A control method and device for a waterworks sedimentation tank, an electronic device, and a storage medium

Abstract

This invention relates to the field of data processing technology, and in particular to a control method, device, electronic equipment, and storage medium for a sedimentation tank in a water treatment plant. The method includes collecting raw water quality parameters through online sensors; inputting the raw water quality parameters into a prediction module; and outputting a predicted coagulant dosage. The embodiments of this application achieve accurate prediction and dynamic adjustment of coagulant dosage through multi-parameter acquisition and algorithmic analysis, effectively reducing chemical consumption and operating costs. Based on dual threshold triggering of sludge level and dry sludge quantity, and feedback adjustment of sludge discharge concentration, it achieves on-demand, efficient, and precise sludge discharge, significantly improving effluent stability. Furthermore, by feeding back effluent turbidity to the dosing unit, a "coagulation-sedimentation" synergistic closed-loop optimization mechanism is constructed, overcoming the defects of process disconnection, response lag, and extensive operation in traditional control at the system level, comprehensively improving water treatment efficiency and economy.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a control method, device, electronic equipment, and storage medium for a sedimentation tank in a water supply plant. Background Technology

[0002] In water treatment plants, coagulation and sedimentation are the core sequential processes for removing suspended solids, colloidal particles, and algae from raw water. First, coagulants are added to cause impurities in the water to form flocs, which are then separated by sedimentation in a settling tank. In traditional water treatment plants, coagulant dosage often relies on beaker tests or manual experience, making it impossible to dynamically adjust based on raw water quality (such as flow rate and turbidity fluctuations). This easily leads to problems such as overdosing (increasing chemical consumption and sludge production) or underdosing (exceeding effluent turbidity standards).

[0003] Currently, the coagulants and sludge removal methods used in sedimentation treatment processes mainly fall into two categories. The first relies entirely on manual operation, where technicians observe the operation of the coagulation sedimentation tank based on experience and manually open or close the sludge pump or valve according to the flocculation situation. This manual control method is not only inefficient and labor-intensive, but also cannot accurately determine the sludge removal time based solely on experience. The second method is single automatic control, which involves setting a time program to add coagulant at regular intervals and controlling the opening and closing of the sludge pump or valve to achieve periodic automatic sludge removal. However, coagulant addition and sedimentation tank sludge removal are often controlled as independent processes without a synergistic optimization mechanism. Adjustments to the coagulant dosage dosing do not consider the subsequent sedimentation tank's sludge load and sedimentation capacity, leading to increased sludge accumulation when the dosage is too high. Furthermore, feedback on the sedimentation tank's sludge removal effect does not provide feedback for optimizing the coagulant dosage. When incomplete sludge removal leads to excessive effluent turbidity, it is impossible to adjust the coagulant dosage parameters in a timely manner to improve floc formation. This disconnect in the process leads to low operating efficiency of the entire coagulation-sedimentation system, making it difficult to achieve the dual goals of meeting water quality standards and controlling costs. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a control method, device, electronic equipment, and storage medium for sedimentation tanks in water supply plants.

[0005] In a first aspect, embodiments of the present invention provide a method for controlling a sedimentation tank in a water supply plant, the method comprising:

[0006] Raw water quality parameters are collected using online sensors;

[0007] The raw water quality parameters are input into the prediction module, which outputs the predicted coagulant dosage. The prediction module includes a sequentially connected feature extraction module and a BP neural network optimized by the particle swarm optimization algorithm.

[0008] Obtain the current sludge level and dry sludge volume in the sedimentation tank after coagulant is added based on the predicted coagulant dosage;

[0009] When a certain amount of sludge reaches a preset threshold, and / or the current sludge level reaches a preset position, the sludge discharge mode is triggered;

[0010] In sludge discharge mode, the sludge discharge execution time is adjusted based on real-time sludge concentration feedback, and the predicted coagulant dosage is adjusted based on the turbidity of the sedimentation tank effluent.

[0011] In conjunction with the first aspect, the feature extraction module includes a correlation screening unit and a feature decomposition unit;

[0012] The steps for inputting raw water quality parameters into the prediction module and outputting the predicted coagulant dosage include:

[0013] The raw water quality parameters were input into the correlation screening unit for correlation analysis and screening, and several parameters that were significantly related to the amount of coagulant added were obtained.

[0014] Multiple parameters are input into the feature decomposition unit, and feature extraction is performed through the adaptive mode decomposition algorithm to obtain the feature vector;

[0015] The feature vector is input into a BP neural network optimized by particle swarm optimization, and the output is the predicted amount of coagulant to be added.

[0016] In conjunction with the first aspect, the BP neural network optimized by the particle swarm optimization algorithm uses the average of the square roots of the mean square errors of the training and test sets as the fitness function to iteratively optimize the parameters of the BP neural network.

[0017] In conjunction with the first aspect, the feature extraction module also includes a preprocessing unit;

[0018] Before the step of inputting multiple parameters into the feature decomposition unit and extracting features using the adaptive mode decomposition algorithm to obtain the feature vector, the following steps are also included:

[0019] The raw water quality parameters are input into the pretreatment unit for normalization to eliminate dimensional differences.

[0020] In conjunction with the first aspect, the steps for obtaining the current sludge level and dry sludge quantity in the sedimentation tank after coagulant dosing based on the predicted coagulant dosage include:

[0021] Obtain the operating parameters of the sedimentation tank after adding coagulant based on the predicted coagulant dosage. The operating parameters include at least: current sludge level, current influent flow rate, current floc turbidity, current sludge density, and current sludge moisture content.

[0022] Calculate the amount of dry sludge based on operating parameters.

[0023] In conjunction with the first aspect, the steps for adjusting the sludge discharge execution time based on real-time sludge concentration feedback include:

[0024] Obtain the current sludge concentration during the sludge discharge process;

[0025] Based on the comparison between the current sludge concentration and the preset threshold, the sludge discharge conditions are determined.

[0026] Based on the preset correspondence, the adjustment time corresponding to the sludge discharge condition is determined;

[0027] The execution duration of the sewage discharge mode is adjusted based on the time adjustment.

[0028] In conjunction with the first aspect, the steps for adjusting and predicting coagulant dosage based on effluent turbidity include:

[0029] If the effluent turbidity is greater than the turbidity threshold, the current raw water quality parameters are input into the first prediction module to output the current predicted dosage of coagulant.

[0030] The coagulant dosing unit is controlled to operate at the current predicted coagulant dosing amount.

[0031] Secondly, embodiments of this application provide a control device for a sedimentation tank in a water supply plant, the device comprising:

[0032] The data acquisition module is used to collect raw water quality parameters through online sensors;

[0033] The prediction module is used to input raw water quality parameters and output the predicted coagulant dosage. The prediction module includes a sequentially connected feature extraction module and a BP neural network optimized by particle swarm optimization algorithm.

[0034] The acquisition module is used to acquire the current sludge level and dry sludge volume of the sedimentation tank after coagulant is added according to the predicted coagulant dosage.

[0035] The trigger module is used to trigger the sludge discharge mode when the dry sludge volume reaches a preset threshold and / or the current sludge level reaches a preset position.

[0036] The adjustment module is used to adjust the sludge discharge execution time based on real-time sludge concentration feedback in sludge discharge mode, and to adjust the predicted coagulant dosage based on the turbidity of the sedimentation tank effluent.

[0037] Thirdly, this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor runs the computer program to cause the electronic device to perform the above-described method.

[0038] Fourthly, this application provides a readable storage medium storing computer program instructions, which are read and executed by a processor to perform the above-described method.

[0039] The embodiments of the present invention bring the following beneficial effects: The present application provides a control method, device, electronic equipment, and storage medium for a sedimentation tank in a water supply plant. The method includes collecting raw water quality parameters through online sensors; inputting the raw water quality parameters into a prediction module and outputting a predicted coagulant dosage; the prediction module includes a sequentially connected feature extraction module and a BP neural network optimized by particle swarm optimization algorithm; obtaining the current sludge level and dry sludge volume of the sedimentation tank after adding coagulant at the predicted dosage; triggering a sludge discharge mode when a certain sludge volume reaches a preset threshold and / or the current sludge level reaches a preset position; in the sludge discharge mode, adjusting the sludge discharge execution time according to real-time sludge concentration feedback, and adjusting the predicted coagulant dosage based on the turbidity of the sedimentation tank effluent.

[0040] This application's embodiments achieve accurate prediction and dynamic adjustment of coagulant dosing through multi-parameter acquisition and energy algorithms, effectively reducing chemical consumption and operating costs; based on dual threshold triggering of sludge level and dry sludge quantity and feedback adjustment of sludge discharge concentration, it achieves on-demand, efficient, and precise sludge discharge, significantly improving effluent stability; and by feeding back effluent turbidity to the dosing unit, it constructs a "coagulation-sedimentation" synergistic closed-loop optimization mechanism, overcoming the defects of process disconnection, response lag, and extensive operation in traditional control at the system level, comprehensively improving water treatment efficiency and economy.

[0041] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention are realized and obtained in accordance with the structures particularly pointed out in the description, claims and drawings.

[0042] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0043] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0044] Figure 1 A schematic flowchart illustrating the control method for a sedimentation tank in a water supply plant provided in an embodiment of the present invention;

[0045] Figure 2 This is a schematic diagram of the control device for a sedimentation tank in a water supply plant provided in an embodiment of the present invention;

[0046] Figure 3 This is a schematic diagram of the electronic device structure provided in an embodiment of the present invention.

[0047] Figure label:

[0048] 10 - Acquisition module, 20 - Prediction module, 30 - Acquisition module, 40 - Triggering module, 50 - Adjustment module;

[0049] 130 - Processor, 131 - Memory, 132 - Bus, 133 - Communication interface. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions 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, 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.

[0051] To facilitate understanding of this embodiment, the technical terms used in this application will be briefly introduced below.

[0052] In the field of water treatment, coagulants are a class of chemical agents used to purify water. Their core function is to "pull" together tiny suspended particles and colloidal impurities in the water that are difficult to remove by natural sedimentation through electrical neutralization and adsorption bridging, and agglomerate them into larger, easily settled flocculent particles (commonly known as "lump flocs"). Common coagulants include aluminum salts (such as polyaluminum chloride (PAC), aluminum sulfate, etc.) and iron salts (such as polyferric sulfate (PFS), ferric chloride, etc.).

[0053] PSO, or Particle Swarm Optimization, is a swarm intelligence optimization algorithm that simulates the social behavior of flocks of birds or schools of fish in nature.

[0054] After introducing the technical terms used in this application, the application scenarios and design concepts of the embodiments of this application will be briefly described below.

[0055] Coagulant addition and sedimentation tank sludge removal are often controlled as independent processes without establishing a synergistic optimization mechanism between the two. This results in low operating efficiency of the entire coagulation-sedimentation system, making it difficult to achieve the dual goals of water quality compliance and cost control.

[0056] Based on this, this application provides a control method, device, electronic equipment, and storage medium for a sedimentation tank in a water supply plant. The method uses intelligent algorithms to achieve precise coagulant dosing and combines a multi-parameter sludge discharge model to achieve on-demand sludge discharge, forming a closed-loop intelligent control that effectively improves the stability of effluent water quality and operational efficiency.

[0057] Example 1

[0058] This application provides a control method for sedimentation tanks in water treatment plants, combined with... Figure 1 As shown, the method includes:

[0059] S110 collects raw water quality parameters through online sensors.

[0060] S120 inputs raw water quality parameters into the prediction module and outputs the predicted coagulant dosage; the prediction module includes a sequentially connected feature extraction module and a BP neural network optimized by particle swarm optimization algorithm.

[0061] S130, obtain the current sludge level and dry sludge volume of the sedimentation tank after coagulant addition based on the predicted coagulant dosage.

[0062] S140, when a certain amount of sludge reaches a preset threshold, and / or when the current sludge level reaches a preset position, the sludge discharge mode is triggered.

[0063] In sludge discharge mode, the S150 adjusts the sludge discharge execution time based on real-time sludge concentration feedback and adjusts the predicted coagulant dosage based on the turbidity of the sedimentation tank effluent.

[0064] This invention uses online sensors to collect multi-dimensional raw water quality parameters in real time and inputs them based on adaptive mode decomposition and PSO. The BP neural network prediction model enables accurate dynamic prediction of coagulant dosage, effectively reducing chemical consumption and operating costs. Furthermore, it combines sludge level and dry sludge quantity as dual thresholds to trigger sludge discharge, and adjusts the discharge time based on real-time sludge concentration feedback, improving discharge efficiency and effluent stability. Simultaneously, it feeds back the turbidity of the sedimentation tank effluent to the dosing unit, forming a "coagulation" effect. The "sedimentation" closed-loop optimization overcomes the problems of process disconnection, response lag and extensive operation in traditional control as a whole, and significantly improves system synergy and economy while ensuring water quality meets standards.

[0065] In step S110, online sensors are deployed at the inlet pipe or channel of the sedimentation tank to capture the original influent water quality entering the sedimentation tank completely and in real time before the coagulation reaction occurs, providing a timely and accurate data source for subsequent intelligent dosing and sludge discharge control. Specifically, these sensors can include flow sensors, turbidity sensors, COD sensors, pH sensors, ammonia nitrogen sensors, and water temperature sensors. Other types of sensors may also be used in practical applications, but will not be elaborated upon here.

[0066] The flow sensor typically has a range of 0-5000m. 3The influent flow rate (Q) is monitored by a sensor with a range of 0-100 NTU. This turbidity level directly reflects the concentration of suspended solids and colloids, and is a key factor in determining the dosage of coagulant. The COD sensor, typically with a range of 0-20 mg / L, assesses the oxygen consumption of organic matter in the water to determine the organic pollution load of the raw water. The pH sensor, with a range of 0-14, accurately measures the acidity or alkalinity of the water. This pH value significantly affects the hydrolysis form and flocculation effect of the coagulant. The ammonia nitrogen sensor monitors the ammonia nitrogen content, and its concentration changes affect the coagulation process and may indicate sudden changes in water quality. The water temperature sensor, typically with a range of 0-50℃, detects the water temperature. Water temperature directly affects the hydrolysis and floc formation rate of the coagulant by influencing the viscosity coefficient and chemical reaction rate of the water.

[0067] In conjunction with the first aspect, the feature extraction module includes a correlation screening unit and a feature decomposition unit.

[0068] Step S120 includes:

[0069] S121, input the raw water quality parameters into the correlation screening unit for correlation analysis and screening to obtain multiple parameters that are significantly related to the amount of coagulant added.

[0070] Understandably, the correlation screening unit is used for preliminary dimensionality reduction and screening of parameters. First, this unit uses Pearson correlation analysis (e.g., Pearson) to calculate the linear correlation between each raw water quality parameter collected in step S110 (such as flow rate, turbidity, COD, pH, ammonia nitrogen, water temperature, etc.) and the final target variable (i.e., coagulant dosage), and obtains its correlation coefficient. Subsequently, the calculated correlation coefficient is subjected to a statistical significance test (two-sided confidence level ≤ 0.01) to determine whether the observed correlation is statistically significant and not caused by accidental factors. Finally, based on the preset significance criteria, the unit automatically screens out the key water quality parameters (such as flow rate, turbidity, COD, ammonia nitrogen, pH, water temperature, etc.) that are significantly correlated with the coagulant dosage, and outputs them as effective features. The fundamental purpose of this unit is to achieve preliminary intelligent dimensionality reduction of the input data. It uses rigorous mathematical statistical methods to identify and retain the core driving factors that truly contribute to the prediction of coagulant dosage from numerous available parameters, while eliminating irrelevant or redundant parameters. This reduces the data dimensionality and noise that subsequent complex algorithms (such as VMD and PSO-BP neural networks) need to process, speeds up model training and prediction, avoids interference from irrelevant variables, and allows the model to focus more on key influencing factors, thereby improving the accuracy and stability of predictions. Optimization is achieved from the data source, eliminating the need to configure expensive sensors or processing resources for unimportant parameters.

[0071] S122: Input multiple parameters into the feature decomposition unit, and extract features through the adaptive mode decomposition algorithm to obtain feature vectors.

[0072] After correlation screening, multiple key water quality parameters (time series data) are input into the feature decomposition unit. An adaptive mode decomposition algorithm (such as VMD, variational mode decomposition) is used to perform deep and adaptive signal decomposition and reconstruction on the data sequence of each parameter, so as to transform the initially screened multidimensional water quality parameters into high-value feature vectors that are more suitable for intelligent prediction.

[0073] S123 inputs the feature vector into a BP neural network optimized by the particle swarm optimization algorithm, and outputs the predicted amount of coagulant to be added.

[0074] The high-dimensional feature vector obtained in the previous step S122, which has undergone deep processing and optimization, is input into a pre-trained, high-performance prediction model to finally calculate the accurate amount of coagulant to be added.

[0075] Step S123 is the decision-making terminal of the entire intelligent dosing process. Its core is to map the deeply processed feature information into precise operational commands through a pre-optimized, globally optimized, high-performance prediction model. This process relies on the PSO-BP algorithm mechanism and specifically includes the following two levels:

[0076] First, the input to the model is a high-value feature vector obtained in step S122. That is, the input to the model is not the original water quality parameters, but a multi-dimensional (28-dimensional in this embodiment) feature vector generated by the adaptive modality number optimization algorithm. This vector is the result of denoising the original data and extracting the core features of trends and fluctuations. It needs to be processed by max-min normalization to eliminate scale differences, thus providing the model with high-quality and standardized input.

[0077] Second, the forward prediction of the PSO-optimized BP neural network is an "offline optimization, online application" process. Prior to this, a PSO-BP model architecture is constructed, and the BP parameters are optimized using the PSO algorithm. Specifically:

[0078] The first parameter encoding: BP weights / biases are mapped to PSO particles; the parameters to be optimized in the BP neural network (input layer → hidden layer weights, hidden layer → output layer weights, hidden layer biases, output layer biases) are encoded as the position vectors of PSO particles; input layer → hidden layer weights: 28 × 6 = 168 parameters; hidden layer → output layer weights: 6 × 1 = 6 parameters; hidden layer biases: 6 parameters; output layer biases: 1 parameter; particle position vector dimension: 168 + 6 + 6 + 1 = 181, that is, each particle corresponds to one set of BP weights and biases, and the position range is set to [-3, 3].

[0079] Secondly, the PSO initialization parameter settings are as follows: Particle swarm size: 60; Maximum number of iterations: 90; Inertia weight ω: A linear decreasing strategy is adopted, with an initial ω... init =0.9, at the end of the iteration; learning factor: c1=c2=2 (classic setting, guiding particles to update towards individual and global optima); particle velocity range: [-0.5, 0.5] (balancing search efficiency and convergence stability).

[0080] In this way, the 181 parameters to be optimized in the BP network (including 168 input-hidden layer weights, 6 hidden-output layer weights, 6 hidden layer biases, and 1 output layer bias) are encoded as the positions of PSO particles. A population of 60 particles is initialized, and evolutionary rules are set (maximum 90 generations, decreasing inertia weights, etc.).

[0081] Then, model training is performed. The PSO algorithm uses the average RMSE of the training and test sets as the fitness function to drive the particle swarm search iteratively. After a maximum of 90 generations of evolution, the algorithm converges and outputs a globally optimal particle position. This position, after decoding, corresponds to a set of optimal BP network parameter combinations. This set of parameters allows the network to achieve the best balance between accuracy and generalization ability.

[0082] Finally, online prediction is performed. During actual operation, the system directly inputs the 28-dimensional normalized feature vector obtained in step S122 into this optimized BP neural network with fixed structure and parameters. The network performs efficient forward propagation calculations: the data is transformed nonlinearly from the input layer (28 nodes) through the hidden layer (6 nodes), and finally a normalized prediction value is generated in the output layer (1 node). This prediction value is then denormalized and converted into a coagulant dosage (in mg / L) with actual physical meaning.

[0083] Based on the above, this application embodiment places the complex global parameter optimization process in the offline stage to ensure the high speed and stability of the online prediction stage. Through the swarm intelligence and global search of PSO, a high starting parameter set that is far superior to that of random initialization is found for the BP network, which directly determines the accuracy, stability and generalization ability of the final prediction model.

[0084] The role of coagulants begins in the coagulation tank, but their effect is ultimately manifested and consolidated in the sedimentation tank. In the coagulation tank (reaction tank), the coagulant added mixes rapidly with the raw water, neutralizing the negative charge on the surface of colloidal particles and disrupting the repulsive forces that stabilize and disperse them (i.e., "destabilization"). The destabilized particles collide with each other through Brownian motion and, under the bridging effect of the coagulant, combine to form tiny flocs. The water carrying these tiny flocs slowly flows into the sedimentation tank. Here, the coagulant's effect continues; the micro-flocs continue to collide and combine in the slow water flow, growing into dense, coarse flocs. Due to their significant weight, these coarse flocs naturally settle to the bottom of the sedimentation tank under static or slow-flowing conditions, thus separating from the clear water. The sludge at the bottom of the tank is ultimately discharged through the sludge removal system.

[0085] In conjunction with the first aspect, step S122 includes:

[0086] S1221 performs adaptive mode number optimization decomposition on multiple input parameters to obtain K modal components for each parameter.

[0087] This step is fundamental to feature extraction. The water quality fluctuation characteristics (such as varying frequency components) of raw water from different times and sources differ. Therefore, the crude approach of fixing the number of modes in traditional signal decomposition methods is abandoned. Instead, a pre-defined search range (K∈[k1, k2]) is used, and the penalty parameter α=180 and the convergence criterion ε=10 are initialized. -7 And key parameters such as the maximum number of iterations T=1000. For each candidate K value, variational mode decomposition is performed, iteratively updating the modal components, center frequency, and Lagrange multiplier in the frequency domain until the convergence condition is met. Specifically:

[0088] First, we set K modal components u k and its center frequency ω k Initialize the Lagrange multiplier λ and the iteration counter n=0.

[0089] Next, iterative updates are performed. The first step is to update the modal components by solving a variational problem in the frequency domain, using the following formula:

[0090] ;

[0091] in, The Fourier transform of the k-th modal component in the (n+1)-th iteration; The Fourier transform of the raw water quality time series; It is the sum of the Fourier transforms of all modal components except the k-th component; Fourier transform of Lagrange multipliers (used for constraint optimization); ω Angular frequency (frequency domain variable); ω k For the first k The center frequency of each modal component.

[0092] Second: The center frequency is updated.

[0093] ;

[0094] in, The center frequency of the k-th modal component in the (n+1)-th iteration; The modulus square of the Fourier transform of the k-th modal component (energy spectral density); the numerator is the frequency-weighted energy integral (reflecting the frequency position of energy concentration); the denominator is the total energy integral (normalized).

[0095] Third, the Lagrange multiplier is used for updating:

[0096] ;

[0097] in, To update the step size (in this embodiment, the value is 0.001);

[0098] For the (n+1)th iteration, the Lagrange multiplier Fourier transform is used.

[0099] This is the reconstruction error (the difference between the original signal and the sum of all modal components).

[0100] Finally, convergence is checked:

[0101] If satisfied Then stop the iteration and output K modal components {u1, u2, ... u3}. K};

[0102] No, not satisfied. If n = n + 1, then return to the previous step and continue iterating and updating.

[0103] in, The L2 norm squared (difference) of the two iteration results of the kth modal component; The L2 norm squared (energy baseline) of the current iteration result of the k-th modal component.

[0104] This process can automatically find the most suitable decomposition scale (i.e., the optimal number of modes K) for the current input water quality data sequence, and decompose it into K modal components from low frequency to high frequency that can reflect the inherent fluctuation law of the raw water, ensuring that the decomposition result can be "tailor-made" to match the current water quality status.

[0105] S1222, calculate the sample entropy and kurtosis of each modal component, and select the optimal number of modes Kopt based on the weighted comprehensive evaluation index.

[0106] Specifically, for each K value, a comprehensive evaluation index is calculated based on the decomposition results.

[0107] First, calculate the sample entropy (SE). k ):

[0108] For the k-th modal component u k Construct an m-dimensional embedding vector: (Xi) = [u k (i), u k (i+1)...u k [(i+m-1)] (m=2); where (Xi) is the i-th m-dimensional embedding vector (composed of m consecutive data points of the modal components); m is the embedding dimension (in this embodiment, the value is 2, used to reflect the local correlation of the signal);

[0109] Define the distance between vectors: d[X] i X j ]=max|X i (l)-X j (l)|(l=1,2...m;where, For vectors and The maximum distance between them (used to measure similarity).

[0110] Calculate the similarity vector ratio: ;

[0111] Where, r = 0.2 × std(u k );Θ is the step function.

[0112] Sample entropy: ;

[0113] in, The sample entropy of the k-th modal component (the smaller the value, the more regular the signal).

[0114] Average sample entropy: ;in, The average sample entropy (overall regularity index) for all modal components.

[0115] Then, kurtosis is calculated:

[0116] Kurtosis of the k-th modal component: ;

[0117] in, is the kurtosis of the k-th modal component (used to measure the degree to which the signal deviates from the normal distribution; the larger the value, the more significant the peak). It is the expectation of the fourth power of the difference between the k-th modal component and the mean (fourth central moment). Let be the variance (second central moment) of the k-th modal component; u be the mean of the k-th modal component; and E be the expectation.

[0118] Average kurtosis: ;

[0119] in, : The average kurtosis of all modal components.

[0120] Finally, the comprehensive evaluation index is calculated: a weighted summation is used to construct the objective function (where the weights are determined using the analytic hierarchy process):

[0121] F(K) = 0.3 × 0.7× (in, The smaller the value, the more stable the feature. The larger the value, the better it can capture mutations.

[0122] This step addresses the crucial issue of automatically determining the optimal K value. It introduces two complementary metrics: sample entropy (measuring the complexity and randomness of a sequence) and kurtosis (measuring the peak of a sequence distribution and its sensitivity to abrupt changes), constructing a weighted comprehensive evaluation function F(K). By iterating through candidate K values ​​and calculating their F(K), the K value that optimizes the comprehensive evaluation function is automatically selected as Kopt. This mechanism intelligently balances the signal's regularity (SE) with its ability to capture abrupt changes (Kurt), ensuring that the modal components decomposed by the selected Kopt are most beneficial for subsequent predictive modeling.

[0123] S1223, retain the top N key components with the highest energy percentage among the Kopt modal components.

[0124] This step is crucial for achieving data dimensionality reduction and noise reduction. Since not all decomposed modal components are equally important, based on the signal processing principle that "most effective information is concentrated in a few key components," only the top N components with the highest energy percentage are retained from the Kopt components (in this embodiment, these are taken as 4 key components (e.g., energy percentage ≥ 95%)). This effectively filters out high-frequency components with weak energy that usually represent noise or irrelevant details. While retaining core information, it significantly reduces the computational load of subsequent neural network models, improving model efficiency and robustness.

[0125] S1224: Sort the key components according to their center frequencies and concatenate them to form a feature vector.

[0126] This step involves organizing the processed signal components into an input format usable by the model, namely, a feature vector. The retained key components are sorted and concatenated according to their center frequencies from low to high, i.e., concatenated in the order of "low-frequency trend components to high-frequency fluctuation components." Low-frequency components represent the long-term trends of water quality parameters, while mid-to-high-frequency components represent periodic fluctuations and short-term changes. The feature vectors formed by concatenating in this order have a clear structure and strong regularity, making them easier for the subsequent PSO-BP neural network model to learn and map, thus directly contributing to the improvement of the final coagulant dosage prediction accuracy.

[0127] In conjunction with the first aspect, the BP neural network optimized by the particle swarm optimization algorithm uses the average of the square roots of the mean square errors of the training and test sets as the fitness function to iteratively optimize the parameters of the BP neural network.

[0128] The specific formula is as follows:

[0129] ;

[0130] Among them: MSE train The mean squared error (MSE) of the BP neural network on the training set (which in this embodiment accounts for 70% of the total data, totaling 183 × 70% ≈ 128 groups); test The mean squared error is calculated on the test set (which accounts for 30% of the total data in this embodiment, comprising 55 groups).

[0131] ;

[0132] in, This represents the actual amount of coagulant added. is the predicted value of the BP neural network, and n is the number of samples.

[0133] The specific working mechanism and iterative optimization process are as follows:

[0134] First, initialization is performed by randomly generating the initial positions (weight / bias combination) and velocities of 60 particles. The position of each particle is decoded into BP neural network parameters, and its fitness value on the training and test sets is calculated. The individual optimal position (pbest) and global optimal position (gbest) of each particle are recorded.

[0135] Next, the velocity and position are updated. For each generation of particles, the velocity and position are updated according to the following formula:

[0136] ;

[0137] ;

[0138] in, Let be the velocity of the i-th particle in the d-th dimension at the t-th iteration; For the corresponding position; , These are random numbers, with values ​​ranging from [0, 1]. Let d be the optimal position of the i-th particle in the d-th dimension. The d-th dimension represents the globally optimal position;

[0139] If a particle's position exceeds [-3,3] or its velocity exceeds [-0.5,0.5], it will be forcibly pulled back to the boundary range.

[0140] If the number of iterations reaches 90, or the global optimal fitness value changes by less than 10 over 10 consecutive iterations. -7 If the iteration stops, output the BP weights and biases corresponding to gbest.

[0141] In other words, each "particle" in the PSO algorithm represents a potential solution of a set of weights and biases in a BP neural network. In each generation of evolution, each particle constructs a temporary BP neural network based on its position (i.e., a set of network parameters) and performs forward computation on both the training and test sets to obtain the prediction error. Subsequently, the system calculates the fitness value of the particle according to the aforementioned fitness function. This value comprehensively reflects the combined performance of the set of parameters in both "learning historical patterns" (training set) and "coping with unknown situations" (test set). Based on this fitness value, the PSO algorithm guides the entire particle swarm to collaboratively search and evolve in the direction of minimizing the comprehensive error. After multiple iterations (e.g., 90 generations), it eventually converges to the globally optimal or suboptimal parameter combination.

[0142] Understandably, if only the training set error is used as the optimization target, the model is prone to overfitting to the details and noise in the training data, leading to a significant drop in prediction accuracy when facing new test data (real-world application scenarios). In this embodiment, to simultaneously improve the model's accuracy (low error) and generalization ability (strong robustness), this method innovatively adopts a composite fitness function as the sole criterion for evaluating the quality of each "particle" (i.e., a set of candidate BP network parameters) in the PSO algorithm. This forces the PSO algorithm not to only look for parameters that perform exceptionally well on the training set (potentially overfitting), but to look for "robust" parameters that maintain low errors on both the training and test sets. This is equivalent to embedding a "generalization ability detector" in the optimization process. The BP neural network optimized through this mechanism inherently possesses good generalization characteristics. When the model is deployed in a real water treatment system, it can still make more reliable and stable predictions for fluctuating new influent water quality that has not appeared in the training set, thus ensuring the long-term reliability of the intelligent dosing system.

[0143] In addition to the first aspect, the feature extraction module also includes a preprocessing unit.

[0144] Before step S122, the following are also included:

[0145] S1220 inputs the raw water quality parameters into the pretreatment unit for normalization to eliminate dimensional differences.

[0146] Understandably, preprocessing is performed before data flows into the intelligent prediction model to address the difficulties in model training and prediction bias caused by the different dimensions and numerical ranges of multiple water quality parameters from different sources. Specifically, multiple raw water quality parameters (such as flow rate, turbidity, COD, ammonia nitrogen, pH, and water temperature) that are significantly related to the coagulant dosage and selected through step S121 are preprocessed. The flow rate unit may be "m³". 3 The unit for turbidity is "NTU" while the unit for COD is "mg / L," and their numerical ranges differ significantly. Without processing, parameters with larger values ​​(such as flow rate) will dominate the model calculations, while the influence of parameters with smaller values ​​(such as pH) may be overwhelmed. By using a max-min normalization algorithm, the values ​​of all these parameters are linearly mapped to the [0,1] interval. The normalized data distribution is more uniform, and the unified data scale can significantly improve the convergence speed of the subsequent PSO-BP neural network during training, making feature extraction more accurate, weight updates more reasonable, and avoiding model oscillations due to inconsistent parameter scales during optimization, thus helping the algorithm find the optimal solution faster and more stably.

[0147] The calculation formula for the max-min normalization algorithm is:

[0148]

[0149] in, These are the original values ​​of the water quality parameters; This represents the minimum value of the original water quality parameters; This represents the maximum value of the original water quality parameters; To map the original values ​​of water quality parameters to values ​​in the interval [0, 1].

[0150] In conjunction with the first aspect, step S130 includes:

[0151] S131, obtain the operating parameters of the sedimentation tank after adding coagulant based on the predicted coagulant dosage. The operating parameters include at least: current sludge level, current influent flow rate, current floc turbidity, current sludge density, and current sludge moisture content.

[0152] S132, calculate the amount of dry sludge based on operating parameters.

[0153] After obtaining the predicted coagulant dosage based on steps S110-S130, the variable frequency metering pump of the coagulant dosing execution module adds coagulant according to the predicted value (i.e., the predicted coagulant dosage). The suspended solids and colloidal particles in the raw water form dense flocs, which enter the sedimentation tank to wait for settling. The formed flocs gradually settle after entering the sedimentation tank, forming a bottom sludge layer.

[0154] The sludge level sensor installed at the bottom of the sedimentation tank monitors the sludge accumulation height in real time, the influent turbidity sensor monitors the turbidity NTU of the floc water entering the sedimentation tank, the influent flow sensor monitors the influent flow rate Q, and the sludge concentration detection device is on standby. The "current sludge level", "current floc water turbidity" and "current influent flow rate" obtained here are directly affected by the execution effect of "predicting the amount of coagulant added" in step S120.

[0155] Subsequently, step S132 integrates the acquired multi-source and multi-dimensional operating parameters through a mechanism model, transforming them into a key decision indicator that can scientifically and objectively characterize sludge load.

[0156] Specifically, the Total Dry Sludge Volume (TDS) model is used for calculation, and the formula is as follows:

[0157] TDS total =Q(T×E1+A×E2×K pH +(COD×E3+NH4 + ×E4)C×K org ×10 -6 ;

[0158] Where TDS is the total dry sludge volume, t / h; Q is the influent volume, m³ / h. 3 / h; T is the raw water turbidity, NTU; E1 is the conversion factor between turbidity and SS, 0.7~2.2; A is the aluminum salt coagulant injection rate, mg / L; E2 is the conversion factor between Al2O3 and Al(OH)3, taken as 1.53, K pH E1 represents the correction factor, categorized according to the raw water pH. When pH = 6.5~7.5 (optimal reaction range), the value is 1.0; when pH = 5.5~6.5 or 7.5~8.5, the value is 0.8~0.9; when pH < 5.5 or > 8.5, the value is 0.5~0.7. E3 represents the conversion factor for the dry sludge amount of COD organic matter complexes, experimentally calibrated to 0.08~0.12 (for every 1 mg / L increase in COD, the dry sludge amount increases by 0.08~0.12 mg / L); E4 represents the conversion factor for the dry sludge amount of ammonia nitrogen reaction products, experimentally calibrated to 0.05~0.07 (for every 1 mg / L increase in ammonia nitrogen, the dry sludge amount increases by 0.05~0.07 mg / L); K orgThis represents the correction factor for the impact of organic matter. It is divided according to the COD value: when COD < 5 mg / L, the value is 0.9~1.0; when 5 mg / L ≤ COD < 10 mg / L, the value is 0.8~0.9; when COD ≥ 10 mg / L, the value is 0.6~0.8.

[0159] To address the calculation deviations caused by sensor errors and fluctuations in reagent dosage, and based on actual operational feedback parameters from the sedimentation tank (effluent turbidity, sludge moisture content), the basic dry sludge quantity was adjusted. Dynamic adjustments are made to form a two-stage model of "theoretical calculation + feedback adjustment", as shown in the following formula:

[0160]

[0161]

[0162] in, This represents the turbidity correction factor for the effluent. This represents the measured value of turbidity in the effluent from the sedimentation tank. Indicates the set value for effluent turbidity; The value ranges from 0 to 0.1; when ≤ hour, =0.

[0163] This represents the sludge moisture content correction factor, with a value range of 0~0.08 (when...). ≤ hour, =0); Indicates the actual moisture content of the sludge (%, collected by an online moisture content sensor); This indicates the design moisture content of the sludge (generally between 99.0% and 99.5%).

[0164] In this way, by comprehensively considering the two major sources of sludge—suspended solids from raw water turbidity and chemical sludge generated by coagulant addition—and combining this with real-time influent flow rate, the absolute dry sludge production in tons per hour can be calculated. This is more scientific and accurate in principle than relying solely on "sludge level" or "time" to determine sludge discharge requirements.

[0165] Understandably, as more and more sludge settles, the accumulation of sludge is large when either of the following conditions is met: "dry sludge volume reaches the preset threshold" or "current sludge level reaches the preset position". At this time, the sludge discharge mode is triggered.

[0166] Specifically, after step S130, the amount of dry sludge in the flocculation sedimentation tank is estimated. Assuming the sludge moisture content and sludge density in the flocculation solid-liquid separation unit... (g / m 3 Then, the sludge production at the bottom of the solid-liquid separation unit... The calculation formula is as follows:

[0167] .

[0168] In this embodiment, when the sludge level detection sensor shows that the sludge accumulation height H reaches the upper limit threshold Hset (the threshold depends on the actual situation), or when the sludge production accumulates to 80% of the effective volume of the solid-liquid separation unit tank, the control unit triggers a sludge discharge command and sets the initial sludge discharge duration to 120s.

[0169] The formula for calculating the sludge accumulation height H is as follows:

[0170] ;

[0171] Where H represents the sludge level height in the sedimentation tank (m); t represents the cumulative time of dry sludge (h); ρ represents the sludge density (g / m³); and S represents the cross-sectional area of ​​the sedimentation tank (m²).

[0172] Sludge discharge trigger condition: when H ≥ Hset (sludge level threshold) or ( When the effective volume of the sedimentation tank is reached, sludge discharge is triggered.

[0173] In conjunction with the first aspect, step S150, which involves adjusting the execution duration of the sewage discharge mode in real time based on the sludge concentration feedback under the sewage discharge mode, includes:

[0174] S151, obtain the current sludge concentration during the sludge discharge process.

[0175] The sludge concentration detection device installed on the sludge discharge pipeline monitors the concentration of the discharged sludge mixture in real time and continuously after the sludge discharge command is triggered, and transmits the monitored concentration data (such as pd1, pd2, pd3) to the data processing unit in real time.

[0176] S152, based on the comparison between the current sludge concentration and the preset threshold, determine the sludge discharge condition.

[0177] The system's data processing unit compares the received current sludge concentration with several preset key threshold concentrations to intelligently determine the current stage of sludge discharge. For example, three key thresholds are preset: initial high concentration pd1, concentration decrease threshold pd2, and sludge discharge termination threshold pd3. When the detected concentration is pd1, it is determined to be "initial stage of high-concentration sludge discharge"; when the concentration drops to pd2, it is determined to be "concentration decrease period"; and when the concentration drops to pd3, it is determined to be "low-concentration discharge period".

[0178] S153, based on the preset correspondence, determines the adjustment time corresponding to the sludge discharge condition.

[0179] Understandably, the control unit stores a table corresponding to the sludge discharge condition and the adjustment time. Based on the condition determined in step S152, the system queries this table to directly determine the execution duration of the next sludge discharge step. Referring to the example above, when in the "initial stage of high-concentration sludge discharge," the preset initial sludge discharge duration is 120 seconds; when the system determines that it has entered the "concentration decrease period" (i.e., the concentration drops from pd1 to pd2), it queries the corresponding table and finds that the adjustment time for this condition is 90 seconds, meaning the remaining sludge discharge time needs to be shortened to 90 seconds.

[0180] S154, Adjust the execution duration of the sewage discharge mode based on the adjustment time.

[0181] Based on the adjustment time determined in S153, a new control command is issued to the sludge discharge valve. For example, if the concentration is detected to drop to pd2 in the 50th second, the control unit immediately and dynamically adjusts the remaining opening time of the sludge discharge valve to 90 seconds. Subsequently, if the sludge concentration continues to drop to pd3, a command to close the valve is immediately issued.

[0182] In this way, instead of mechanically running at a fixed time, the sludge discharge operation is intelligently determined based on the "value" (concentration) of the actual discharged material. Sufficient sludge discharge is ensured at high concentrations, and adjustments or termination are made promptly when the concentration drops significantly. This ensures effective sludge discharge while maximizing water conservation and energy savings for subsequent sludge treatment units, resolving the inherent contradiction of traditional timed sludge discharge (i.e., if the discharge time is set too short, sludge will not be completely discharged; if it is set too long, a large amount of clean water will be wasted).

[0183] In conjunction with the first aspect, step S150, which involves adjusting the predicted coagulant dosage based on effluent turbidity, specifically includes:

[0184] S155, if the effluent turbidity is greater than the turbidity threshold, input the current raw water quality parameters into the prediction module to output the current predicted coagulant dosage.

[0185] S156, control the coagulant dosing unit to operate at the current predicted coagulant dosing amount.

[0186] Understandably, using the turbidity of the sedimentation tank effluent as the final performance indicator as the evaluation standard, when the monitored value exceeds the preset value (e.g., 2.0 NTU), it indicates that the current operating state of the entire "coagulation-sedimentation" system has deviated from the optimal range. At this point, an optimization program is immediately triggered to fine-tune the intelligent prediction model based on the feedback information of substandard effluent. By increasing the weight of the "turbidity" input parameter in the model, the model pays more attention to the high turbidity or abnormal turbidity change patterns in the raw water during recalculation. Subsequently, the model combines the latest raw water quality parameters with the adjusted internal weight configuration to perform a new, more targeted forward calculation, thereby outputting a new, feedback-optimized command value, namely the "current predicted coagulant dosage." This new value focuses more on resolving the main problem causing the effluent turbidity to exceed the standard.

[0187] The "current predicted coagulant dosage" output by S155 is sent to the coagulant dosing execution module (such as a variable frequency metering pump) so that the execution module can accurately adjust the dosage to the new set value. The adjusted coagulant mixes thoroughly with the raw water to form flocs with better settling performance. These flocs settle more thoroughly in the sedimentation tank. The direct result is that the turbidity of the effluent from the subsequent sedimentation tank begins to decrease and eventually stabilizes below the preset safety threshold. This result verifies the effectiveness of the feedback adjustment and marks the completion of this collaborative optimization closed loop. Thus, a direct feedback link is established from the end of the process (sedimentation effluent) to the beginning of the process (coagulant dosing), integrating two originally independent units into an organic whole that can respond collaboratively. Through this continuous, small-step, rapid feedback fine-tuning, the system can proactively adapt to the slow time-varying and sudden fluctuations in the raw water quality, thereby continuously ensuring the stable compliance of the effluent quality in long-term operation and avoiding long-term waste of chemicals due to control lag, achieving a balance between stability and economy.

[0188] Secondly, embodiments of this application provide a control method for a sedimentation tank in a water supply plant, combined with... Figure 2 As shown, the device includes: a data acquisition module 10, a prediction module 20, an acquisition module 30, a triggering module 40, and an adjustment module 50.

[0189] The data acquisition module 10 is used to acquire raw water quality parameters through online sensors.

[0190] The prediction module 20 is used to input raw water quality parameters into the prediction module and output the predicted coagulant dosage. The prediction module includes a sequentially connected feature extraction module and a BP neural network optimized by particle swarm optimization algorithm.

[0191] The acquisition module 30 is used to acquire the current sludge level and dry sludge volume of the sedimentation tank after coagulant is added according to the predicted coagulant dosage.

[0192] The trigger module 40 is used to trigger the sludge discharge mode when the dry sludge volume reaches a preset threshold and / or the current sludge level reaches a preset position.

[0193] The adjustment module 50 is used to adjust the sludge discharge execution time based on real-time sludge concentration feedback in sludge discharge mode, and to adjust the predicted coagulant dosage based on the turbidity of the sedimentation tank effluent.

[0194] Thirdly, embodiments of this application provide an electronic device, combined with Figure 3 As shown, the electronic device includes a memory 131 and a processor 130. The memory 131 stores a computer program, and the processor 130 runs the computer program to make the electronic device perform the above-described method.

[0195] Furthermore, combined Figure 3 The electronic device shown also includes a bus 132 and a communication interface 133, with the processor 130, the communication interface 133 and the memory 131 connected via the bus 132.

[0196] The memory 131 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 133 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc. The bus 132 may be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0197] Processor 130 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 130 or by instructions in software form. Processor 130 may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software module can reside in a readily available storage medium in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 131. The processor 130 reads the information from memory 131 and, in conjunction with its hardware, completes the steps of the method described in the foregoing embodiments.

[0198] Fourthly, embodiments of this application provide a readable storage medium storing computer program instructions, which are read and executed by a processor to perform the above-described method.

[0199] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system and apparatus described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0200] Furthermore, in the description of the embodiments of the present invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in the present invention based on the specific circumstances.

[0201] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0202] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0203] Finally, it should be noted that the above embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for controlling a sedimentation tank in a water supply plant, characterized in that, The method includes: Raw water quality parameters are collected using online sensors; The raw water quality parameters are input into the prediction module, which outputs the predicted coagulant dosage. The prediction module includes a sequentially connected feature extraction module and a BP neural network optimized by the particle swarm optimization algorithm. Obtain the operating parameters of the sedimentation tank after adding coagulant at the predicted coagulant dosage. The operating parameters include at least: current sludge level, current influent flow rate, current floc turbidity, current sludge density, and current sludge moisture content. Calculate the dry sludge quantity based on the operating parameters. If the amount of dry sludge reaches a preset threshold, and / or the current sludge level reaches a preset position, the sludge discharge mode is triggered; In the sludge discharge mode, the current sludge concentration during the sludge discharge process is obtained; based on the comparison relationship between the current sludge concentration and a preset threshold, the sludge discharge condition is determined; based on a preset correspondence, the adjustment time corresponding to the sludge discharge condition is determined; based on the adjustment time, the execution duration of the sludge discharge mode is adjusted; and based on the turbidity of the sedimentation tank effluent, the predicted coagulant dosage is adjusted. The formula for calculating the amount of dry mud is: TDS total = Q[T x E1 + A x E2 x K pH + (COD x E3 + NH4 + x E4) C x K org ] x 10 -6 ; Where TDS is the total dry sludge volume, t / h; Q is the influent volume, m³ / h. 3 / h; T is the raw water turbidity, NTU; E1 is the conversion factor between turbidity and SS; A is the aluminum salt coagulant injection rate, mg / L; E2 is the conversion factor between Al2O3 and Al(OH)3, K pH E3 represents the conversion factor for the dry sludge amount of COD organic matter complexes; E4 represents the conversion factor for the dry sludge amount of ammonia nitrogen reaction products; K represents the correction factor. org This indicates the correction factor for the influence of organic matter; The formula for dynamically correcting the amount of dry mud in the foundation is: in, This represents the turbidity correction factor for the effluent. This represents the measured value of turbidity in the effluent from the sedimentation tank. Indicates the set value for effluent turbidity; This represents the correction factor for sludge moisture content; This indicates the actual moisture content of the sludge; This indicates the design moisture content of the sludge.

2. The method according to claim 1, characterized in that, The feature extraction module includes a correlation screening unit and a feature decomposition unit; The steps of inputting the raw water quality parameters into the prediction module and outputting the predicted coagulant dosage include: The raw water quality parameters are input into the correlation screening unit for correlation analysis and screening to obtain multiple parameters that are significantly related to the amount of coagulant added. The multiple parameters are input into the feature decomposition unit, and feature extraction is performed through an adaptive mode decomposition algorithm to obtain a feature vector; The feature vector is input into the BP neural network optimized by the particle swarm optimization algorithm, and the predicted coagulant dosage is output.

3. The method according to claim 2, characterized in that, The BP neural network optimized by the particle swarm optimization algorithm uses the average of the square roots of the mean square errors of the training and test sets as the fitness function to iteratively optimize the parameters of the BP neural network.

4. The method according to claim 2, characterized in that, The feature extraction module also includes a preprocessing unit; Before the step of inputting the multiple parameters into the feature decomposition unit and extracting features using an adaptive mode decomposition algorithm to obtain the feature vector, the method further includes: The raw water quality parameters are input into the pretreatment unit for normalization to eliminate dimensional differences.

5. The method according to claim 1, characterized in that, The step of adjusting the predicted coagulant dosage based on the effluent turbidity includes: If the turbidity of the effluent is greater than the turbidity threshold, the current raw water quality parameters are input into the first prediction module to output the current predicted dosage of coagulant. The coagulant dosing unit is controlled to operate at the current predicted coagulant dosage.

6. A control device for a sedimentation tank in a water supply plant, characterized in that, The device includes: The data acquisition module is used to collect raw water quality parameters through online sensors; The prediction module is used to input the raw water quality parameters into the prediction module and output the predicted coagulant dosage; the prediction module includes a sequentially connected feature extraction module and a BP neural network optimized by particle swarm optimization algorithm; The acquisition module is used to acquire the operating parameters of the sedimentation tank after coagulant is added according to the predicted coagulant dosage. The operating parameters include at least: current sludge level, current influent flow rate, current floc turbidity, current sludge density, and current sludge moisture content. Based on the operating parameters, the dry sludge quantity is calculated. The triggering module is used to trigger the sludge discharge mode when the amount of dry sludge reaches a preset threshold and / or the current sludge level reaches a preset position. The adjustment module is used to: obtain the current sludge concentration during the sludge discharge process in the sludge discharge mode; determine the sludge discharge condition based on the comparison relationship between the current sludge concentration and a preset threshold; determine the adjustment time corresponding to the sludge discharge condition based on a preset correspondence; adjust the execution duration of the sludge discharge mode based on the adjustment time; and adjust the predicted coagulant dosage based on the turbidity of the sedimentation tank effluent. The formula for calculating the amount of dry mud is: TDS total=Q[T×E1+A×E2×K pH +(COD×E3+NH4 + ×E4)C×K org ]×10 -6 ; Where TDS is the total dry sludge volume, t / h; Q is the influent volume, m³ / h. 3 / h; T is the raw water turbidity, NTU; E1 is the conversion factor between turbidity and SS; A is the aluminum salt coagulant injection rate, mg / L; E2 is the conversion factor between Al2O3 and Al(OH)3, K pH E3 represents the conversion factor for the dry sludge amount of COD organic matter complexes; E4 represents the conversion factor for the dry sludge amount of ammonia nitrogen reaction products; K represents the correction factor. org This indicates the correction factor for the influence of organic matter; The formula for dynamically correcting the amount of dry mud in the foundation is: in, This represents the turbidity correction factor for the effluent. This represents the measured value of turbidity in the effluent from the sedimentation tank. Indicates the set value for effluent turbidity; This represents the correction factor for sludge moisture content; This indicates the actual moisture content of the sludge; This indicates the design moisture content of the sludge.

7. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program and the processor running the computer program to cause the electronic device to perform the method of any one of claims 1 to 5.

8. A storage medium, characterized in that, The storage medium stores computer program instructions, which, when read and executed by a processor, perform the method described in any one of claims 1 to 5.

Images