A system and method for predicting the amount of sludge in a settling basin

Abstract

The present application provides a kind of sedimentation tank sludge quantity prediction system and method, it is related to sedimentation tank sludge control technical field, comprising: for each sedimentation period of sedimentation tank, at least one group of historical water inlet parameters and corresponding historical sludge-water interface height in non-sludge discharge period are obtained respectively, and total actual sludge production after sludge discharge period ends and corresponding total sludge-water interface height;According to each historical sludge-water interface height, total actual sludge production and total sludge-water interface height, the real-time actual sludge production corresponding to historical water inlet parameters is obtained by processing;Real-time water inlet parameters are obtained as input, and the sludge quantity prediction model corresponding to the real-time actual sludge production as output is trained;A group of real-time water inlet parameters of sedimentation tank in non-sludge discharge period is obtained, and real-time water inlet parameters are input into sludge quantity prediction model, to obtain the sludge quantity prediction result of sedimentation tank.The beneficial effect is that the real-time sludge production of sedimentation tank can be effectively predicted;Self-adaptive training and correction of sludge quantity prediction model can be realized.

Description

Technical Field

[0001] This invention relates to the field of sedimentation tank sludge control technology, and in particular to a sedimentation tank sludge quantity prediction system and method. Background Technology

[0002] In water treatment plants, flocculation and sedimentation are crucial steps in removing impurities from water, and their structure varies slightly depending on process requirements. Therefore, proper maintenance and management of sedimentation tanks are essential. Among these, proper sludge removal from sedimentation tanks, and improving sludge removal efficiency, are vital for enhancing the cleanliness of the sedimentation tanks and fully realizing their sedimentation function.

[0003] In theory, sedimentation tanks can precisely control the sludge discharge process based on real-time sludge production, thereby improving sludge discharge efficiency. However, currently, due to the lack of methods for real-time detection or prediction of real-time sludge production, the sludge discharge process in sedimentation tanks of water treatment plants mainly begins under the following conditions: the sedimentation tank's operating time exceeds the set value; or the post-sedimentation turbidity is higher than the set value; or the upper control system forces sludge discharge. However, these set values ​​are usually determined based on experience, lacking scientific guidance and exhibiting considerable arbitrariness. Because experience-determined sludge discharge-related parameter settings are often unreasonable, excessively short sludge discharge cycles or excessively long sludge discharge times can lead to unnecessary energy waste and cost losses; conversely, inadequate settings cannot guarantee that the effluent water quality consistently meets standards.

[0004] Therefore, there is an urgent need for a technical solution that can predict the real-time sludge production of sedimentation tanks in order to solve the aforementioned technical problems that exist in the current experience-based approach. Summary of the Invention

[0005] To address the problems existing in the prior art, the present invention provides a sedimentation tank sludge quantity prediction system, comprising:

[0006] The data acquisition module is used to continuously monitor the non-sludge discharge period and sludge discharge period of a sedimentation tank. It takes a single non-sludge discharge period and its corresponding sludge discharge period as a sedimentation cycle. For each sedimentation cycle, it acquires at least one set of historical influent parameters and the corresponding historical mud-water interface height of the sedimentation tank during the non-sludge discharge period, as well as the total actual sludge production and the corresponding total mud-water interface height after the end of the sludge discharge period.

[0007] The data processing module, connected to the data acquisition module, is used to process the historical mud-water interface height, the total actual mud production, and the total mud-water interface height to obtain the real-time actual mud production of the sedimentation tank during the period between the acquisition time of the historical influent parameters and the previous acquisition time.

[0008] The model training module is connected to the data acquisition module and the data processing module respectively. It is used to add each of the real-time influent parameters and the corresponding real-time actual sludge production to a sample set, and train a sludge production prediction model based on the sample set, with the real-time influent parameters as input and the corresponding real-time actual sludge production as output.

[0009] The sludge quantity prediction module is connected to the data acquisition module and the model training module, respectively. It is used to obtain a set of real-time influent parameters of the sedimentation tank during the non-sludge discharge period, and input the real-time influent parameters into the sludge quantity prediction model to obtain the current predicted sludge production of the sedimentation tank as the sludge quantity prediction result of the sedimentation tank.

[0010] Preferably, the historical influent parameters include influent flow rate, influent turbidity, pH value, temperature, and dosage.

[0011] Preferably, the formula for calculating the real-time actual sludge production is as follows:

[0012] S p =(HH′) / H 总 *S 总

[0013] Among them, S p H represents the real-time actual sludge production at the current acquisition time of the historical influent parameters; H represents the historical sludge-water interface height at the current acquisition time; H′ represents the historical sludge-water interface height at the previous acquisition time. 总 Used to represent the total mud-water interface height; S 总 Used to represent the total actual sludge production.

[0014] Preferably, the mud volume prediction model is a BP neural network model, and during the model training process, an sigmoid transfer function is selected and the network weights and thresholds of the BP neural network model are adjusted through a backpropagation error function.

[0015] Preferably, the BP neural network model includes an input layer, a hidden layer, and an output layer connected in sequence, wherein the activation function of the neurons in the hidden layer is a sigmoid tangent function, and the activation function of the neurons in the output layer is a sigmoid logarithmic function.

[0016] Preferably, it further includes a model correction module, which is connected to the data acquisition module, the data processing module, the model training module, and the mud volume prediction module, respectively. The model correction module includes:

[0017] The data acquisition unit is used to acquire the predicted sludge volume corresponding to each of the real-time influent parameters of the sedimentation tank during the entire non-sludge discharge period of each sedimentation cycle, and the total actual sludge production after the end of the sludge discharge period within the same sedimentation cycle.

[0018] The first processing unit, connected to the data acquisition unit, is used to accumulate and add up the prediction results of each mud quantity to obtain the total predicted mud quantity of the entire non-mud discharge stage of each sedimentation cycle, and process it to obtain the ratio between the total predicted mud quantity and the total actual mud production corresponding to the same sedimentation cycle.

[0019] A data extraction unit, connected to the first processing unit, is used to extract the real-time influent parameters associated with the total predicted mud volume corresponding to the ratio when the ratio is not within a preset standard ratio range;

[0020] The second processing unit is connected to the data extraction unit and is used to obtain the real-time actual sludge production corresponding to each of the extracted real-time influent parameters and add it as a new sample to the sample set to update the sample set.

[0021] The third processing unit, connected to the second processing unit, is used to calculate the proportion of the newly added samples in the updated sample set, and generate a correction signal when the proportion of the samples reaches a preset value.

[0022] The model training module retrains the mud volume prediction model based on the correction signal and the updated sample set.

[0023] Preferably, the model correction module further includes a setting port connected to the data extraction unit, used to receive external adjustment instructions to update the standard ratio range.

[0024] This invention also provides a method for predicting the amount of sludge in a sedimentation tank, applied to the aforementioned sedimentation tank sludge prediction system. The method includes:

[0025] Step S1: Continuously monitor the non-sludge discharge period and sludge discharge period of a sedimentation tank. Take a single non-sludge discharge period and its corresponding sludge discharge period as a sedimentation cycle. For each sedimentation cycle, obtain at least one set of historical influent parameters and the corresponding historical mud-water interface height of the sedimentation tank during the non-sludge discharge period, as well as the total actual sludge production and the corresponding total mud-water interface height after the end of the sludge discharge period.

[0026] Step S2: Based on the historical mud-water interface height, the total actual mud production, and the total mud-water interface height, the real-time actual mud production of the sedimentation tank during the period between the historical influent parameter collection time and the previous collection time is obtained.

[0027] Step S3: Add each of the real-time influent parameters and the corresponding real-time actual sludge production to a sample set, and train a sludge production prediction model based on the sample set, using the real-time influent parameters as input and the corresponding real-time actual sludge production as output.

[0028] Step S4: Obtain a set of real-time influent parameters for the sedimentation tank during the non-sludge discharge period, and input the real-time influent parameters into the sludge quantity prediction model to obtain the current predicted sludge production of the sedimentation tank as the sludge quantity prediction result of the sedimentation tank.

[0029] Preferably, the historical influent parameters include influent flow rate, influent turbidity, pH value, temperature, and dosage.

[0030] Preferably, after performing step S4, a model correction process is further included, comprising:

[0031] Step A1: Obtain the predicted sludge volume corresponding to each of the real-time influent parameters of the sedimentation tank during the entire non-sludge discharge period of each sedimentation cycle, and the total actual sludge production after the end of the sludge discharge period within the same sedimentation cycle.

[0032] Step A2: The predicted mud volume results of each sedimentation cycle are accumulated and added together to obtain the total predicted mud volume of the entire non-sludge discharge stage of each sedimentation cycle, and the ratio between the total predicted mud volume and the total actual mud production corresponding to the same sedimentation cycle is obtained.

[0033] Step A3: Determine whether the ratio is within a preset standard ratio range.

[0034] If not, extract the real-time influent parameters associated with the total predicted sludge volume corresponding to the ratio, and then proceed to step A4;

[0035] If so, return to step A2;

[0036] Step A4: Obtain the actual real-time sludge production corresponding to each of the extracted real-time influent parameters and add it as a new sample to the sample set to update the sample set;

[0037] Step A5: Calculate the proportion of the newly added samples in the updated sample set, and generate a correction signal when the proportion of the samples reaches a preset value.

[0038] In step S4, the mud volume prediction model is retrained based on the corrected signal and the updated sample set.

[0039] The above technical solution has the following advantages or beneficial effects:

[0040] 1) It can effectively predict the real-time sludge production of the sedimentation tank, thereby providing effective data support for the optimization and control of the sludge discharge process, assisting in the refined control of the sludge discharge process in water plants, and reducing the intensity of manual operation.

[0041] 2) It can achieve adaptive training and correction of the mud volume prediction model to continuously improve the prediction accuracy of the mud volume prediction model. Attached Figure Description

[0042] Figure 1 A schematic diagram of a sedimentation tank sludge quantity prediction system is shown in a preferred embodiment of the present invention.

[0043] Figure 2 A flowchart illustrating a method for predicting sludge volume in a sedimentation tank, as described in a preferred embodiment of the present invention.

[0044] Figure 3 This is a flowchart illustrating the model correction process in a preferred embodiment of the present invention. Detailed Implementation

[0045] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. The present invention is not limited to this embodiment; other embodiments that conform to the spirit of the present invention may also fall within the scope of the present invention.

[0046] In a preferred embodiment of the present invention, based on the above-mentioned problems existing in the prior art, a sedimentation tank sludge quantity prediction system is provided, such as... Figure 1 As shown, it includes:

[0047] Data acquisition module 1 is used to continuously monitor the non-sludge discharge period and sludge discharge period of a sedimentation tank. It takes a single non-sludge discharge period and its corresponding sludge discharge period as a sedimentation cycle. For each sedimentation cycle, it acquires at least one set of historical influent parameters and the corresponding historical mud-water interface height of the sedimentation tank during the non-sludge discharge period in each sedimentation cycle, as well as the total actual sludge production and the corresponding total mud-water interface height after the end of the sludge discharge period.

[0048] Data processing module 2, connected to data acquisition module 1, is used to process the historical mud-water interface height, total actual mud production, and total mud-water interface height to obtain the real-time actual mud production of the sedimentation tank during the period between the acquisition time of the historical influent parameters and the previous acquisition time.

[0049] Model training module 3 is connected to data acquisition module 1 and data processing module 2 respectively. It is used to add each real-time influent parameter and the corresponding real-time actual sludge production to a sample set, and train a sludge production prediction model with real-time influent parameters as input and corresponding real-time actual sludge production as output based on the sample set.

[0050] The sludge quantity prediction module 4 is connected to the data acquisition module 1 and the model training module 3 respectively. It is used to obtain a set of real-time influent parameters of the sedimentation tank during non-sludge discharge periods, and input the real-time influent parameters into the sludge quantity prediction model to obtain the current predicted sludge production of the sedimentation tank as the sludge quantity prediction result of the sedimentation tank.

[0051] Specifically, in this embodiment, based on the working principle of the sedimentation tank, its working process is divided into a sedimentation stage and a sludge discharge stage. The sedimentation stage is a continuous sludge production process, while the sludge discharge stage is the sludge discharge operation performed on the sludge produced in the corresponding sedimentation stage. Each sedimentation stage corresponds to a sludge discharge stage. Therefore, a single non-sludge discharge period of the sedimentation tank and its corresponding sludge discharge period are treated as a sedimentation cycle. The non-sludge discharge period is used to characterize the duration of the sedimentation stage, and the sludge discharge period is used to characterize the duration of the sludge discharge stage.

[0052] Furthermore, each sedimentation cycle is used as the monitoring object to obtain a sample set for model training. It is preferred to collect the corresponding historical influent parameters at a certain frequency during the non-sludge discharge period of each sedimentation cycle in the sedimentation tank, and collect the mud-water interface height of the sedimentation tank at the time of the historical influent parameter collection as the historical mud-water interface height. That is, each set of historical influent parameters corresponds to a historical mud-water interface height.

[0053] In a preferred embodiment of the present invention, the historical influent parameters include influent flow rate, influent turbidity, pH value, temperature and dosage. Preferably, the historical influent parameters may also include one or more of the following: influent suspended solids concentration, influent ammonia nitrogen concentration, oxygen consumption and color. More preferably, all nine of the above historical influent parameters may be used as historical influent parameters.

[0054] Preferably, for each sedimentation cycle, after collecting the aforementioned historical influent parameters and corresponding historical sludge-water interface heights during non-sludge discharge periods, the method further includes collecting the total sludge-water interface height of the sedimentation tank before sludge discharge begins, and calculating the total actual sludge production after the corresponding sludge discharge period ends. Based on the historical sludge-water interface heights, the total sludge-water interface height, and the total actual sludge production, the real-time actual sludge production at the time of collection for each set of historical influent parameters is then calculated. More preferably, the data acquisition module 1 is also used to collect the suspended solids concentration at the sedimentation port of the sedimentation tank in real time during the sludge discharge period, and considers the sludge discharge period to end when the suspended solids concentration is lower than a preset threshold, at which point the total actual sludge production is obtained.

[0055] In a preferred embodiment of the present invention, the formula for calculating the real-time actual sludge production is as follows:

[0056] S p =(HH′) / H 总 *S 总

[0057] Among them, S p H represents the real-time actual sludge production at the current acquisition time of historical influent parameters; H represents the historical sludge-water interface height at the current acquisition time; H′ represents the historical sludge-water interface height at the previous acquisition time. 总 Used to indicate the total mud-water interface height; S 总 Used to represent the total actual sludge production.

[0058] It is understandable that the real-time actual sludge production is the real-time actual sludge production of the sedimentation tank during the period between the historical influent parameter collection time and the previous collection time. For example, if historical influent parameters are collected three times at a certain frequency during the non-sludge discharge period, then the corresponding collection time includes collection time one, collection time two, and collection time three. If collection time one is the start time of the non-sludge discharge period, then the real-time actual sludge production associated with the corresponding historical influent parameters is zero. If collection time one is not the start time of the non-sludge discharge period, then the corresponding real-time actual sludge production is the sludge production of the sedimentation tank from the start time of the non-sludge discharge period to collection time one. In other words, the start time of the non-sludge discharge period is the previous collection time of collection time one, and the corresponding historical sludge-water interface height is theoretically zero. Alternatively, the historical sludge-water interface height at the start time of the non-sludge discharge period can be pre-configured according to the actual situation and used to participate in the calculation of the real-time actual sludge production corresponding to the first collection time when historical influent parameters are collected for the first time during the non-sludge discharge period. At time two, the corresponding real-time actual mud production is the actual mud production during the period from time one to time two, and so on for the others. This will not be elaborated further here.

[0059] When acquiring multiple sets of historical influent parameters and corresponding real-time actual sludge production for multiple sedimentation cycles, each set of historical influent parameters and corresponding real-time actual sludge production is used as a training sample in the sample set to train a sludge production prediction model that can predict the real-time sludge production of the sedimentation tank. When in use, the collected real-time influent parameters are simply input into the sludge production prediction model to predict the corresponding sludge production.

[0060] In a preferred embodiment of the present invention, the mud volume prediction model is a BP neural network model, and during the model training process, an sigmoid transfer function is selected and the network weights and thresholds of the BP neural network model are adjusted through a backpropagation error function.

[0061] Specifically, in this embodiment, the aforementioned BP neural network model, also known as a backpropagation neural network, continuously adjusts the network weights and thresholds through training with sample data, causing the error function to decrease along the negative gradient direction, thus approximating the desired output. It is a widely used neural network model, consisting of an input layer, hidden layers, and an output layer, with one or more hidden layers. This network uses a sigmoid transfer function: f(x) = 1 / (1+e^(-x) / x).-x (-x is an exponent) Through backpropagation error function: E=∑ i (T i +O i ) 2 / 2(T i For the desired output, O i (This is the computational output of the network), and during the training process, the network weights and thresholds are continuously adjusted to minimize the error function E.

[0062] In a preferred embodiment of the present invention, the BP neural network model includes an input layer, a hidden layer, and an output layer connected in sequence. The activation function of the neurons in the hidden layer is a sigmoid tangent function, and the activation function of the neurons in the output layer is a sigmoid logarithmic function.

[0063] Specifically, in this embodiment, considering that the model's input contains 9 parameters and the output contains 1 parameter, the number of neurons in the input layer of the BP neural network model can be designed to be 9, and the number of neurons in the output layer can be designed to be 1. The number of neurons in the hidden layer is selected according to the following formula:

[0064] l = (n + m) 1 / 2 +a

[0065] Where l represents the number of neurons in the hidden layer, n represents the number of neurons in the input layer, m represents the number of neurons in the output layer, and a is a constant between [1, 10].

[0066] Based on the above formula, the number of neurons in the hidden layer can be calculated to be between 5 and 14. In this technical solution, the number of neurons in the hidden layer is 7.

[0067] During model training, a learning sample is first selected from the sample set and its input is propagated forward. Then, the inputs and outputs of neurons in each layer of the model are calculated, and the output error is calculated. Based on the backpropagation error, the connection weights and thresholds are adjusted. Next, it is determined whether the learning samples in the sample set have been exhausted. If not, another learning sample is selected and the above process is repeated. If exhausted, the average error of the network is calculated. Then, based on the average error, it is determined whether the accuracy of the trained model meets the requirements. If it does, the trained mud volume prediction model is output. If it does not meet the requirements, it is determined whether the maximum number of iterations has been reached. If the maximum number of iterations has been reached, training ends. If the maximum number of iterations has not been reached, another learning sample is selected and the above process is repeated.

[0068] In a preferred embodiment of the present invention, a model correction module 5 is further included, which is connected to the data acquisition module 1, the data processing module 2, the model training module 3, and the mud volume prediction module 4, respectively. The model correction module 5 includes:

[0069] The data acquisition unit 51 is used to acquire the predicted sludge volume corresponding to each real-time influent parameter during the entire non-sludge discharge period of each sedimentation cycle, as well as the total actual sludge production after the end of the sludge discharge period within the same sedimentation cycle.

[0070] The first processing unit 52 is connected to the data acquisition unit 51. It is used to accumulate and add up the prediction results of each mud quantity to obtain the total predicted mud quantity of the entire non-mud discharge stage of each sedimentation cycle, and process it to obtain the ratio between the total predicted mud quantity and the total actual mud production corresponding to the same sedimentation cycle.

[0071] The data extraction unit 53 is connected to the first processing unit 52 and is used to extract the real-time influent parameters associated with the total predicted mud volume corresponding to the ratio when the ratio is not within a preset standard ratio range.

[0072] The second processing unit 54 is connected to the data extraction unit 53 and is used to obtain the real-time actual sludge production corresponding to each real-time influent parameter and add it as a new sample to the sample set to update the sample set.

[0073] The third processing unit 55 is connected to the second processing unit 54 and is used to calculate the proportion of newly added samples in the updated sample set and generate a correction signal when the proportion of samples reaches a preset value.

[0074] Model training module 4 retrains the mud volume prediction model based on the corrected signal and the updated sample set.

[0075] Specifically, in this embodiment, to accurately obtain the total predicted sludge volume for the entire non-sludge discharge stage of each sedimentation cycle, it is preferable to set the last time the real-time influent parameters are collected during the non-sludge discharge stage as the end time of the non-sludge discharge stage. Since the predicted sludge volume result corresponding to the real-time influent parameters is the predicted sludge production volume between the corresponding collection time and the previous collection time, the sum of the predicted sludge volume results for the entire non-sludge discharge stage can be used as the total predicted sludge volume for the entire non-sludge discharge stage. Considering that the real-time actual sludge production volume corresponding to each collection time is derived from the total actual sludge production volume and the corresponding mud-water interface height, it inevitably contains a certain error. Therefore, in this technical solution, the ratio between the total predicted sludge volume and the total actual sludge production volume corresponding to the same sedimentation cycle is used as the evaluation index of the sludge volume prediction model to further improve the accuracy.

[0076] Preferably, the above-mentioned standard ratio range can be obtained by fluctuating around a preset standard, preferably 0.95, then the corresponding standard ratio range is [0.95, 1.05], which can be adjusted according to needs. That is, the total predicted mud amount can be larger or smaller than the total actual mud amount, and the ratio of the deviation between the two does not exceed the above-mentioned standard ratio range, indicating that the prediction result meets the model accuracy requirements.

[0077] In a preferred embodiment of the present invention, the model correction module 5 further includes a setting port 56 connected to the data extraction unit 53, for receiving external adjustment instructions to update the standard ratio range.

[0078] Specifically, in this embodiment, as the model is continuously corrected, the accuracy of its prediction results will gradually improve. In order to speed up the model correction efficiency, the above standard ratio range can be adjusted accordingly after the model is optimized. For example, the above standard ratio range [0.95, 1.05] can be adjusted to [0.98, 1.02]. The specific settings can be made according to the requirements.

[0079] In a preferred embodiment, the adjustment of the standard ratio range can also be configured to be automatic. Specifically, multiple different standard ratio ranges can be configured in the model correction module 5. Each standard ratio range can be associated with a corresponding preset threshold. During the usage phase of each mud quantity prediction model, the number of ratios outside the standard ratio range is continuously counted. When the trend of the change of the proportion of the number of these ratios to the total number of mud quantity prediction results of the mud quantity prediction model over time indicates that the accuracy of the mud quantity prediction model reaches the corresponding preset threshold, the existing standard ratio range is adjusted to the standard ratio range associated with the preset threshold. It is understood that the above embodiment is only one example of this technical solution and is not intended to limit the technical solution.

[0080] This invention also provides a method for predicting the amount of sludge in a sedimentation tank, applicable to the aforementioned sedimentation tank sludge prediction system, such as... Figure 2 As shown, the methods for predicting the amount of sludge in sedimentation tanks include:

[0081] Step S1: Continuously monitor the non-sludge discharge period and sludge discharge period of a sedimentation tank. Take a single non-sludge discharge period and its corresponding sludge discharge period as a sedimentation cycle. For each sedimentation cycle, obtain at least one set of historical influent parameters and the corresponding historical mud-water interface height of the sedimentation tank during the non-sludge discharge period in each sedimentation cycle, as well as the total actual sludge production and the corresponding total mud-water interface height after the sludge discharge period ends.

[0082] Step S2: Based on the historical mud-water interface height, total actual mud production, and total mud-water interface height, the real-time actual mud production of the sedimentation tank during the period between the historical influent parameter collection time and the previous collection time is obtained.

[0083] Step S3: Add each real-time influent parameter and the corresponding real-time actual sludge production to a sample set, and train a sludge production prediction model based on the sample set, with the real-time influent parameter as input and the corresponding real-time actual sludge production as output.

[0084] Step S4: Obtain a set of real-time influent parameters for the sedimentation tank during non-sludge discharge periods, and input the real-time influent parameters into the sludge quantity prediction model to obtain the current predicted sludge production of the sedimentation tank as the sludge quantity prediction result of the sedimentation tank.

[0085] In a preferred embodiment of the present invention, the historical influent parameters include influent flow rate, influent turbidity, pH value, temperature, and dosage.

[0086] In a preferred embodiment of the present invention, a model correction process is further included after step S4, such as... Figure 3 As shown, it includes:

[0087] Step A1: Obtain the predicted sludge volume for each real-time influent parameter during the entire non-sludge discharge period of each sedimentation cycle, as well as the total actual sludge production after the end of the sludge discharge period within the same sedimentation cycle.

[0088] Step A2: The predicted mud volume results of each mud volume are summed up to obtain the total predicted mud volume of the entire non-sludge discharge stage of each sedimentation cycle, and the ratio between the total predicted mud volume and the total actual mud production corresponding to the same sedimentation cycle is obtained.

[0089] Step A3: Determine whether the ratio is within a preset standard ratio range.

[0090] If not, extract the real-time influent parameters associated with the total predicted sludge volume corresponding to the ratio, and then proceed to step A4;

[0091] If so, return to step A2;

[0092] Step A4: Obtain the actual real-time sludge production corresponding to each extracted real-time influent parameter and add it to the sample set as a new sample to update the sample set.

[0093] Step A5: Calculate the proportion of newly added samples in the updated sample set, and generate a correction signal when the proportion of samples reaches a preset value.

[0094] In step S4, the mud volume prediction model is retrained based on the corrected signal and the updated sample set.

[0095] The above description is merely a preferred embodiment of the present invention and does not limit the implementation and protection scope of the present invention. Those skilled in the art should realize that any equivalent substitutions and obvious changes made using the content of this specification and illustrations should be included within the protection scope of the present invention.

Claims

1. A system for predicting the amount of sludge in a sedimentation basin, characterized by, include: The data acquisition module is used to continuously monitor the non-sludge discharge period and sludge discharge period of a sedimentation tank. It takes a single non-sludge discharge period and its corresponding sludge discharge period as a sedimentation cycle. For each sedimentation cycle, it acquires at least one set of historical influent parameters and the corresponding historical mud-water interface height of the sedimentation tank during the non-sludge discharge period, as well as the total actual sludge production and the corresponding total mud-water interface height after the end of the sludge discharge period. The data processing module, connected to the data acquisition module, is used to process the historical mud-water interface height, the total actual mud production, and the total mud-water interface height to obtain the real-time actual mud production of the sedimentation tank during the period between the acquisition time of the historical influent parameters and the previous acquisition time. The model training module is connected to the data acquisition module and the data processing module respectively. It is used to add each real-time influent parameter and the corresponding real-time actual sludge production to a sample set, and train a sludge production prediction model based on the sample set, with the real-time influent parameter as input and the corresponding real-time actual sludge production as output. The sludge quantity prediction module is connected to the data acquisition module and the model training module, respectively. It is used to obtain a set of real-time influent parameters of the sedimentation tank during the non-sludge discharge period, and input the real-time influent parameters into the sludge quantity prediction model to obtain the current predicted sludge production of the sedimentation tank as the sludge quantity prediction result of the sedimentation tank.

2. The sedimentation basin sludge amount prediction system according to claim 1, characterized by, The historical influent parameters include influent flow rate, influent turbidity, pH, temperature, and dosage.

3. The sedimentation basin sludge quantity prediction system according to claim 1, characterized by, The formula for calculating the real-time actual sludge production is as follows: ； in, The real-time actual sludge production amount is used to represent the current acquisition time of the historical influent parameters; The historical mud-water interface height is used to indicate the current acquisition time. The historical mud-water interface height used to represent the previous acquisition time at the current acquisition time; Used to indicate the total mud-water interface height; Used to represent the total actual sludge production.

4. The sedimentation basin sludge quantity prediction system according to claim 1, characterized by, The mud volume prediction model is a BP neural network model, and during the model training process, an sigmoid transfer function is selected and the network weights and thresholds of the BP neural network model are adjusted through a backpropagation error function.

5. The sedimentation basin sludge quantity prediction system according to claim 4, characterized by, The BP neural network model includes an input layer, a hidden layer, and an output layer connected in sequence. The activation function of the neurons in the hidden layer is a sigmoid tangent function, and the activation function of the neurons in the output layer is a sigmoid logarithmic function.

6. The sedimentation basin sludge quantity prediction system according to claim 1, characterized by, It also includes a model correction module, which is connected to the data acquisition module, the data processing module, the model training module, and the mud volume prediction module. The model correction module includes: The data acquisition unit is used to acquire the predicted sludge volume corresponding to each of the real-time influent parameters of the sedimentation tank during the entire non-sludge discharge period of each sedimentation cycle, as well as the total actual sludge production after the end of the sludge discharge period within the same sedimentation cycle. The first processing unit, connected to the data acquisition unit, is used to accumulate and add up the prediction results of each mud quantity to obtain the total predicted mud quantity for the entire non-mud discharge period of each sedimentation cycle, and process it to obtain the ratio between the total predicted mud quantity and the total actual mud production corresponding to the same sedimentation cycle. A data extraction unit, connected to the first processing unit, is used to extract the real-time influent parameters associated with the total predicted mud volume corresponding to the ratio when the ratio is not within a preset standard ratio range. The second processing unit is connected to the data extraction unit and is used to obtain the real-time actual sludge production corresponding to each of the extracted real-time influent parameters and add it as a new sample to the sample set to update the sample set. The third processing unit, connected to the second processing unit, is used to calculate the proportion of the newly added samples in the updated sample set, and generate a correction signal when the proportion of the samples reaches a preset value. The model training module retrains the mud volume prediction model based on the correction signal and the updated sample set.

7. The sedimentation basin sludge quantity prediction system according to claim 6, characterized by, The model correction module also includes a setting port connected to the data extraction unit, used to receive external adjustment instructions to update the standard ratio range.

8. A method for predicting the amount of sludge in a sedimentation tank, characterized in that, The sedimentation tank sludge quantity prediction system as described in any one of claims 1-7, wherein the sedimentation tank sludge quantity prediction method comprises: Step S1: Continuously monitor the non-sludge discharge period and sludge discharge period of a sedimentation tank. Take a single non-sludge discharge period and its corresponding sludge discharge period as a sedimentation cycle. For each sedimentation cycle, obtain at least one set of historical influent parameters and the corresponding historical mud-water interface height of the sedimentation tank during the non-sludge discharge period, as well as the total actual sludge production and the corresponding total mud-water interface height after the end of the sludge discharge period. Step S2: Based on the historical mud-water interface height, the total actual sludge production, and the total mud-water interface height, the real-time actual sludge production of the sedimentation tank during the period between the historical influent parameter collection time and the previous collection time is obtained. Step S3: Add each of the real-time influent parameters and the corresponding real-time actual sludge production to a sample set, and train a sludge production prediction model based on the sample set, using the real-time influent parameters as input and the corresponding real-time actual sludge production as output. Step S4: Obtain a set of real-time influent parameters for the sedimentation tank during the non-sludge discharge period, and input the real-time influent parameters into the sludge quantity prediction model to obtain the current predicted sludge production of the sedimentation tank as the sludge quantity prediction result of the sedimentation tank.

9. The method according to claim 8, wherein The historical influent parameters include influent flow rate, influent turbidity, pH, temperature, and dosage.

10. The method according to claim 8, wherein After performing step S4, a model correction process is also included, including: Step A1: Obtain the predicted sludge volume corresponding to each of the real-time influent parameters of the sedimentation tank during the entire non-sludge discharge period of each sedimentation cycle, and the total actual sludge production after the end of the sludge discharge period within the same sedimentation cycle. Step A2: The predicted mud volume results of each sedimentation cycle are accumulated and added together to obtain the total predicted mud volume for the entire non-sludge discharge period of each sedimentation cycle, and the ratio between the total predicted mud volume and the total actual mud production corresponding to the same sedimentation cycle is obtained. Step A3: Determine whether the ratio is within a preset standard ratio range. If not, extract the real-time influent parameters associated with the total predicted sludge volume corresponding to the ratio, and then proceed to step A4; If so, return to step A2; Step A4: Obtain the actual real-time sludge production corresponding to each of the extracted real-time influent parameters and add it as a new sample to the sample set to update the sample set; Step A5: Calculate the proportion of the newly added samples in the updated sample set, and generate a correction signal when the proportion of the samples reaches a preset value. In step S4, the mud volume prediction model is retrained based on the corrected signal and the updated sample set.

Images