An adaptive control method and system for a photovoltaic driven photocatalytic reactor

Abstract

The present application relates to sewage treatment and intelligent control field, specifically disclose a kind of adaptive control method and system of photovoltaic driving photocatalytic reactor, including real-time acquisition photovoltaic power supply state and effluent quality parameter, whether trigger adjustment is judged according to the weighted comprehensive score of COD and ammonia nitrogen removal rate;Pre-constructed response surface model through offline test, a small amount of data is collected after triggering and is updated online, construct the response surface model with illumination intensity, hydraulic retention time and catalyst surface update frequency as independent variable, with predicted comprehensive score as dependent variable, solve optimal parameter combination and execute;Through closed-loop feedback continuous monitoring, until effluent standard is reached.The present application deeply fuses photovoltaic clean energy and artificial intelligence adaptive optimization, realizes the independent dynamic matching of operating parameter under power fluctuation and water quality disturbance, significantly improves the effluent stability and environmental adaptability of reactor.

Description

Technical Field

[0001] This invention belongs to the field of wastewater treatment and intelligent control, specifically an adaptive control method and system for a photovoltaic-driven photocatalytic reactor. Background Technology

[0002] A photovoltaic-driven photocatalytic reactor is a wastewater treatment device that combines solar photovoltaic power generation with photocatalytic oxidation technology. It is widely used for on-site purification of rural domestic sewage, small water bodies, and decentralized wastewater in remote areas. Its basic principle is as follows: photovoltaic modules absorb sunlight and convert it into electrical energy, which powers the light source (such as ultraviolet lamps or LED lamps) and auxiliary equipment within the reactor. This excites photocatalytic materials (such as TiO2, g-C3N4, etc.) to produce highly oxidizing active species (such as hydroxyl radicals ·OH), thereby oxidizing and degrading organic pollutants and harmful substances such as ammonia nitrogen in the wastewater into CO2, H2O, and small-molecule inorganic substances. This technology has advantages such as no need for external power grid connection, low operating costs, and no secondary pollution, making it particularly suitable for rural areas with weak power infrastructure.

[0003] In actual operation, photovoltaic-driven photocatalytic reactors face two typical dynamic disturbances: First, photovoltaic power supply is affected by weather conditions (cloud cover, rainy days) and changes in the angle of sunlight, resulting in significant intermittency and fluctuations in output power, making it impossible to maintain a constant light intensity within the reactor. Second, the influent water quality (COD concentration, ammonia nitrogen concentration, etc.) exhibits large random fluctuations due to changes in residents' daily routines and seasonal variations. Traditional control methods typically employ fixed parameter modes (such as fixed hydraulic residence time, fixed light source power) or simple on / off control (such as shutting down the equipment when the light intensity is below a certain threshold), which are insufficient to simultaneously address these dual dynamic disturbances. When photovoltaic power supply is insufficient or the influent load suddenly increases, fixed parameter control cannot adjust key operating parameters such as hydraulic residence time and catalyst surface renewal frequency in a timely manner, leading to excessive COD and ammonia nitrogen concentrations in the effluent and unstable purification effects. Currently, there is a lack of dynamic adaptive control schemes that can simultaneously adapt to both photovoltaic fluctuations and water quality changes. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides an adaptive control method and system for a photovoltaic-driven photocatalytic reactor, thereby resolving the issue of unstable effluent quality caused by the inability to simultaneously adapt to fluctuations in photovoltaic power supply and dynamic changes in influent water quality in the prior art.

[0005] This invention provides an adaptive control method for a photovoltaic-driven photocatalytic reactor, comprising the following steps:

[0006] Step S1: Real-time acquisition of the operating status parameter set and effluent water quality parameter set of the photocatalytic reactor; the operating status parameter set includes the output voltage, output current and light intensity of the photovoltaic power supply module; the effluent water quality parameter set includes the chemical oxygen demand (COD) concentration and ammonia nitrogen concentration;

[0007] Step S2: Calculate the comprehensive score of the current effluent water quality according to the preset effluent water quality evaluation function. The evaluation function is a weighted sum of COD removal rate and ammonia nitrogen removal rate, wherein the COD removal rate and ammonia nitrogen removal rate are calculated from the current COD concentration, ammonia nitrogen concentration and the influent COD concentration, ammonia nitrogen concentration, respectively. If the comprehensive score is higher than the preset first threshold, the current control parameters are maintained unchanged and the process returns to step S1. If the comprehensive score is lower than or equal to the first threshold, the adaptive adjustment process is triggered.

[0008] Step S3: In advance, at least m sets of different control parameter combinations and their corresponding effluent water quality data are collected through offline experiments, where m is greater than the number of undetermined coefficients in the response surface model, to construct an initial response surface model; During actual system operation, when the adaptive adjustment process is triggered, at least one set of actual effluent water quality data corresponding to the control parameters under the current operating conditions is collected, and the initial response surface model is incrementally updated online using the recursive least squares method; The control parameters include the light intensity setpoint, hydraulic residence time setpoint, and catalyst surface update frequency setpoint in the photocatalytic reaction chamber;

[0009] Step S4: Based on the response surface model, solve for the candidate control parameter combination that yields the highest predicted comprehensive score of effluent water quality within the feasible region of the current control parameters, and send the candidate control parameter combination to the actuator of the photocatalytic reactor to update the current control parameters;

[0010] Step S5: Under the condition of executing the updated control parameters, continuously monitor the actual comprehensive score of the effluent water quality. If the actual comprehensive score recovers to above the first threshold after a preset stabilization period, lock the current control parameters and return to step S1; if the actual comprehensive score still does not recover to above the first threshold, return to step S3 for a new round of adjustment until the standard is met or the preset maximum number of adjustment rounds is reached.

[0011] Preferably, the weights of the preset effluent water quality evaluation function in step S2 are fixed values, wherein the weight of COD removal rate is 0.6 and the weight of ammonia nitrogen removal rate is 0.4.

[0012] Preferably, step S2, before calculating the comprehensive score, further includes a state grading strategy:

[0013] Based on the output voltage of the photovoltaic power supply module and the preset voltage level threshold, the current photovoltaic state is divided into three levels: "sufficient state", "fluctuating state" and "insufficient state".

[0014] The value of the first threshold is dynamically adjusted according to different levels: the original first threshold is used in the "sufficient state"; the first threshold is lowered to the second threshold in the "fluctuating state", and the second threshold is lower than the first threshold; the first threshold is lowered to the third threshold in the "insufficient state", and the third threshold is lower than the second threshold.

[0015] Preferably, before solving for the candidate control parameter combinations in step S4, a multi-constraint collaborative boundary verification step is also included:

[0016] Obtain the physical constraint boundaries of each control parameter, including the upper and lower limits of light intensity, the minimum and maximum values ​​of hydraulic residence time, and the upper limit of catalyst surface renewal frequency;

[0017] Identify and construct the correlation constraints between various control parameters, including: the product of hydraulic residence time and catalyst surface renewal frequency is not less than a preset minimum total reaction contact threshold, and the product of light intensity setting value and hydraulic residence time setting value is not less than a preset minimum total light energy input threshold.

[0018] The physical constraint boundaries and associated constraint relationships are jointly constructed into a constraint boundary management system;

[0019] When solving for candidate combinations of control parameters, only candidate solutions that simultaneously satisfy all physical constraint boundaries and associated constraint relationships are retained; if the current optimal candidate solution violates any associated constraint, it is shrunk to the inside of the boundary of the associated constraint before being adopted.

[0020] Preferably, when returning to step S3 in step S5 for a new round of adjustments, a historical trend awareness and early switching strategy is introduced:

[0021] Record the improvement of the actual comprehensive score of the effluent water quality after each round of adjustment during the first N rounds of adjustment, N≥2;

[0022] Calculate the rate of change of improvement between two adjacent rounds of adjustment. If the rate of change is negative and the absolute value exceeds a preset threshold for diminishing returns, then it is determined that the marginal return of the current adjustment direction has significantly decreased.

[0023] After determining that the marginal returns have decreased significantly, the adjustment strategy is actively switched: the optimization interval of the response surface model is switched from the local neighborhood of the current control parameters to the global feasible region, or the polynomial order of the response surface model is increased to capture nonlinear features.

[0024] Preferably, after locking the current control parameters in step S5, the method further includes an abnormal root cause tracing and targeted intervention step:

[0025] When the actual comprehensive score of the effluent water quality fluctuates around the first threshold for more than three consecutive adjustment cycles;

[0026] Check the following candidate root causes in sequence: photovoltaic power supply fluctuation exceeds the preset fluctuation limit, influent COD concentration fluctuation exceeds the preset load change threshold, and photocatalytic material catalytic efficiency decays to below the preset efficiency lower limit.

[0027] The candidate root cause that contributes the most to the current fluctuation in effluent water quality during the inspection is identified as the abnormal root cause.

[0028] Based on the type of abnormal root cause, targeted intervention actions are performed: if the root cause is fluctuations in photovoltaic power supply, the auxiliary energy storage unit is activated to compensate for power supply; if the root cause is changes in influent load, the residence time of the homogenization and conditioning tank at the front end is adjusted; if the root cause is the degradation of catalyst material, an in-situ catalyst regeneration or replacement prompt is triggered.

[0029] Preferably, the weights of the effluent water quality evaluation function in step S2 are not fixed values, but are dynamically calibrated.

[0030] Multiple sub-dimensions for evaluating effluent water quality were set, including COD removal rate, ammonia nitrogen removal rate, and effluent turbidity.

[0031] In each sampling period, the normalized achievement gap between the actual value of each sub-dimension and its corresponding achievement target value is calculated;

[0032] The evaluation weight for the current period is dynamically calculated based on the normalized achievement gap of each sub-dimension: the sub-dimension with the larger achievement gap is assigned a higher evaluation weight, and the sub-dimension with the smaller achievement gap is assigned a lower evaluation weight.

[0033] The weights are determined as follows:

[0034] like ,but ;

[0035] like ,but ;

[0036] The overall score is the weighted sum of the actual values ​​of each sub-dimension and their current periodic dynamic evaluation weights.

[0037] The present invention also provides an adaptive control system for a photovoltaic-driven photocatalytic reactor, for performing the above-described method, comprising:

[0038] Data acquisition module: It is connected to the water quality sensor in the photocatalytic reactor and the light and electrical parameter sensors of the photovoltaic power supply module, respectively, to acquire the set of operating status parameters and the set of effluent water quality parameters in real time;

[0039] Status assessment module: connected to the data acquisition module, with a pre-set effluent water quality evaluation function and a first threshold, used to calculate the comprehensive score of the current effluent water quality and compare it with the first threshold to trigger the adaptive adjustment process;

[0040] Response surface modeling module: connected to the state assessment module, used to store the initial response surface model constructed in advance by collecting at least m sets of different control parameter combinations and their corresponding effluent water quality data through offline experiments; after triggering the adaptive adjustment process, it collects at least one set of actual effluent water quality data corresponding to the control parameters under the current operating conditions, and uses the recursive least squares method to perform online incremental updates on the initial response surface model;

[0041] The optimization solution module is connected to the response surface modeling module and is used to solve for the candidate control parameter combination that maximizes the predicted comprehensive score of the effluent water quality within the feasible region of the control parameters. The optimization solution module integrates a constraint boundary management unit for performing multi-constraint collaborative boundary checks.

[0042] Execution control module: connected to the optimization solution module and the actuator of the photocatalytic reactor, used to send candidate control parameter combinations to the actuator and update the current control parameters;

[0043] Feedback monitoring module: Connected to the data acquisition module and execution control module, it is used to continuously monitor the actual comprehensive score of the effluent water quality after executing the updated control parameters, and decide to lock the current parameters or trigger a new round of adjustments based on the compliance status.

[0044] Preferably, the constraint boundary management unit within the optimization solution module includes:

[0045] The physical constraint subunit is used to store the physical constraint boundaries of each control parameter;

[0046] The associated constraint subunit is used to store and identify the preset associated constraint relationships between various control parameters;

[0047] The verification execution subunit is used to verify whether the candidate solution satisfies both physical and associated constraints when solving for candidate control parameter combinations, and to perform boundary shrinkage correction on candidate solutions that violate the constraints.

[0048] Preferably, the feedback monitoring module further includes a trend analysis unit and a strategy switching unit:

[0049] The trend analysis unit is used to record the improvement rate of the actual comprehensive score of the effluent water quality during the first N rounds of adjustment, and to calculate the rate of change of the improvement rate.

[0050] The strategy switching unit is connected to the trend analysis unit and is used to actively send a strategy switching signal to the optimization solution module when the rate of change shows a significant decrease in marginal returns, instructing the optimization solution module to switch the optimization interval from the current local neighborhood to the global feasible region.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] By introducing an adaptive parameter adjustment mechanism based on the response surface model, the optimization process is triggered in real time according to the comprehensive score of the effluent water quality, enabling the control parameters to automatically seek optimization as the environment changes. This effectively overcomes the shortcomings of traditional fixed parameter control methods in dynamic scenarios, which have poor adaptability.

[0053] This invention employs a multi-dimensional state-level strategy to dynamically adjust the target threshold for effluent water quality based on photovoltaic power supply capacity. This allows the system to pursue high-standard purification effects when there is sufficient sunlight and automatically switch to an energy-saving mode when there is insufficient sunlight, thus achieving an adaptive balance between treatment quality and energy consumption costs.

[0054] This invention integrates physical constraints and parameter coupling constraints into the optimization process through multi-constraint collaborative boundary management, ensuring that every control adjustment is executed within the engineering feasible domain and avoiding execution failure or equipment damage caused by parameter conflicts.

[0055] This invention also improves adjustment efficiency, fault handling capabilities, and multi-objective collaborative optimization level by using historical trend perception to switch optimization directions in advance, tracing the root causes of anomalies to achieve precise targeted intervention, and dynamically evaluating weights to focus on weak indicators.

[0056] In summary, this invention endows photovoltaic-driven photocatalytic reactors with the ability to sense environmental changes, make autonomous decisions and adjustments, and continuously optimize operation, significantly improving the equipment adaptability, operational reliability and economy in decentralized wastewater treatment scenarios, and providing green and intelligent technical support for rural revitalization and ecological environmental protection. Attached Figure Description

[0057] Figure 1 This is a schematic diagram of the adaptive control method in Embodiment 1 of the present invention. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] Example 1: This example provides an adaptive control method for a photovoltaic-driven photocatalytic reactor, which is described below in conjunction with... Figure 1 The flowchart shown will be explained in detail.

[0060] Step S1: Real-time acquisition of operating status parameter set and effluent water quality parameter set

[0061] The following sensors were installed on the photocatalytic reactor:

[0062] Photovoltaic power supply module: Equipped with voltage sensor, current sensor and light intensity sensor (model: BH1750) to collect output voltage data in real time. (Unit: V) Output current (Unit: A) and ambient light intensity (Unit: Lux).

[0063] Water quality monitoring module: A COD sensor (such as an ultraviolet absorption COD sensor), an ammonia nitrogen sensor (ion-selective electrode method), and a turbidity sensor are installed at the inlet and outlet respectively. The influent COD concentration is collected in real time. (Unit: mg / L), influent ammonia nitrogen concentration (Unit: mg / L), Effluent COD concentration Ammonia nitrogen concentration in effluent and effluent turbidity (Unit: NTU)

[0064] The sampling period of the above sensors is set according to the sensor type: the sampling period of light intensity sensor and voltage and current sensor is 1 minute; the sampling period of COD sensor and ammonia nitrogen sensor is 5 minutes (to meet the sensor stabilization time requirements, the sampling value is the average value within the stabilization period); the sampling period of turbidity sensor is 2 minutes. The collected data is transmitted to the embedded controller (such as STM32F407 or Raspberry Pi) through a data acquisition card (such as NI USB-6009).

[0065] Step S2: Calculate the overall water quality score and determine whether to trigger adjustments.

[0066] First, we define two basic removal rates for the effluent quality:

[0067] COD removal rate:

[0068]

[0069] when At that time, take .

[0070] Ammonia nitrogen removal rate:

[0071]

[0072] The same applies to the case of zero input.

[0073] This embodiment uses a fixed-weight effluent water quality evaluation function for comprehensive scoring. Defined as:

[0074]

[0075] The weights satisfy In this embodiment, , This means that the COD removal rate has a slightly higher weight, which is in line with the actual needs of domestic sewage treatment, where COD is the main indicator.

[0076] Preset first threshold If the current overall score This indicates that the effluent water quality meets the standards. The system maintains the current control parameters unchanged and returns to step S1 to continue monitoring. If If this occurs, the adaptive adjustment process is triggered, and step S3 is entered.

[0077] Step S3: Collect data and build a response surface model

[0078] Step S3: Build and update the response surface model online.

[0079] This embodiment uses a combination of offline pre-modeling and online incremental updates.

[0080] (a) Offline pre-modeling: During the system debugging phase, three adjustable control parameters are defined:

[0081] Light intensity setting value Unit is This is achieved by adjusting the photovoltaic output duty cycle or the power of the auxiliary LED lights.

[0082] Hydraulic residence time set value Unit is This is achieved by adjusting the speed of the inlet peristaltic pump.

[0083] Catalyst surface refresh frequency setting Unit is This is achieved by adjusting the speed of the stirrer inside the reaction chamber.

[0084] The central composite design (CCD) is used to generate 12 sets of control parameter combinations. (The number of undetermined coefficients is greater than 10 in the second-order response surface model). Each set of parameters runs for 30 minutes, and the average comprehensive score is collected after the effluent water quality stabilizes. The response surface model adopts a second-order polynomial form:

[0085]

[0086] The initial model parameters are obtained by fitting using the least squares method:

[0087]

[0088] in To design the matrix, This is the actual comprehensive score vector.

[0089] (II) Online Incremental Update: After the system enters actual operation, each time the adaptive adjustment process is triggered, only one set of actual effluent water quality comprehensive score data under the current control parameters is collected. The response surface model parameters are updated using the recursive least squares (RLS) method.

[0090]

[0091] in For the gain vector, This is the current control parameter vector. This method provides an actual comprehensive score. It does not require interrupting the normal processing flow, thus avoiding substandard emissions caused by prolonged data collection.

[0092] Step S4: Solve for the optimal combination of control parameters

[0093] The response surface model established in step S3 Solving within the feasible region of the current control parameters will result in a predicted comprehensive score. The largest possible combination of parameters. The feasible region is defined by engineering constraints:

[0094]

[0095] The value used in this embodiment is: , h, .

[0096] The solution is obtained using the Sequential Quadratic Programming (SQP) algorithm, with the objective function being: To obtain the optimal solution Then, the control module issues commands to adjust the duty cycle of the photovoltaic MPPT controller or the LED drive current to achieve the desired effect. Adjust the frequency of the inlet peristaltic pump to achieve Adjusting the PWM duty cycle of the stirring motor to achieve... .

[0097] Step S5: Execution Feedback and Monitoring

[0098] After implementing the updated control parameters, the system continuously monitors the overall score of the actual effluent water quality. Set the settling time. Hours (i.e., 3 sampling periods). If in back If the adjustment is successful, the current control parameters are locked, and the system returns to step S1 to enter normal monitoring mode. If the target is still not met, the system returns to step S3 for a new round of adjustment. To avoid infinite loops, a maximum number of adjustment rounds is set. If the target is not met even after reaching the maximum number of cycles, the system will issue a maintenance alarm signal, prompting manual intervention.

[0099] Example 2: This example adds a state hierarchical strategy based on Example 1.

[0100] S2.1, Photovoltaic state classification:

[0101] Based on the output voltage of the photovoltaic power supply module With preset voltage level threshold and (This embodiment takes) , The photovoltaic state is divided into:

[0102] Sufficient condition:

[0103] Fluctuation state:

[0104] Insufficient status:

[0105] S2.2 Dynamically adjust the discrimination threshold:

[0106] Original first threshold Use only when power supply is sufficient. Under fluctuating conditions, due to reduced power capacity, maintaining a high standard cannot be achieved; therefore, the threshold is lowered to the second threshold. If the threshold is insufficient, it will be further lowered to the third threshold. In this way, when photovoltaic power is insufficient, the system automatically lowers the target water quality for the effluent, avoiding frequent and ineffective adjustments; when photovoltaic power recovers, the threshold rises again to ensure the final effluent quality.

[0107] Real-time monitoring of status hierarchy unit Update in each sampling period The actual value is then passed to the state evaluation module.

[0108] The second threshold (75%) and the third threshold (60%) mentioned above are exemplary values ​​of the present invention. In actual applications, they can be adjusted according to local wastewater discharge standards (such as GB 18918) and reactor design indicators. The third threshold should not be lower than the minimum removal rate corresponding to the statutory discharge standard. When the system is in an "insufficient state" for a long time and the actual effluent quality is close to the discharge limit, the system should issue a maintenance alarm to prompt manual intervention or switch to backup power. The thresholds in the present invention are only used for internal control decisions and do not represent that the final discharge is allowed to exceed the standard.

[0109] Example 3: This example adds a constraint boundary check step before solving for the optimal parameters in step S4.

[0110] S3.1 Physical Constraints:

[0111] Each control parameter has its own physical limitations, for example:

[0112] The light intensity must not exceed 500 (To prevent overheating and damage to the photocatalytic material), i.e. .

[0113] The hydraulic retention time should not be less than 0.5 h (to prevent insufficient reaction) and not more than 8 h (to prevent overflow or sludge deposition). .

[0114] The stirring speed should not exceed 200 rpm (to prevent excessive catalyst wear). .

[0115] S3.2, Association Constraints

[0116] There is a coupling relationship between the two parameters, such as the hydraulic residence time. With stirring speed The product must be greater than or equal to a minimum total contact amount. This ensures that the wastewater and the photocatalyst are in full contact.

[0117] In this embodiment, the hydraulic residence time Catalyst surface renewal frequency (stirring speed) Product constraints Based on a simplified mass transfer kinetics model, the effective number of contacts between pollutants and the catalyst surface in a photocatalytic reactor is determined. Proportional to and The product of, i.e. To achieve the target degradation rate, the following conditions must be met. From this, we can derive Based on the experimental data of the mass transfer characteristics of this reactor, when At that time, the COD removal rate decreased by more than 15% compared to the baseline condition, therefore... .

[0118] This embodiment takes ,Right now:

[0119]

[0120] like For smaller residence times, the stirring speed must be increased to enhance mass transfer; conversely, the same applies.

[0121] In addition, light intensity Duration of stay There is an energy consumption correlation: when the light intensity is high, the dwell time can be appropriately shortened, but the total light energy input must not be lower than the lower limit. :

[0122]

[0123] This embodiment takes .

[0124] S3.3 Inspection Process

[0125] In the SQP solution process, candidate solutions are generated each time. Then, the constraint boundary management unit checks in sequence:

[0126] Do all physical constraints meet? .

[0127] Does the association constraint satisfy: and .

[0128] If any associated constraint is violated, the candidate solution is projected onto the constraint boundary for correction. For example, violating... At time, fixed Then Upgraded to ; or fixed Then Upgraded to The corrected solution falls on the boundary and continues to participate in the optimization.

[0129] Example 4: In this example, a trend-aware mechanism is added when returning to step S3 from step S5.

[0130] Definition of the first The actual improvement in the overall effluent quality score after the adjustment:

[0131]

[0132] Record recent Improvement in wheel design: , , Calculate the rate of change of improvement between two adjacent rounds:

[0133]

[0134] in The time required for each round of adjustments (approximately 1 hour).

[0135] like and (Pick ),at the same time If the value is negative or zero, it indicates that the marginal benefit of the current adjustment direction has significantly decreased. At this point, instead of waiting for natural convergence, the strategy is actively switched:

[0136] Switch the optimization range of the response surface model from the current local neighborhood (e.g., ±20% of the current parameter) to the global feasible region.

[0137] Alternatively, the polynomial order of the response surface model can be increased (from second to third order) to capture more complex nonlinear relationships.

[0138] This mechanism can effectively avoid ineffective iterations in the direction of diminishing returns, and can usually reduce the number of convergence rounds from 5-6 rounds to 3-4 rounds.

[0139] Example 5: This example addresses the situation where repeated fluctuations prevent convergence.

[0140] S5.1 Triggering conditions:

[0141] The actual comprehensive score of effluent water quality over three consecutive adjustment cycles Repeatedly crossing the first threshold The pattern is: the first Wheel of standard ( ), No. If the wheel fails to meet the standard, the first The system meets the standards again, resulting in oscillations. These oscillations are usually not due to simply unsuitable parameters, but rather to external disturbances or system aging.

[0142] S5.2, Candidate Root Cause Check:

[0143] The following three candidate root causes will be examined in turn:

[0144] Photovoltaic power supply fluctuations: Calculate the photovoltaic output voltage over the past hour. variance ,like (If a pre-set upper limit for fluctuation is set), then photovoltaic fluctuations are considered a potential root cause.

[0145] Influent load variation: Calculate the ratio of the standard deviation to the average value of the influent COD concentration over the past hour (coefficient of variation). If it is greater than 0.3, the influent load is considered to fluctuate greatly.

[0146] Photocatalyst degradation: Compare the COD removal rate of the effluent under the same control parameters with the removal rate during the initial commissioning. If the decrease exceeds 20% and the cumulative operating time exceeds the recommended replacement cycle of the catalyst (e.g., 500 hours), then the catalyst is considered to have degraded.

[0147] S5.3 Contribution Assessment:

[0148] Each candidate root cause is assigned a contribution score (0~1), using fuzzy logic:

[0149] Photovoltaic fluctuation contribution ;

[0150] Inflow load contribution ;

[0151] Catalyst attenuation contribution If the decay condition is met, otherwise 0;

[0152] The one with the highest contribution score is taken as the root cause of the anomaly.

[0153] S5.4 Targeted Intervention:

[0154] The root cause is photovoltaic fluctuations: start auxiliary energy storage units (such as batteries) to supplement power to the reactor, and set the energy storage discharge depth to not exceed 70%.

[0155] The root cause is changes in the influent load: adjust the residence time in the front-end equalization tank (for example, reduce the opening of the equalization tank effluent valve to allow the wastewater to mix in the equalization tank for a longer time) to smooth out concentration fluctuations.

[0156] The root cause is catalyst degradation: trigger in-situ catalyst regeneration (e.g., ozone oxidation for 30 minutes) or issue a prompt message "Please replace the photocatalytic material".

[0157] After the intervention is completed, the system re-enters step S3 and readjusts using the new parameters.

[0158] Example 6: In this example, the fixed weight evaluation function is changed to dynamic weight to adapt to the different emphasis requirements of different indicators at different operating stages.

[0159] COD removal rate and ammonia nitrogen removal rate In addition, it increases the turbidity of the effluent. As the third evaluation sub-dimension, the turbidity target value is set to... The turbidity evaluation index is defined as follows:

[0160]

[0161] That is, the lower the turbidity, The higher.

[0162] In the For each sampling period, calculate the normalized achievement gap for each sub-dimension:

[0163]

[0164]

[0165]

[0166] Where the target value is taken , , .

[0167] Dynamic weight calculation:

[0168] If the sum of the normalized compliance gaps of each sub-dimension is greater than zero, then the weights are distributed proportionally:

[0169]

[0170] If the normalized compliance gap for all sub-dimensions is zero (i.e., all indicators have met the requirements), then the weights of each dimension are equal.

[0171]

[0172] In the formula These represent three sub-dimensions: COD, ammonia nitrogen, and turbidity.

[0173] Final overall score:

[0174]

[0175] This dynamic weighting mechanism ensures that the system always prioritizes improving the weakest indicator during operation, achieving multi-objective collaborative optimization.

[0176] Example 7: This example provides an adaptive control system for a photovoltaic-driven photocatalytic reactor. The system includes the following modules:

[0177] Data acquisition module: Composed of a multi-channel sensor signal conditioning circuit and an ADS1115 analog-to-digital converter, it connects to water quality sensors (COD, ammonia nitrogen, turbidity), photovoltaic voltage and current sensors, and a light sensor. The sampling frequency is 1Hz, and the average value is calculated every 60 seconds as the current sample value.

[0178] The status assessment module is implemented using an embedded microcontroller (STM32F407), which internally integrates the effluent water quality evaluation function (fixed weight or dynamic weight) and status classification unit. This module receives parameters from the data acquisition module, calculates a comprehensive score, compares it with a dynamic threshold, and outputs a trigger signal.

[0179] Response surface modeling module: Implemented on a PC or high-performance embedded platform (such as Raspberry Pi 4B), using Python's numpy and scikit-learn libraries for least squares regression. It employs a combination of offline pre-modeling and online recursive least squares updating; see step S3 of Example 1 for details.

[0180] The optimization solution module is integrated within the response surface modeling module and uses the SLSQP solver from the SciPy library. This module includes constraint boundary management elements, which are further divided into:

[0181] Physically constrained sub-unit: storage , , , , .

[0182] Associative constraint sub-unit: stores associative constraint expressions and .

[0183] Verify execution subunits: Verify and correct constraints in each iteration.

[0184] Execution control module: includes a PWM signal generator, relays, and drive circuits. Output:

[0185] A 0-10V analog signal controls the constant current source of the LED driver to adjust the light intensity.

[0186] The pulse signal controls the stepper motor driver to adjust the peristaltic pump speed (dwell time).

[0187] The PWM signal controls the DC motor speed controller to adjust the speed of the mixer.

[0188] Feedback monitoring module: Communicates bidirectionally with the data acquisition module and execution control module. Internally includes a trend analysis unit and a strategy switching unit.

[0189] Trend Analysis Unit: Records the improvement sequence of the most recent N rounds of adjustments and calculates the rate of change.

[0190] Strategy switching unit: When diminishing marginal returns are detected, it sends a signal to the optimization solution module to "switch to global optimization" or "increase the model order".

[0191] Workflow:

[0192] After the system powers on, it enters the initialization phase and loads the default control parameters. , , Then execute in a loop:

[0193] The data acquisition module collects sensor data every 60 seconds and stores it in a circular buffer.

[0194] The status assessment module calculates a comprehensive score and compares it with the current threshold.

[0195] If the target is not met, the response surface modeling module, optimization solution module, and execution control module will be called sequentially to adjust the parameters.

[0196] After adjustments, the monitoring module will provide feedback on compliance status and trigger trend detection or root cause analysis if necessary.

[0197] All operational data is uploaded to the host computer monitoring system via RS485.

[0198] The embodiments of the present invention are given for the purposes of illustration and description. Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Any changes, modifications, substitutions and variations made by those skilled in the art to the above embodiments within the scope of the present invention should be included within the protection scope of the present invention.

Claims

1. A self-adaptive control method of a photocatalytic reactor driven by photovoltaics, characterized in that, Includes the following steps: Step S1: Real-time acquisition of the operating status parameter set and effluent water quality parameter set of the photocatalytic reactor; the operating status parameter set includes the output voltage, output current and light intensity of the photovoltaic power supply module; the effluent water quality parameter set includes the chemical oxygen demand (COD) concentration and ammonia nitrogen concentration; Step S2: Calculate the comprehensive score of the current effluent water quality according to the preset effluent water quality evaluation function. The evaluation function is a weighted sum of COD removal rate and ammonia nitrogen removal rate, wherein the COD removal rate and ammonia nitrogen removal rate are calculated from the current COD concentration, ammonia nitrogen concentration and the influent COD concentration, ammonia nitrogen concentration, respectively. If the comprehensive score is higher than the preset first threshold, the current control parameters are maintained unchanged and the process returns to step S1. If the comprehensive score is lower than or equal to the first threshold, the adaptive adjustment process is triggered. Step S3: In advance, at least m sets of different control parameter combinations and their corresponding effluent water quality data are collected through offline experiments, where m is greater than the number of undetermined coefficients in the response surface model, to construct an initial response surface model; During actual system operation, when the adaptive adjustment process is triggered, at least one set of actual effluent water quality data corresponding to the control parameters under the current operating conditions is collected, and the initial response surface model is incrementally updated online using the recursive least squares method; The control parameters include the light intensity setpoint, hydraulic residence time setpoint, and catalyst surface update frequency setpoint in the photocatalytic reaction chamber; Step S4: Based on the response surface model, solve for the candidate control parameter combination that yields the highest predicted comprehensive score of effluent water quality within the feasible region of the current control parameters, and send the candidate control parameter combination to the actuator of the photocatalytic reactor to update the current control parameters; Step S5: Under the condition of executing the updated control parameters, continuously monitor the actual comprehensive score of the effluent water quality. If the actual comprehensive score recovers to above the first threshold after a preset stabilization period, lock the current control parameters and return to step S1; if the actual comprehensive score still does not recover to above the first threshold, return to step S3 for a new round of adjustment until the standard is met or the preset maximum number of adjustment rounds is reached.

2. The method of claim 1, wherein, In step S2, the weights of the preset effluent water quality evaluation function are fixed values, with the weight of COD removal rate being 0.6 and the weight of ammonia nitrogen removal rate being 0.

4.

3. The method according to claim 1, characterized in that, The state grading strategy is also included before calculating the comprehensive score in step S2: Based on the output voltage of the photovoltaic power supply module and the preset voltage level threshold, the current photovoltaic state is divided into three levels: "sufficient state", "fluctuating state" and "insufficient state". The value of the first threshold is dynamically adjusted according to different levels: the original first threshold is used in the "sufficient state"; the first threshold is lowered to the second threshold in the "fluctuating state" and the second threshold is lower than the first threshold; the first threshold is lowered to the third threshold in the "insufficient state" and the third threshold is lower than the second threshold.

4. The method according to claim 1, characterized in that, Before solving for the candidate control parameter combination in step S4, a multi-constraint collaborative boundary verification step is also included: Obtain the physical constraint boundaries of each control parameter, including the upper and lower limits of light intensity, the minimum and maximum values ​​of hydraulic residence time, and the upper limit of catalyst surface renewal frequency; Identify and construct the correlation constraints between various control parameters, including: the product of hydraulic residence time and catalyst surface renewal frequency is not less than a preset minimum total reaction contact threshold, and the product of light intensity setting value and hydraulic residence time setting value is not less than a preset minimum total light energy input threshold. The physical constraint boundaries and associated constraint relationships are jointly constructed into a constraint boundary management system; When solving for candidate combinations of control parameters, only candidate solutions that simultaneously satisfy all physical constraint boundaries and associated constraint relationships are retained; if the current optimal candidate solution violates any associated constraint, it is shrunk to the inside of the boundary of the associated constraint before being adopted.

5. The method according to claim 1, characterized in that, When returning to step S3 in step S5 for a new round of adjustments, a historical trend awareness and early switching strategy is introduced: Record the improvement of the actual comprehensive score of the effluent water quality after each round of adjustment during the first N rounds of adjustment, N≥2; Calculate the rate of change of improvement between two adjacent rounds of adjustment. If the rate of change is negative and the absolute value exceeds a preset threshold for diminishing returns, then it is determined that the marginal return of the current adjustment direction has significantly decreased. After determining that the marginal returns have decreased significantly, the adjustment strategy is actively switched: the optimization interval of the response surface model is switched from the local neighborhood of the current control parameters to the global feasible region, or the polynomial order of the response surface model is increased to capture nonlinear features.

6. The method according to claim 1, characterized in that, After locking the current control parameters in step S5, the process also includes steps for tracing the root cause of the anomaly and providing targeted intervention: When the actual comprehensive score of the effluent water quality fluctuates around the first threshold for more than three consecutive adjustment cycles; Check the following candidate root causes in sequence: photovoltaic power supply fluctuation exceeds the preset fluctuation limit, influent COD concentration fluctuation exceeds the preset load change threshold, and photocatalytic material catalytic efficiency decays to below the preset efficiency lower limit. The candidate root cause that contributes the most to the current fluctuation in effluent water quality during the inspection is identified as the abnormal root cause. Based on the type of abnormal root cause, perform targeted intervention actions: if the root cause is fluctuations in photovoltaic power supply, activate the auxiliary energy storage unit to compensate for power supply; if the root cause is changes in influent load, adjust the residence time of the homogenization and conditioning tank at the front end. If the catalyst material degrades, a prompt will be triggered to regenerate or replace the catalyst in situ.

7. The method according to claim 1, characterized in that, The weights of the effluent quality evaluation function in step S2 are not fixed values, but are dynamically calibrated. Multiple sub-dimensions for evaluating effluent water quality were set, including COD removal rate, ammonia nitrogen removal rate, and effluent turbidity. In each sampling period, the normalized achievement gap between the actual value of each sub-dimension and its corresponding achievement target value is calculated; The evaluation weight for the current period is dynamically calculated based on the normalized achievement gap of each sub-dimension: the sub-dimension with the larger achievement gap is assigned a higher evaluation weight, and the sub-dimension with the smaller achievement gap is assigned a lower evaluation weight. The weights are determined as follows: like ,but ; like ,but ; The overall score is the weighted sum of the actual values ​​of each sub-dimension and their current periodic dynamic evaluation weights.

8. An adaptive control system for a photovoltaic-driven photocatalytic reactor, used to execute the method according to any one of claims 1 to 7, characterized in that, include: Data acquisition module: It is connected to the water quality sensor in the photocatalytic reactor and the light and electrical parameter sensors of the photovoltaic power supply module, respectively, to acquire the set of operating status parameters and the set of effluent water quality parameters in real time; Status assessment module: connected to the data acquisition module, with a pre-set effluent water quality evaluation function and a first threshold, used to calculate the comprehensive score of the current effluent water quality and compare it with the first threshold to trigger the adaptive adjustment process; Response surface modeling module: connected to the state assessment module, used to store the initial response surface model constructed in advance by collecting at least m sets of different control parameter combinations and their corresponding effluent water quality data through offline experiments; after triggering the adaptive adjustment process, it collects at least one set of actual effluent water quality data corresponding to the control parameters under the current operating conditions, and uses the recursive least squares method to perform online incremental updates on the initial response surface model; The optimization solution module is connected to the response surface modeling module and is used to solve for the candidate control parameter combination that maximizes the predicted comprehensive score of the effluent water quality within the feasible region of the control parameters. The optimization solution module integrates a constraint boundary management unit for performing multi-constraint collaborative boundary checks. Execution control module: connected to the optimization solution module and the actuator of the photocatalytic reactor, used to send candidate control parameter combinations to the actuator and update the current control parameters; Feedback monitoring module: Connected to the data acquisition module and execution control module, it is used to continuously monitor the actual comprehensive score of the effluent water quality after executing the updated control parameters, and decide to lock the current parameters or trigger a new round of adjustments based on the compliance status.

9. The system according to claim 8, characterized in that, The constraint boundary management unit inside the optimization solution module includes: The physical constraint subunit is used to store the physical constraint boundaries of each control parameter; The associated constraint subunit is used to store and identify the preset associated constraint relationships between various control parameters; The verification execution subunit is used to verify whether the candidate solution satisfies both physical constraints and associated constraints when solving for candidate control parameter combinations, and to perform boundary shrinkage correction on candidate solutions that violate constraints.

10. The system according to claim 8, characterized in that, The feedback monitoring module also includes a trend analysis unit and a strategy switching unit. The trend analysis unit is used to record the improvement rate of the actual comprehensive score of the effluent water quality during the first N rounds of adjustment, and to calculate the rate of change of the improvement rate. The strategy switching unit is connected to the trend analysis unit and is used to actively send a strategy switching signal to the optimization solution module when the rate of change shows a significant decrease in marginal returns, instructing the optimization solution module to switch the optimization interval from the current local neighborhood to the global feasible region.

Images