of understanding, the following hierarchical table is given for illustration:

[0070] Table 1: Adjustment level and comparison table of adjustment action:

[0071]

[0072] The introduced task integrity weight function is used to balance the work effect and work efficiency, and the way is:

[0073] W_iy=w1×(V_ac / V_ta)+w2×(R_ac / R_ta)+w3×(t_ac / t_ta); In the formula, W_iy represents the task integrity evaluation, and the value is in [0, 1]; w1, w2 and w3 are weight coefficients, and 1> w1> w2> w3> 0;

[0074] Among them, according to the specific scene setting, when the underground well is polluted, the original w2 is appropriately increased, for example: the original w1=0.5, w2=0.3, w3=0.2, the original w2 is appropriately increased, so that w2 changes from 0.3 to 0.4, to emphasize cleaning; When the underground well is clean, the original w2 is appropriately reduced, which is consistent with the increase, which will not be repeated here, and when it is impossible to judge whether the underground well is clean or polluted, the original w2 is maintained;

[0075] V_ac and V_ta are the actual execution sampling volume and the target sampling volume respectively, the target sampling volume is the user's preset standard sampling volume, which is the reference value in the ideal state; R_ac and R_ta are the actual execution number of washing and the target number of washing respectively, the target number of washing is the preset standard number of washing, which ensures that the pipeline is fully cleaned; t_ac and t_ta are the actual execution of emptying and back blowing time and target emptying and back blowing time respectively, the target emptying and back blowing time is the standard back blowing time, which is enough to remove the attachments of the sampling head.

[0076] Effect: by introducing the task integrity weight function, the following is realized:

[0077] Quantitative evaluation: the fuzzy influence on sampling quality is converted into a calculable value;

[0078] Intelligent decision: select the scheme with the smallest data impact in the multiple energy saving paths;

[0079] Configurability: the weight coefficient can be preset by the user or the scene, which is suitable for different monitoring needs;

[0080] Traceability: each adjustment has a complete integrity score record, which is convenient for later audit.

[0081] The comparison result of the task integrity evaluation and the preset standard threshold value is as follows:

[0082] When the task integrity estimate W_iy exceeds the preset standard threshold, indicating that it is within an acceptable range, it is retained;

[0083] When the task integrity estimate W_iy does not exceed the preset standard threshold, it is rejected;

[0084] According to the constraint condition, the corresponding adjustment level is selected from the first adjustment set, that is, the retained first adjustment set is summarized, thereby obtaining the second adjustment set;

[0085] S4.3, adjustment coefficient feedback correction:

[0086] The historical sampling feedback data set is counted and a parameter correction model is constructed to generate a correction result for the original first adjustment coefficient and the original second adjustment coefficient, which is used in the next water quality sampling work;

[0087] The historical sampling feedback data set includes: the number of effective samples and the total number of samples under the condition of using the first adjustment coefficient, the number of blockage alarms and the total number of emptying backflushes under the condition of using the second adjustment coefficient; wherein the number of effective samples indicates that the corresponding collection volume meets the standard, then the number of invalid samples is found to be insufficient for the corresponding collection volume, and the sum of the effective and invalid samples is the total number of samples; when the system performs emptying backflush, if it encounters blockage, an alarm will be triggered, which is a regular setting, so it is not introduced or explained too much, and the number of alarms triggered due to blockage is the number of blockage alarms;

[0088] The process of constructing and running the parameter correction model is as follows:

[0089] When the first adjustment coefficient is dealt with, an effectiveness function is established: η_V(b1)=N_vd(b1) / N_tl(b1); in the formula, η_V(b1) represents the sample effectiveness rate under the b1, N_vd(b1) and N_tl(b1) represent the number of samples that are effective in detection and the total number of samples respectively when the b1 is used; a target threshold is set: η_V(b1)≥ηmin needs to be met, wherein ηmin represents the target threshold; wherein the value of ηmin in the embodiment is 0.95, that is, more than 95% of the samples are effective; the correction setting is: when the set target threshold is not met in the continuous S times of tasks, it is indicated that the original b1 is too small, and needs to be adjusted upwards, and the adjustment basis is: b1_new=b1_old×(1+Δb1), Δb1∈(0, 0.1); in the embodiment, Δb1=0.05; wherein the value of S is greater than 1, and S=3 in the embodiment; in the formula, b1_new represents the new first adjustment coefficient, b1_old represents the original first adjustment coefficient, and Δb1 is the amplitude value; otherwise, it needs to be adjusted downwards, and the adjustment basis is: b1_new=b1_old×(1-Δb1), Δb1∈(0, 0.1); when the second adjustment coefficient is dealt with, a clogging risk function is established: ρ_cl(b2)=C_cl(b2) / N_bl(b2); in the formula, ρ_cl(b2) represents the clogging occurrence rate per emptying and back flushing operation, C_cl(b2) represents the number of clogging alarm times under the b2 condition, and N_bl(b2) represents the total emptying and back flushing times; a safety threshold is set: ρ_cl(b2)≤ρmax needs to be met, wherein ρmax represents the safety threshold; wherein the value of ρmax in the embodiment is 0.02, that is, at most 1 alarm per 50 emptying and back flushing operations; the correction setting is: when the set safety threshold is not met, it is indicated that the original b2 is too small, and needs to be adjusted upwards, and the adjustment basis is: b2_new=min(b2_old×(1+Δb2), 1), Δb2∈(0, 1); in the embodiment, Δb2=0.1; in the formula, b2_new represents the new second adjustment coefficient, b2_old represents the original second adjustment coefficient, and Δb2 is the adjustment amplitude; otherwise, it needs to be adjusted downwards, and the adjustment basis is: b2_new=max(b2_old×(1-Δb2), 0.6), Δb2∈(0, 1); in specific actual application, S4.3 can be triggered once every 10 sampling tasks or after 30 days of cumulative operation to realize feedback and adjustment of self-learning.

[0090] Effect description: The present application introduces safety margin judgment and multi-level adjustment strategy, including reducing sampling volume, reducing washing frequency, shortening emptying and back flushing time, etc. On the premise of ensuring sample quality, energy consumption is effectively saved; on the one hand, this method significantly improves the energy efficiency ratio of single task, prolongs the endurance time of related water quality sampling equipment in the field; on the other hand, through learning and parameter correction of historical data, the application scenarios and threshold settings of each adjustment level are further optimized, so that the whole system can continuously improve itself in long-term operation, and finally realizes the balance adjustment operation between task completion rate and energy consumption, ensuring the effectiveness of the whole control work.

[0091] S5, sampling adaptive control:

[0092] S5 is used to adopt an adaptive feedback control model to improve precision and stability during the execution phase of washing, sampling and emptying (i.e. emptying and back flushing);

[0093] S5.1, washing process flow control:

[0094] The goal is to ensure that each washing extracts an accurate volume of water sample, for example: 100mL, which is not affected by water pressure fluctuations; the defined variables include: target washing volume, actual volume extracted at time t, volume error, control output and PID controller parameter set; wherein the target washing volume is denoted as Q _ta, the actual volume extracted at time t is denoted as Q _ac(t), and the pump speed wz(t) and time integral are obtained: ; In the formula, ku represents the pump flow coefficient, with the unit of mL / rpm, and dτ represents the differential symbol of the integral variable; the volume error e(t) is: e(t)=Q _ta-Q _ac(t); the control output u(t) is used to adjust the pump driving electric; the PID controller parameter set is the initial control parameter of the adaptive feedback control model mentioned in S1.2: proportional gain Kp, integral gain Ki and differential gain Kd;

[0095] Control output u(t): ; In the formula, the part before the first plus sign is the proportional term, the larger the error, the faster the pump speed increases, and the response is accelerated; the part before the second plus sign is the integral term, which eliminates long-term cumulative error and prevents under-drawing; the part after the second plus sign is the differential term, which predicts the error trend and slows down in advance when approaching the target to prevent overshoot;

[0096] S5.2, adaptive adjustment of emptying and back flushing time:

[0097] The target is to dynamically adjust the emptying backflushing time according to the ambient temperature, to prevent sludge sticking at high temperature or energy waste at low temperature; the defined variables include: the final executed emptying backflushing time t_ac', the actually executed emptying backflushing time t_ac, the current ambient temperature T_e, the reference ambient temperature T_ref and the temperature compensation coefficient d1, which represents the proportion of the required increase in emptying backflushing time per 1℃ increase; the mode for self-adaptive adjustment of the emptying backflushing time is:

[0098] t_ac'=t_ac×[1+d1×(T_e-T_ref)]; in the formula, when T_e>T_ref: the sludge viscosity is reduced, but it is easy to deposit, and the purging time needs to be extended on the original basis; when T_e

[0099] Effect description: through the precise time sequence sliding window algorithm and the closed-loop adjustment mechanism based on the extended PID, the sampling time and process variables such as flow and temperature compensation can be accurately controlled, thereby greatly improving the accuracy of data acquisition; on the one hand, this helps to eliminate errors caused by external interference factors, ensuring the consistency and repeatability of data; on the other hand, intelligent decision-making combined with the task integrity weight function can minimize the influence of various energy-saving measures on the final analysis results, and ensure the authenticity and reliability of the monitoring data.

[0100] S6, multi-model collaborative decision-making operation:

[0101] The collaborative work model cluster cooperates through event-driven and data-sharing methods;

[0102] S6.1, model data flow definition:

[0103] Between the state recognition classification model and the sampling time prediction model: output the interrupt recovery signal; between the sampling time prediction model and the improved resource optimization scheduling model: provide the remaining sampling number N_r; between the improved resource optimization scheduling model and the adaptive feedback control model: transfer the adjusted sampling volume, the number of rinse times and the preliminary emptying backflushing time; between the adaptive feedback control model and the state recognition classification model: feedback the execution success / failure state;

[0104] S6.2, collaborative control cycle:

[0105] The period of each model in the collaborative operation model cluster can be set according to actual needs: for example, taking 1s as a basic control period, each model is executed in turn: the adaptive feedback control model performs PID calculation once every 10ms; the improved resource optimization scheduling model updates the energy consumption state every 1s; the state recognition classification model and the sampling time prediction model are updated every min; after each sampling, the log can be updated, and online learning can also be performed as needed.

[0106] S7, operation log recording and online learning can also be included in the actual running process; at this time, the system has preliminary learning ability, optimizes the model parameters through historical data, realizes the structured storage of the log, and can complete backtracking and optimization of the parameters of part of the model, for example: after completing 3 tasks, the following operations are regularly performed: the average error of the adaptive feedback control model is counted, and the PID parameters are fine-tuned; the power-off frequency is analyzed, and the time tolerance window delta of the sampling time prediction model is optimized; the battery attenuation model is established, and the single standard energy consumption E_s of the improved resource optimization scheduling model is updated; the specific content of backtracking and optimization is not described here, because it is a conventional technical means.

[0107] Dynamic adaptation is realized through multiple models, solving the problem that fixed logic cannot cope with dynamic environment; on the one hand, the application realizes intelligentization and adaptive operation of water quality sampling control, and through the linkage of each model in the collaborative operation model cluster, the water quality sampling work can independently adjust the operation strategy under complex working conditions such as power-off, low power, water pressure fluctuation, etc., ensuring task integrity and data representativeness; on the other hand, the application realizes learning behavior and long-term performance optimization, and through running log accumulation and model parameter backtracking, it realizes continuous evolvable ability; experimental data shows that after 90 days of continuous operation, the task completion rate is increased from 90% to about 95%, the average energy consumption can be reduced by about 11%, and the pipe blockage rate is significantly reduced by about 80%.

[0108] Embodiment 2

[0109] Based on embodiment 1, the application also provides a water quality sampling control system for environmental detection, comprising:

[0110] An initialization module: a system state vector is constructed, and a collaborative operation model cluster is loaded; wherein the collaborative operation model cluster at least includes: a state recognition classification model, a sampling time prediction model, an improved resource optimization scheduling model and an adaptive feedback control model;

[0111] A state recognition module: according to the system state vector, the state recognition classification model runs a classification rule engine, and outputs a mode determination result;

[0112] The sampling prediction module runs a sampling time prediction model, constructs a predetermined sampling time sequence, and completes a sliding window matching action.

[0113] The energy consumption and task coordination module runs an improved resource optimization scheduling model, compares the product of the residual task energy consumption and the safety margin coefficient with the current available power, selects whether to trigger the hierarchical adjustment model according to the comparison result, and the hierarchical adjustment model adopts a multi-level simulation adjustment strategy, simulates according to different adjustment levels, filters the corresponding adjustment level that meets the condition that the current available power is greater than or equal to the product of the residual task energy consumption and the safety margin coefficient, and forms a first adjustment set; the task integrity weight function is introduced to obtain the task integrity evaluation, and the comparison result between the task integrity evaluation and the preset standard threshold is used as a constraint condition, and the corresponding adjustment level with the maximum task integrity evaluation is selected from the first adjustment set according to the constraint condition to form a second adjustment set.

[0114] The feedback correction sub-module: statistics historical sampling feedback data set, and construct parameter correction model, realize the correction of multi-level simulation adjustment strategy; it needs to be explained that the feedback correction sub-module is executed or run as a sub-module of the energy consumption and task coordination module; the adaptive control module: adopts an adaptive feedback control model to realize the flow control and emptying backflush time adaptive adjustment of the rinsing process.

[0115] The above embodiments can be realized by software, hardware, firmware or any combination thereof, in whole or in part. When realized by software, the above embodiments can be realized in the form of a computer program product in whole or in part. Those skilled in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized by hardware or software depends on the specific application and design constraints of the technical solutions.

[0116] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, which can be located in one place or distributed on multiple network units. Part or all of the units can be selected to achieve the purpose of the embodiment according to actual needs.

[0117] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any skilled person in the art can easily think of changes or replacements within the technical range disclosed in the present application, which should be covered within the protection scope of the present application.

Claims

1. A water quality sampling control method for coping with environmental detection, characterized by, The steps include: constructing a system state vector, using a collaborative job model cluster and loading; wherein the collaborative job model cluster at least includes: a state recognition classification model, a sampling time prediction model, an improved resource optimization scheduling model, and a self-adaptive feedback control model; According to the system state vector, the state recognition classification model runs the classification rule engine, and outputs the mode judgment result; Run the sampling time prediction model, construct the scheduled sampling time sequence, and complete the sliding window matching action; Run the improved resource optimization scheduling model, compare the product of the remaining task energy consumption and the safety margin coefficient with the current available power, and select whether to trigger the hierarchical adjustment model according to the comparison result, and the hierarchical adjustment model uses a multi-level simulation adjustment strategy, simulates according to different adjustment levels, selects the corresponding adjustment level that satisfies: the current available power ≥ the product of the remaining task energy consumption and the safety margin coefficient, and forms a first adjustment set; Introducing the task integrity estimate value obtained by the task integrity weight function, taking the comparison result between the preset standard threshold value as the constraint condition, and according to the constraint condition, the corresponding adjustment level of the maximum task integrity estimate value is selected from the second adjustment set; The multi-level simulation adjustment strategy includes: adjustment level L1: reduce the sampling volume according to the preset gradient: from the original sampling volume V_set to the present sampling volume V_set×b1; Wherein, b1 is the first adjustment coefficient corresponding to the preset gradient, and b1∈(0, 1); Adjustment level L2: reduce the number of rinsing according to the preset interval: from the original rinsing number R_wash to the present rinsing number R_wash-1; Wherein, the minimum value of the original rinsing number R_wash is 2, and the original rinsing number R_wash is a positive integer; Adjustment level L3: shorten the blowdown blowback time according to the preset proportion: from the original blowdown blowback time t_blow to the present blowdown blowback time t_blow×b2; Wherein, b2 is the second adjustment coefficient corresponding to the preset proportion, and b2∈(0, 1); Statistical historical sampling feedback data set, and construct parameter correction model to realize the correction of multi-level simulation adjustment strategy; Adopting the self-adaptive feedback control model, realizing the flow control and the adaptive adjustment of the blowdown blowback time in the rinsing process.

2. The water quality sampling control method for environmental detection according to claim 1, characterized in that: Construct a system state vector, which includes a multi-dimensional vector at least including: battery voltage, environmental temperature, last task state, last sampling timestamp, and current sampling mode code; In the last task state: 0=not completed, 1=completed; The contents loaded after using the collaborative job model cluster at least include: the preset time period of the sampling time prediction model; The energy consumption coefficient of the improved resource optimization scheduling model, which is used as the standard for single standard sampling power consumption; The initial control parameters of the self-adaptive feedback control model: proportional gain Kp, integral gain Ki and differential gain Kd.

3. The water quality sampling control method for environmental detection according to claim 1, characterized in that: The process of running the classification rule engine is: if the last task is not completed and the last sampling time stamp is earlier than the current time, it is determined that the interruption recovery mode is entered; if the last task is completed or sampling has never been performed, it is determined that the standby configuration mode is entered; if the current time is equal to the next preset sampling time point, it is determined that the immediate sampling mode is entered.

4. The water quality sampling control method for environmental detection according to claim 2, characterized in that: When the predetermined sampling time sequence is constructed, the predetermined sampling time sequence is generated based on the preset time period Δt; the execution process of the sliding window matching action is: a time tolerance window δ is defined, and it is checked whether the current time falls within the tolerance interval of any unfinished predetermined sampling time Tn; wherein the tolerance interval is: Tn−δ≤current time≤Tn+δ; if yes, the supplementary sampling action is triggered immediately; otherwise, it is skipped to the next Tn+1.

5. The water quality sampling control method for environmental detection according to claim 1, characterized in that: The manner of introducing the task integrity weight function is: W_iy=w1×(V_ac / V_ta)+w2×(R_ac / R_ta)+w3×(t_ac / t_ta); wherein, W_iy is the task integrity evaluation value, and the value is in [0, 1]; w1, w2 and w3 are weight coefficients, and 1>w1>w2>w3>0; V_ac and V_ta are the actual executed sampling volume and the target sampling volume respectively, R_ac and R_ta are the actual executed rinsing frequency and the target rinsing frequency respectively, t_ac and t_ta are the actual executed emptying back flushing time length and the target emptying back flushing time length respectively; the comparison result of the task integrity evaluation value W_iy and the preset standard threshold value is as follows: when the task integrity evaluation value W_iy exceeds the preset standard threshold value, it is retained; otherwise, it is rejected.

6. The water quality sampling control method for environmental detection according to claim 1, characterized in that: The historical sampling feedback data set at least includes: the number of effective samples and the total number of samples under the condition of using the first adjustment coefficient, the number of blockage alarms and the total number of emptying back flushing under the condition of using the second adjustment coefficient; The process of constructing and running the parameter correction model is: In response to the first adjustment coefficient, an effectiveness function η_V(b1)=N_vd(b1) / N_tl(b1) is established; wherein, η_V(b1) is the sample effectiveness rate under b1, N_vd(b1) and N_tl(b1) are the number of effective samples and the total number of samples detected under b1 respectively; a target threshold is set: η_V(b1)≥ηmin must be met; a correction setting is made: when the set target threshold is not met for S consecutive tasks, an upward action is performed; otherwise, a downward action is performed; In response to the second adjustment coefficient, a blockage risk function ρ_cl(b2)=C_cl(b2) / N_bl(b2) is established; wherein, ρ_cl(b2) is the blockage occurrence rate per unit emptying back flushing operation, C_cl(b2) is the number of blockage alarms under b2, and N_bl(b2) is the total number of emptying back flushing; a safety threshold is set: ρ_cl(b2)≤ρmax must be met; a correction setting is made: when the set safety threshold is not met, an upward action is performed; otherwise, a downward action is performed.

7. The water quality sampling control method for environmental detection according to claim 2, characterized in that: When the rinsing process flow control is implemented: the target is to ensure that a set volume of water sample is extracted each time. The defined variables at least include: a target rinsing volume, an actual volume extracted at time t, a volume error, a control output, and a PID controller parameter set; wherein the target rinsing volume is denoted as Q _ta, the actual volume extracted at time t is denoted as Q _ac(t), the volume error e(t) is obtained by integrating the pump speed wz(t) and time, the control output u(t) is used to adjust the pump driving electricity, and the PID controller parameter set is the initial control parameter of the adaptive feedback control model; and the control output u(t) is obtained according to the PID controller parameter set.

8. The water quality sampling control method for environmental detection according to claim 5, characterized in that: When the emptying backflush time is adaptively adjusted, the target is to dynamically adjust the emptying backflush time according to the ambient temperature; The defined variables at least include: a final executed emptying backflush time t_ac', an actually executed emptying backflush time t_ac, a current ambient temperature T_e, a reference ambient temperature T_ref, and a temperature compensation coefficient d1; and the way according to which the emptying backflush time is adaptively adjusted is: t_ac'=t_ac×[1+d1×(T_e-T_ref)]; wherein the temperature compensation coefficient d1 has a value range of (0, 0.1).

9. A water quality sampling control system for environmental detection, characterized by: The initialization module includes: The initialization module: constructs a system state vector, adopts a cooperative job model cluster, and loads; wherein the cooperative job model cluster at least includes: a state recognition classification model, a sampling time prediction model, an improved resource optimization scheduling model, and an adaptive feedback control model; The state recognition module: according to the system state vector, the state recognition classification model runs a classification rule engine, and outputs a mode judgment result; The sampling prediction module: runs the sampling time prediction model, constructs a predetermined sampling time sequence, and completes a sliding window matching action; The energy consumption and task coordination module: running an improved resource optimization scheduling model, comparing the product of the remaining task energy consumption and the safety margin coefficient with the current available power, selecting whether to trigger the hierarchical adjustment model according to the comparison result, and the hierarchical adjustment model adopts a multi-level simulation adjustment strategy, simulates according to different adjustment levels, filters the corresponding adjustment level that meets: the current available power ≥ the product of the remaining task energy consumption and the safety margin coefficient, and forms a first adjustment set; introducing the task integrity estimate value obtained by the task integrity weight function, taking the comparison result between the task integrity estimate value and the preset standard threshold as a constraint condition, and according to the constraint condition, the corresponding adjustment level of the second adjustment set is selected from the first adjustment set, and the adjustment level corresponding to the maximum task integrity estimate value is selected from the second adjustment set; the multi-level simulation adjustment strategy includes: adjustment level L1: reducing the sampling volume according to the preset gradient: from the original sampling volume V_set to the present sampling volume V_set×b1; wherein, b1 is the first adjustment coefficient corresponding to the preset gradient, and b1∈(0, 1); adjustment level L2: reducing the number of rinsing according to the preset interval: from the original rinsing number R_wash to the present rinsing number R_wash-1; wherein, the minimum value of the original rinsing number R_wash is 2, and the original rinsing number R_wash is a positive integer; adjustment level L3: shortening the blowdown blowback time according to the preset proportion: from the original blowdown blowback time t_blow to the present blowdown blowback time t_blow×b2; wherein, b2 is the second adjustment coefficient corresponding to the preset proportion, and b2∈(0, 1); The feedback correction sub-module: statistics historical sampling feedback data set, and construct parameter correction model, realize the correction of multi-level simulation adjustment strategy; The self-adaptive control module: using an adaptive feedback control model to realize the flow control and the adaptive adjustment of the blowdown blowback time.

Images