Automatic equipment control method for water treatment equipment

Abstract

The invention relates to the technical field of water quality monitoring, and discloses an automatic equipment control method for water treatment equipment, which is characterized in that a water treatment process is modeled into a Markov decision process, and dynamic and optimal control on the water treatment equipment is realized. The model comprehensively considers a plurality of dimensions such as raw water quality, treatment water quality, equipment operation state and treatment flow, and defines comprehensive state space and action space. By designing a reward function, multiple indexes such as the water quality standard-reaching rate, the equipment operation efficiency and the energy consumption are evaluated, and a deep reinforcement learning algorithm is adopted to continuously optimize a control strategy so as to maximize accumulated rewards. According to the method, the defects of response lag, strategy stiffness, lack of global optimization capability and the like of a traditional control mode based on a threshold value and manual intervention are overcome, the method can be adaptive to complex and changeable water quality conditions and equipment operation states, the treatment efficiency is improved, the water quality safety is ensured, and the operation cost is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of water quality monitoring, and in particular to a method for controlling automated equipment of water treatment equipment. Background Art

[0002] In the field of water treatment, automated equipment control is crucial to improving treatment efficiency, ensuring water quality safety and reducing operating costs. The current water treatment equipment control method is based on preset threshold rules. By monitoring key water quality parameters (such as pH value, turbidity, dissolved oxygen content, etc.) and comparing them, when the water quality indicators exceed the preset range, corresponding control actions are triggered, such as adjusting the dosage, turning on and off the pump, etc.

[0003] Judgments based on fixed thresholds cannot make immediate adjustments based on the trend of water quality changes or the dynamic fluctuations of equipment performance. This results in delayed system response when faced with sudden water quality deterioration or equipment failure, making it difficult to adopt the optimal control strategy in a timely manner, affecting water quality safety and treatment efficiency. Setting reasonable thresholds and formulating emergency treatment plans also require a lot of reliance on the experience and knowledge of operators, which not only increases the risk of human error, but also limits the further improvement of the level of automated control. With the increase in the scale and complexity of water treatment, it is difficult to meet the needs of efficient and precise control based solely on human experience.

[0004] Moreover, current control strategies usually only focus on the control of a single indicator, ignoring the multi-objective optimization issues in the water treatment process, such as simultaneously considering water quality compliance, energy efficiency, equipment operating costs, etc. This local optimization approach often leads to low overall system efficiency, and in some cases may even sacrifice other important performances due to over-adjustment of a certain indicator. Summary of the invention

[0005] The object of the present invention is to provide a water treatment equipment automation equipment control method to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: a water treatment equipment automation equipment control method, the method comprising:

[0007] Construct an intelligent control model for automated control of water treatment equipment, model the water treatment process as a Markov decision process, and dynamically adjust the operation strategy of the water treatment equipment by analyzing the real-time data of the water treatment system;

[0008] The steps to build an intelligent control model include:

[0009] Define the state space, including the raw water quality dimension, treated water quality dimension, equipment operation status dimension and treatment flow dimension of the water treatment system;

[0010] The raw water quality dimensions include pH value, turbidity, dissolved oxygen content, residual chlorine content and other relevant water quality indicators of raw water;

[0011] The treated water quality dimensions include the pH value, turbidity, dissolved oxygen content, residual chlorine content, water quality compliance and other relevant water quality indicators of the treated water;

[0012] The equipment operation status dimension includes the working status, operation parameters, fault conditions and maintenance records of each water treatment equipment;

[0013] The processing flow dimensions include current processing flow, average processing flow and maximum processing flow;

[0014] Define the action space, which includes the possible actions that the agent can take when dealing with abnormal water quality and equipment operation failures, including adjusting equipment parameters, switching equipment operation modes, starting backup equipment, performing equipment maintenance, and notifying operators;

[0015] Design a reward function to evaluate the results of actions based on water quality compliance rate, equipment operating efficiency, energy consumption index, maintenance cost and emergency response capability, and give corresponding rewards or penalties;

[0016] Formulate strategies, rules for selecting actions in a given state, and continuously optimize them through machine learning algorithms to maximize cumulative rewards;

[0017] A deep reinforcement learning algorithm is used to train the intelligent control model.

[0018] The water quality compliance rate is evaluated based on the proportion of water quality reaching the prescribed standards after treatment;

[0019] The equipment operation efficiency is evaluated based on the equipment operation time, processing flow rate and processing effect;

[0020] The energy consumption index is evaluated based on the energy consumption and processing efficiency of the equipment;

[0021] The maintenance cost is evaluated based on the number of maintenance times, maintenance costs and maintenance time of the equipment;

[0022] The emergency response capability is evaluated based on the response time and processing effect when equipment fails.

[0023] Preferably, raw water quality dimensional data, treated water quality dimensional data, equipment operation status dimensional data and processed flow dimensional data are collected in real time through sensors and monitoring systems installed on the water treatment equipment as the original data of the state space S.

[0024] Preferably, the components of the reward function include water quality compliance rate reward, equipment operation efficiency reward, energy consumption index penalty, maintenance cost penalty and emergency response capability reward:

[0025] The calculation formula of the water quality compliance rate reward function is: R_water = k1*(number of water quality indicators that meet the standards / total number of water quality indicators); k1 is the water quality compliance rate reward coefficient, which is adjusted according to the actual application scenario;

[0026] The calculation formula of the equipment operation efficiency reward function is: R_efficiency = k2*(actual processing flow / equipment maximum processing flow)*(equipment operation time / total operation time); where k2 is the equipment operation efficiency reward coefficient, which is adjusted according to the actual application scenario;

[0027] The calculation formula of the energy consumption index penalty function is: P_energy = k3*(equipment energy consumption / standard energy consumption); where k3 is the energy consumption index penalty coefficient, which is adjusted according to the actual application scenario; standard energy consumption is the energy consumption standard value set according to the equipment type and operating parameters;

[0028] The calculation formula of the maintenance cost penalty function is: P_maintenance = k4*(maintenance times / expected maintenance times) + k5*(maintenance cost / expected maintenance cost); where k4 is the maintenance times penalty coefficient, and k5 is the maintenance cost penalty coefficient, which are adjusted according to the actual application scenario; the expected maintenance times and expected maintenance costs are expected values ​​set according to the equipment type, operating time, and maintenance plan;

[0029] The calculation formula of the emergency response capability reward function is: R_response = k6*(response time threshold - actual response time) / response time threshold; where k6 is the emergency response capability reward coefficient, which is adjusted according to the actual application scenario; the response time threshold is the maximum value of the set equipment failure response time;

[0030] Through the above reward and penalty functions, the final reward function R is obtained: R = R_water + R_efficiency - P_energy - P_maintenance + R_response.

[0031] Preferably, a deep Q network DQN algorithm is used to train the intelligent control model.

[0032] Preferably, the step of training the intelligent control model includes:

[0033] S1: Initialize the deep Q network: Build a deep neural network as a Q-value function approximator, with the input as the state s and the output as the Q value corresponding to each action a; initialize the network parameters θ;

[0034] S2: Initialize the experience playback buffer D: Set the buffer size N to store the experience generated by the intelligent control model during the training process;

[0035] S3: Using the collected training data, the parameters of the neural network are updated through the back-propagation algorithm and the gradient descent method. During the training process, the model continuously tries different actions and adjusts its strategy based on the feedback obtained from the reward function to maximize the cumulative reward.

[0036] S4: During the training process, the performance of the intelligent control model is regularly evaluated, and the weight coefficient of the reward function and the learning rate of the model are adjusted according to the evaluation results;

[0037] S5: Repeat steps S3 and S4 until a preset number of training rounds is reached.

[0038] Preferably, in step S3, the formula for updating the parameters of the gradient descent algorithm is:

[0039]

[0040] Among them, θ represents the model parameters, including all weights and biases that the model needs to learn; ω represents the learning rate, which is used to control the step size of parameter update; represents the gradient of the loss function J(θ) with respect to the parameter θ, is a vector pointing to the direction where the loss function grows fastest; := represents the assignment operation, that is, updating the value of the parameter θ.

[0041] Preferably, after the intelligent control model training is completed, the model is evaluated by combining simulation and field application.

[0042] Preferably, the method for performing simulation testing on the intelligent control model includes:

[0043] Step 1: Simulation environment construction: Use MATLAB / Simulink simulation software to build a simulation environment for water treatment equipment. This environment can be used to simulate the actual operating status of the equipment, including water flow dynamics, water quality changes, equipment energy consumption, and equipment failures.

[0044] Step 2: Simulation case design: In the constructed simulation environment, design a variety of test cases to cover various situations that may be encountered in actual operations, including treatment effects under different water quality conditions, operation strategies for different equipment combinations, optimization plans under different energy consumption constraints, and emergency treatment in the event of equipment failure;

[0045] Step 3, simulation data generation: Run the simulation environment and generate simulation data according to the designed test cases, including the state changes of the equipment, the dynamic changes of water quality indicators, the energy consumption records of the equipment and the operation records of the equipment;

[0046] Step 4: Compare model predictions with simulation results: Input simulation data into the trained deep Q network model to obtain the operation strategy predicted by the model, and simulate the strategy in the simulation environment to obtain the model's prediction results; at the same time, simulate the traditional control method under the same conditions to obtain the simulation results of the traditional method;

[0047] Step 5: Performance index evaluation: Compare the model prediction results with the traditional method simulation results in terms of multiple performance indicators, including the stability of water quality compliance, energy saving, equipment life, system response speed, and equipment failure handling efficiency.

[0048] Preferably, the method for field application verification of the intelligent control model includes: based on simulation evaluation, selecting some typical water treatment equipment or scenarios for field application verification, deploying the control strategy of the deep Q network model to the actual equipment, conducting long-term tracking monitoring, collecting actual operation data, and further verifying the actual application effect of the model.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] The present invention realizes dynamic adjustment of the operation strategy of water treatment equipment by modeling the water treatment process as a Markov decision process and analyzing system data in real time. Through the dynamic response mechanism, the trend of water quality changes and equipment performance fluctuations are effectively captured, ensuring that the system can quickly adopt the optimal control strategy in the face of sudden water quality deterioration or equipment failure, significantly improving water quality safety and treatment efficiency.

[0051] The intelligent control model takes into account multiple factors such as water quality compliance, equipment operating efficiency, energy consumption indicators, maintenance costs, etc. when designing. By comprehensively evaluating the results of actions and giving corresponding rewards or penalties, global optimization is achieved, avoiding the problem of low overall system efficiency caused by excessive focus on a single indicator in traditional control strategies, and achieving the best processing effect under multi-objective coordination. The intelligent control model can monitor the equipment operating status and water quality indicators in real time. Once an abnormal situation is found, it can quickly trigger the early warning mechanism and take corresponding emergency response actions, effectively shortening the time for fault discovery and processing. By comprehensively considering the equipment operating efficiency and energy consumption indicators, the intelligent control model can achieve a more energy-saving and efficient equipment operation strategy while ensuring that the water quality meets the standards. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a working principle diagram of a water treatment equipment automation equipment control method according to the present invention;

[0053] Figure 2 Schematic diagram of the design of the reward function;

[0054] Figure 3 This is the training flow chart of the intelligent control model;

[0055] Figure 4 Evaluation flow chart of the intelligent control model. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] See also Figure 1-4 The present invention provides a technical solution: a water treatment equipment automation equipment control method, the method comprising:

[0058] ① Construct an intelligent control model and regard the water treatment process as a Markov decision process, in which the state space, action space and reward function are the key elements of the model.

[0059] The state space covers multiple dimensions of the water treatment system, including but not limited to:

[0060] Raw water quality dimensions: including key water quality indicators such as raw water pH value, turbidity, dissolved oxygen content, residual chlorine content, etc. These indicators can comprehensively reflect the water quality of raw water.

[0061] Treatment water quality dimensions: including the pH value, turbidity, dissolved oxygen content, residual chlorine content and water quality compliance of the treated water, which are used to evaluate the treatment effect.

[0062] Equipment operating status dimension: covers the working status, operating parameters, fault conditions and maintenance records of each water treatment equipment to understand the operating status of the equipment in real time.

[0063] Processing flow dimension: includes current processing flow, average processing flow and maximum processing flow to reflect the processing capacity and load of the equipment.

[0064] ② Define the action space, which contains all the actions that the agent may take when dealing with abnormal water quality and equipment operation failures, including:

[0065] Adjust equipment parameters: such as adjusting dosage, changing filtration speed, etc.

[0066] Switch the equipment operation mode: such as switching from automatic mode to manual mode, or from energy-saving mode to high-efficiency mode.

[0067] Start backup equipment: When the main equipment fails or is under maintenance, start the backup equipment to ensure the continuity of the processing process.

[0068] Perform equipment maintenance: Arrange appropriate maintenance activities based on the equipment's operating status and maintenance records.

[0069] Notify the operator: When a serious abnormality or failure occurs, promptly notify the operator so that emergency measures can be taken.

[0070] ③ Design a reward function, which evaluates the results of actions based on multiple factors such as water quality compliance rate, equipment operating efficiency, energy consumption index, maintenance cost and emergency response capability, and gives corresponding rewards or penalties. Specific evaluation indicators include:

[0071] Water quality compliance rate: evaluated based on the proportion of treated water that meets the specified standards. The higher the compliance rate, the greater the reward.

[0072] Equipment operation efficiency: Evaluated based on the equipment's operating time, processing flow, and processing effect. The higher the efficiency, the greater the reward.

[0073] Energy consumption index: Evaluate based on the energy consumption and processing efficiency of the equipment. The lower the energy consumption and the higher the processing efficiency, the greater the reward; otherwise, a penalty will be given.

[0074] Maintenance cost: Evaluated based on the number of maintenance times, maintenance costs and maintenance time of the equipment. The lower the maintenance cost, the greater the reward; otherwise, a penalty will be given.

[0075] Emergency response capability: Evaluated based on the response time and handling effect when equipment fails. The faster the response and the better the handling effect, the greater the reward.

[0076] ④ Formulate a strategy. The rules for selecting actions in a given state are called strategies. This embodiment uses machine learning algorithms, especially deep reinforcement learning algorithms, to continuously optimize strategies to maximize cumulative rewards. Through continuous trial and error and learning, the intelligent control model can gradually master the strategy of taking the best action in different states.

[0077] ⑤ Train the intelligent control model and use the deep reinforcement learning algorithm to train the intelligent control model. During the training process, the model interacts with the environment (i.e., the water treatment system), continuously tries different actions and receives feedback (i.e., rewards or penalties) from the environment. Through repeated iterations and learning, the model gradually learns how to take the best action strategy in different states to achieve the global optimization goal.

[0078] ⑥After training, the intelligent control model is deployed to the actual water treatment system for real-time control. The model can monitor the equipment operation status and water quality indicators in real time, and select the optimal action strategy for adjustment according to the current status. In the face of sudden water quality deterioration or equipment failure, the model can respond quickly and adopt the optimal control strategy to ensure water quality safety and treatment efficiency.

[0079] This outbreak is further described below in conjunction with Examples 1 to 3:

[0080] Embodiment 1:

[0081] This example is used to describe the design of the reward function, which directly affects the learning effect of the model and the final control strategy. The reward function mainly consists of five parts: water quality compliance rate reward, equipment operation efficiency reward, energy consumption index penalty, maintenance cost penalty and emergency response capability reward.

[0082] ① Water quality compliance rate reward: The design of the water quality compliance rate reward function is to encourage the intelligent control model to prioritize the quality of treated water. In specific implementation, a water quality compliance rate reward coefficient k1 can be set, and the value range of this coefficient can be adjusted according to the actual application scenario. For example, set k1 = 0.5. Then, the water quality compliance rate reward is calculated by the following formula:

[0083] R_water = k1*(number of water quality indicators that meet the standards / total number of water quality indicators)

[0084] Assuming that in a certain control cycle, the total number of water quality indicators is 5, and the number of water quality indicators that meet the standards is 4, then R_water = 0.5*(4 / 5) = 0.4.

[0085] ② Equipment operation efficiency reward: The design of the equipment operation efficiency reward function is to encourage the intelligent control model to improve the equipment operation efficiency. In specific implementation, an equipment operation efficiency reward coefficient k2 can be set, for example, k2 = 0.3. Then, the equipment operation efficiency reward is calculated by the following formula:

[0086] R_efficiency = k2*(actual processing flow / maximum processing flow of the equipment)*(equipment operation time / total operation time)

[0087] Assume that in a certain control cycle, the maximum processing flow of the device is 100m 3 / h, the actual processing flow is 80m 3 / h, the equipment running time is 8 hours, and the total running time is 24 hours, then R_efficiency=0.3*(80 / 100)*(8 / 24)=0.08.

[0088] ③ Energy consumption index penalty: The design of the energy consumption index penalty function is to limit the excessive energy consumption of the intelligent control model in the pursuit of water quality and equipment operation efficiency. In specific implementation, an energy consumption index penalty coefficient k3 can be set, for example, k3 = 0.2. Then, the energy consumption index penalty is calculated by the following formula:

[0089] P_energy=k3*(equipment energy consumption / standard energy consumption)

[0090] Assuming that in a certain control cycle, the equipment energy consumption is 120 kWh and the standard energy consumption is 100 kWh, then P_energy=0.2*(120 / 100)=0.24.

[0091] ④ Maintenance cost penalty: The design of the maintenance cost penalty function is to encourage the intelligent control model to reduce the number and cost of equipment maintenance. In specific implementation, the maintenance number penalty coefficient k4 and the maintenance cost penalty coefficient k5 can be set, for example, k4 = 0.1, k5 = 0.15. Then, the maintenance cost penalty is calculated by the following formula:

[0092] P_maintenance = k4*(maintenance times / expected maintenance times)+k5*(maintenance cost / expected maintenance cost)

[0093] Assume that in a certain control cycle, the actual maintenance times are 3 times, the expected maintenance times are 2 times, the actual maintenance cost is 300 yuan, and the expected maintenance cost is 200 yuan, then P_maintenance = 0.1*(3 / 2)+0.15*(300 / 200)=0.15+0.225=0.375.

[0094] ⑤ Emergency response capability reward: The design of the emergency response capability reward function is to encourage the intelligent control model to respond quickly and take measures when the equipment fails. In specific implementation, an emergency response capability reward coefficient k6 can be set, for example, k6 = 0.1. Then, the emergency response capability reward is calculated by the following formula:

[0095] R_response = k6*(response time threshold - actual response time) / response time threshold

[0096] Assume that in a certain equipment failure, the response time threshold is 10 minutes and the actual response time is 5 minutes, then R_response = 0.1*(10-5) / 10 = 0.05.

[0097] Combining the above reward and penalty functions, we get the final reward function R:

[0098] R=R_water+R_efficiency-P_energy-P_maintenance+R_response

[0099] Substituting the above specific values, we obtain R=0.4+0.08-0.24-0.375+0.05=-0.085.

[0100] During the training process of the intelligent control model, the value of the reward function R is continuously calculated, and the control strategy of the model is adjusted according to the value of R, thereby realizing automatic control of the water treatment equipment.

[0101] Embodiment 2:

[0102] The following describes in detail the training implementation steps of the intelligent control model of water treatment equipment in combination with a specific water treatment equipment control scenario.

[0103] Assume that a water treatment system consists of three main devices: an inlet pump, a filter, and a reaction tank. The inlet pump is responsible for pumping raw water into the system, the filter is used to remove impurities in the water, and the reaction tank treats the water by adding chemicals. The goal of the system is to control the operating parameters of these devices, such as the flow rate of the inlet pump, the pressure of the filter, the amount of chemicals added to the reaction tank, etc., so that the treated water quality meets the specified standards and minimizes energy consumption and maintenance costs.

[0104] Model training implementation steps:

[0105] ① Define the state space S: The state space S includes water quality indicators such as the flow rate of the water inlet pump, the pressure of the filter, the pH value of the reaction tank, turbidity, residual chlorine, and the operating status of each device (on / off, fault / normal). For example, the state s can be expressed as:

[0106] s = [inlet pump flow, filter pressure, reaction pool pH value, reaction pool turbidity, reaction pool residual chlorine, inlet pump status, filter status, reaction pool status]

[0107] ② Define action space A: Action space A includes operations such as adjusting the flow of the water inlet pump, adjusting the pressure of the filter, adjusting the dosage of the reaction tank, switching the equipment operation mode (such as switching from automatic to manual), starting or stopping the backup equipment, etc. For example, action a can be expressed as:

[0108] a=[adjust the water inlet pump flow +10%, adjust the filter pressure -5%, increase the amount of reagent added to the reaction tank, switch to manual mode, start the standby filter]

[0109] Among them, each action corresponds to a specific device operation.

[0110] ③ Design reward function R: The reward function R is evaluated based on water quality compliance rate, equipment operation efficiency, energy consumption index, maintenance cost and emergency response capability. For example, the following reward and penalty rules can be set:

[0111] Water quality compliance rate: For every 1% increase, the reward is +1;

[0112] Equipment operation efficiency: For every 10% increase, the reward is +5;

[0113] Energy consumption index: For every 10% above the standard energy consumption, the penalty is -3;

[0114] Maintenance cost: Each maintenance, penalty -10;

[0115] Emergency response capability: Respond within 5 minutes after equipment failure, reward +20; otherwise, penalty -20.

[0116] Combining the above rules, calculate the reward r for each time step.

[0117] ④ Initialize the deep Q network: Build a deep neural network, the input layer receives the state s, and the output layer is the Q value corresponding to each action a. Use random parameters to initialize the network parameters θ.

[0118] ⑤ Initialize the experience playback buffer D: set the buffer size N, for example N = 10000, to store the experience (s, a, r, s').

[0119] ⑥Start the training cycle:

[0120] a) The initialization state s is the actual state of the current water treatment equipment, such as data obtained by sensors.

[0121] b) Use the ε-greedy strategy to select action a. For example, set ε = 0.1, that is, select an action randomly with a probability of 10%, and select the action with the maximum Q value with a probability of 90%.

[0122] c) Execute action a and observe the reward r and next state s'. For example, after adjusting the flow of the water inlet pump, the new water quality index and equipment status are obtained through the sensor.

[0123] d) Store the experience (s, a, r, s') into the experience replay buffer D. If the buffer is full, overwrite the oldest experience.

[0124] e) Randomly extract an experience batch (minibatch) from the experience replay buffer D, for example, M=64.

[0125] f) For each extracted experience (s_i, a_i, r_i, s'_i), calculate the target Q value y_i. If s'_i is a terminal state (such as water quality has reached the standard or the equipment has been shut down), then y_i = r_i; otherwise, y_i = r_i + γ * max_a'Q (s'_i, a'; θ^-); where γ is the discount factor and θ^- is the target network parameter of the Q network, which is copied from θ regularly.

[0126] g) Use the mean square error loss function to update the parameters θ of the deep Q network through the gradient descent method. The optimization goal is to minimize the error between the target Q value and the predicted Q value: L(θ) = 1 / M*Σ_i(y_i-Q(s_i,a_i;θ))^2. The formula for updating the parameters of the gradient descent algorithm is:

[0127]

[0128] Among them, θ represents the model parameters, including all weights and biases that the model needs to learn; ω represents the learning rate, which is used to control the step size of parameter update; represents the gradient of the loss function J(θ) with respect to the parameter θ, is a vector pointing to the direction where the loss function grows fastest; := represents the assignment operation, that is, updating the value of the parameter θ.

[0129] h) Every 1000 steps, copy the parameters θ of the deep Q network to the target network parameters θ^-.

[0130] i) Update state s to s', and repeat steps b) to i) until a preset number of training rounds (such as 100,000 rounds) is reached or other stopping conditions are met (such as the water quality compliance rate is stable at more than 95%).

[0131] ⑦ After the training is completed, the trained deep Q network is used as the intelligent control model. In the actual operation process, the optimal action a (that is, the action with the maximum Q value) is selected according to the current state s to realize the automatic control of the water treatment equipment. For example, if the current state shows that the flow rate of the water inlet pump is too low and the water turbidity is high, the model will choose to increase the flow rate of the water inlet pump and increase the dosage of the reaction tank.

[0132] Embodiment 3:

[0133] The following describes in detail the implementation of the intelligent control model evaluation method with reference to specific examples.

[0134] This embodiment selects a city sewage treatment plant as the evaluation object. The plant has multiple sets of water treatment equipment, including grids, sedimentation tanks, biological reaction tanks, secondary sedimentation tanks, disinfection tanks, etc. The treatment process is complex and has high requirements for automated control.

[0135] ①Simulation environment construction

[0136] Using MATLAB / Simulink simulation software, a simulation environment for water treatment equipment is constructed based on the actual process flow and equipment parameters of the sewage treatment plant. The simulation environment includes water flow dynamics model, water quality change model, equipment energy consumption model, equipment failure model, etc., which can fully simulate the actual operating status of the equipment.

[0137] ②Simulation case design

[0138] In the constructed simulation environment, design the following test cases:

[0139] Treatment effect under different water quality conditions: simulate water quality from different sources (such as domestic sewage, industrial wastewater, etc.) and evaluate the treatment effect of the model under different water quality conditions.

[0140] Operation strategies for different equipment combinations: Evaluate the operation strategies of the model under various equipment combinations (such as screen + sedimentation tank, biological reactor + secondary sedimentation tank, etc.) to find the optimal equipment combination.

[0141] Optimization schemes under different energy consumption constraints: Set different energy consumption constraints and evaluate the optimization schemes of the model under energy consumption constraints to achieve the goal of energy saving and consumption reduction.

[0142] Emergency response in case of equipment failure: simulate equipment failure (such as pump failure, valve blockage, etc.) and evaluate the model's emergency response capabilities and recovery strategies in case of failure.

[0143] ③Simulation data generation

[0144] According to the designed test cases, the simulation environment is run to generate simulation data. The simulation data includes the state changes of the equipment (such as opening, closing, operating load, etc.), the dynamic changes of water quality indicators (such as COD, BOD, ammonia nitrogen, etc.), the energy consumption records of the equipment, the operation records of the equipment, etc.

[0145] ④Comparison between model prediction and simulation results

[0146] Input simulation data into the trained deep Q network model to obtain the operation strategy predicted by the model. Simulate and execute the prediction strategy in the simulation environment to obtain the prediction results of the model. At the same time, simulate the traditional control method (such as PID control, fuzzy control, etc.) or manual operation strategy under the same conditions to obtain the simulation results of the traditional method.

[0147] ⑤Performance index evaluation

[0148] Compare the model prediction results with the traditional method simulation results on the following performance indicators:

[0149] Stability of water quality compliance: Evaluate the treatment effect of the model under different water quality conditions, with the water quality compliance rate as the main indicator.

[0150] Energy saving degree: Compare the energy consumption data of the model prediction strategy and the traditional method in actual operation, with the energy saving rate as the main indicator.

[0151] Equipment life: Evaluate the impact of model prediction strategies on equipment life, with equipment failure rate and maintenance frequency as the main indicators.

[0152] System response speed: Evaluate the response speed of the model when the equipment status changes or the water quality changes, with the response time as the main indicator.

[0153] Equipment failure handling efficiency: Simulate equipment failure situations and evaluate the model's emergency handling capabilities and recovery strategies in failure situations, with fault handling time and recovery efficiency as the main indicators.

[0154] ⑥ Field application verification

[0155] Based on the simulation evaluation, some typical equipment or scenarios of the sewage treatment plant were selected for field application verification. The control strategy of the deep Q network model was deployed to the actual equipment for long-term tracking and monitoring. Actual operation data, including water quality indicators, equipment status, energy consumption data, etc., were collected to further verify the actual application effect of the model.

[0156] ⑦Evaluation report generation

[0157] According to the results of simulation evaluation and field application verification, a detailed evaluation report is written. The report content includes the evaluation purpose, evaluation method, evaluation results, problem analysis, improvement suggestions, etc. The report should objectively and comprehensively reflect the performance of the model in all aspects and provide a scientific basis for the subsequent optimization and promotion of the model.

[0158] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0159] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A water treatment equipment automation equipment control method, characterized in that: The method comprises: Construct an intelligent control model for automated control of water treatment equipment, model the water treatment process as a Markov decision process, and dynamically adjust the operation strategy of the water treatment equipment by analyzing the real-time data of the water treatment system; The steps to build an intelligent control model include: Define the state space, including the raw water quality dimension, treated water quality dimension, equipment operation status dimension and treatment flow dimension of the water treatment system; The raw water quality dimensions include pH value, turbidity, dissolved oxygen content, residual chlorine content and other relevant water quality indicators of raw water; The treated water quality dimensions include the pH value, turbidity, dissolved oxygen content, residual chlorine content, water quality compliance and other relevant water quality indicators of the treated water; The equipment operation status dimension includes the working status, operation parameters, fault conditions and maintenance records of each water treatment equipment; The processing flow dimensions include current processing flow, average processing flow and maximum processing flow; Define the action space, which includes the possible actions that the agent can take when dealing with abnormal water quality and equipment operation failures, including adjusting equipment parameters, switching equipment operation modes, starting backup equipment, performing equipment maintenance, and notifying operators; Design a reward function to evaluate the results of actions based on water quality compliance rate, equipment operating efficiency, energy consumption index, maintenance cost and emergency response capability, and give corresponding rewards or penalties; Formulate strategies, rules for selecting actions in a given state, and continuously optimize them through machine learning algorithms to maximize cumulative rewards; A deep reinforcement learning algorithm is used to train the intelligent control model.

2. A water treatment equipment automation equipment control method according to claim 1, characterized in that: The water quality compliance rate is evaluated based on the proportion of water quality reaching the prescribed standards after treatment; The equipment operation efficiency is evaluated based on the equipment operation time, processing flow rate and processing effect; The energy consumption index is evaluated based on the energy consumption and processing efficiency of the equipment; The maintenance cost is evaluated based on the number of maintenance times, maintenance costs and maintenance time of the equipment; The emergency response capability is evaluated based on the response time and processing effect when equipment fails.

3. A water treatment equipment automation equipment control method according to claim 2, characterized in that: Through the sensors and monitoring systems installed on the water treatment equipment, the raw water quality dimension data, the treated water quality dimension data, the equipment operation status dimension data and the processed flow dimension data are collected in real time as the original data of the state space S.

4. A water treatment equipment automation equipment control method according to claim 3, characterized in that: The components of the reward function include water quality compliance rate reward, equipment operation efficiency reward, energy consumption index penalty, maintenance cost penalty and emergency response capability reward: The calculation formula of the water quality compliance rate reward function is: R_water = k1*(number of water quality indicators that meet the standards / total number of water quality indicators); where k1 is the water quality compliance rate reward coefficient, which is adjusted according to the actual application scenario; The calculation formula of the equipment operation efficiency reward function is: R_efficiency = k2*(actual processing flow / equipment maximum processing flow)*(equipment operation time / total operation time); where k2 is the equipment operation efficiency reward coefficient, which is adjusted according to the actual application scenario; The calculation formula of the energy consumption index penalty function is: P_energy = k3*(equipment energy consumption / standard energy consumption); where k3 is the energy consumption index penalty coefficient, which is adjusted according to the actual application scenario; standard energy consumption is the energy consumption standard value set according to the equipment type and operating parameters; The calculation formula of the maintenance cost penalty function is: P_maintenance = k4*(maintenance times / expected maintenance times) + k5*(maintenance cost / expected maintenance cost); where k4 is the maintenance times penalty coefficient, and k5 is the maintenance cost penalty coefficient, which are adjusted according to the actual application scenario; the expected maintenance times and expected maintenance costs are expected values ​​set according to the equipment type, operating time, and maintenance plan; The calculation formula of the emergency response capability reward function is: R_response = k6*(response time threshold - actual response time) / response time threshold; where k6 is the emergency response capability reward coefficient, which is adjusted according to the actual application scenario; the response time threshold is the maximum value of the set equipment failure response time; Through the above reward and penalty functions, the final reward function R is obtained: R = R_water + R_efficiency - P_energy - P_maintenance + R_response.

5. A water treatment equipment automation equipment control method according to claim 4, characterized in that: The deep Q network DQN algorithm is used to train the intelligent control model.

6. A water treatment equipment automation equipment control method according to claim 5, characterized in that: The steps to train the intelligent control model include: S1: Initialize the deep Q network: Build a deep neural network as a Q-value function approximator, with the input as the state s and the output as the Q value corresponding to each action a; initialize the network parameters θ; S2: Initialize the experience playback buffer D: Set the buffer size N to store the experience generated by the intelligent control model during the training process; S3: Using the collected training data, the parameters of the neural network are updated through the back-propagation algorithm and the gradient descent method. During the training process, the model continuously tries different actions and adjusts its strategy based on the feedback obtained from the reward function to maximize the cumulative reward. S4: During the training process, the performance of the intelligent control model is regularly evaluated, and the weight coefficient of the reward function and the learning rate of the model are adjusted according to the evaluation results; S5: Repeat steps S3 and S4 until a preset number of training rounds is reached.

7. A water treatment equipment automation equipment control method according to claim 6, characterized in that: In step S3, the formula for updating parameters of the gradient descent algorithm is: Among them, θ represents the model parameters, including all weights and biases that the model needs to learn; ω represents the learning rate, which is used to control the step size of parameter update; represents the gradient of the loss function J(θ) with respect to the parameter θ, is a vector pointing to the direction where the loss function grows fastest; := represents the assignment operation, that is, updating the value of the parameter θ.

8. A water treatment equipment automation equipment control method according to claim 1, characterized in that: After the intelligent control model training is completed, the model is evaluated by combining simulation and field application.

9. A water treatment equipment automation equipment control method according to claim 8, characterized in that: The method of performing simulation test on the intelligent control model includes: Step 1: Simulation environment construction: Use MATLAB / Simulink simulation software to build a simulation environment for water treatment equipment. This environment can be used to simulate the actual operating status of the equipment, including water flow dynamics, water quality changes, equipment energy consumption, and equipment failures. Step 2: Simulation case design: In the constructed simulation environment, design a variety of test cases to cover various situations that may be encountered in actual operations, including treatment effects under different water quality conditions, operation strategies for different equipment combinations, optimization solutions under different energy consumption constraints, and emergency treatment in the event of equipment failure; Step 3, simulation data generation: Run the simulation environment and generate simulation data according to the designed test cases, including the state changes of the equipment, the dynamic changes of water quality indicators, the energy consumption records of the equipment and the operation records of the equipment; Step 4: Compare model predictions with simulation results: Input simulation data into the trained deep Q network model to obtain the operation strategy predicted by the model, and simulate the strategy in the simulation environment to obtain the model's prediction results; at the same time, simulate the traditional control method under the same conditions to obtain the simulation results of the traditional method; Step 5: Performance index evaluation: Compare the model prediction results with the traditional method simulation results in terms of multiple performance indicators, including the stability of water quality compliance, energy saving, equipment life, system response speed, and equipment failure handling efficiency.

10. A water treatment equipment automation equipment control method according to claim 9, characterized in that: The method of field application verification of the intelligent control model includes: based on simulation evaluation, selecting some typical water treatment equipment or scenarios for field application verification, deploying the control strategy of the deep Q network model to the actual equipment, conducting long-term tracking monitoring, collecting actual operation data, and further verifying the actual application effect of the model.

Images