Self-adaptive intelligent digital dosing method and system based on water demand prediction

Abstract

The invention discloses a self-adaptive intelligent digital dosing method and system based on water demand prediction, and relates to the technical field of water treatment dosing. According to the method, modeling indexes are screened from operation indexes, and a dosing model is obtained based on the modeling indexes; determining the water demand of a prediction period through the water demand prediction time sequence model, and further generating a water taking scheme of the prediction period; based on the water taking scheme, using a dosing model to obtain a dosing amount of a prediction period; generating a water plant agent adding equipment scheduling plan according to the agent adding amount in the prediction period; based on the real-time modeling index data, the real-time predicted dosage is obtained; and determining the real-time dosage according to the dosage predicted in real time and the control rule of the medicament adding equipment, and issuing the real-time dosage to the medicament adding equipment based on the scheduling plan of the medicament adding equipment of the water plant to adjust the dosage. The accuracy of the dosage can be improved in the overall dispatching process of the water plant, and the outgoing water quality is stabilized.

Description

Technical Field

[0001] The present application relates to the technical field of water treatment dosing, and in particular to an adaptive intelligent digital dosing method and system based on water demand prediction. Background Art

[0002] In the water treatment process, dosing is an important step. By adding chemicals, additives or other liquid treatment agents to water, impurities and pollutants in the water can be effectively removed and the water quality can be improved. For example, adding chemicals to boiler water can prevent scaling and corrosion and ensure the normal operation of the boiler. In a water plant, the dosing process usually includes building a dosing tank, using a diaphragm dosing pump and a mixer to fully mix the liquid medicine and the water to be treated, reduce the turbidity of the water, and make the water clear, so as to meet the standards for drinking water.

[0003] The dosage directly affects the quality of produced water. The dosage of existing dosing systems is usually determined based on the water volume. The water volume data comes from the actual flow data measured by the sensor meter. Based on the actual flow data, only the dosage calculation and adjustment can be performed at the current measurement time, and it cannot adapt to the overall scheduling process of the water plant in the future. Moreover, the measurement of water volume data depends on the sensor meter. If the sensor meter itself has a measurement error, it will lead to deviations in the water volume data. Based on the water volume data, a relatively fixed formula model is usually used to calculate the dosage. Some fixed coefficients of the formula also rely on manual maintenance and correction, which will cause calculation distortion due to human factors. Summary of the invention

[0004] The purpose of this application is to provide an adaptive intelligent digital dosing method and system based on water demand prediction, which can improve the accuracy of dosing during the overall scheduling of the water plant and stabilize the water quality out of the factory.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] In the first aspect, the present application provides an adaptive intelligent digital dosing method based on water demand prediction, including: creating a dosing scenario model, and receiving historical data of operating indicators and historical data of dosing amounts of the water plant during the actual water treatment process through the dosing scenario model; performing correlation analysis on the operating indicators and the dosing amounts, screening modeling indicators from the operating indicators based on the correlation analysis results, and forming a sample set with the historical data of the modeling indicators, the statistical characteristics of the modeling indicators and the historical data of the dosing amounts; the modeling indicators at least include the instantaneous flow rate of inlet water, the quality of raw water, the quality of process water and the quality of outlet water; using the sample set to train a neural network model to obtain a dosing model; determining the water demand of the prediction period through a water demand prediction time series model; the water demand of the prediction period includes the instantaneous outgoing water flow rate at each prediction moment in a future period of time; generating a water intake plan for the prediction period based on the water demand of the prediction period and the target liquid level setting of the reservoir; the water intake plan for the prediction period includes the inlet water flow at each prediction moment in a future period of time amount; inputting the modeling index data including at least the water intake scheme, the average value of raw water quality, the set value of process water quality and the set value of outgoing water quality of at least the forecast period into the dosing model together with the statistical characteristics of the modeling index to obtain the dosing amount of the forecast period; the dosing amount of the forecast period is the dosing amount at each forecast moment in the future period; determining the dosing control scheme of the forecast period according to the dosing amount of the forecast period and the control rule of the dosing equipment, and generating the scheduling plan of the dosing equipment of the water plant according to the dosing control scheme of the forecast period; receiving the modeling index data in real time through the dosing scenario model, and inputting the modeling index data including at least the instantaneous flow rate of the incoming water, the raw water quality, the process water quality and the outgoing water quality collected in real time into the dosing model together with the statistical characteristics of the modeling index to obtain the real-time predicted dosing amount; determining the real-time dosing amount according to the real-time predicted dosing amount and the control rule of the dosing equipment, and sending the real-time dosing amount to the dosing equipment based on the scheduling plan of the dosing equipment of the water plant to adjust the dosing amount.

[0007] In the second aspect, the present application provides an adaptive intelligent digital dosing system based on water demand prediction, which supports local deployment and cloud deployment; the adaptive intelligent digital dosing system based on water demand prediction executes the above-mentioned adaptive intelligent digital dosing method based on water demand prediction.

[0008] According to the specific embodiments provided in this application, this application has the following technical effects:

[0009] The present application provides an adaptive intelligent digital dosing method and system based on water demand prediction. The water demand in the future is predicted by a water demand prediction time series model. The dosing model can be used to adaptively predict the dosing amount in the future. The dosing amount at the current moment can be adaptively predicted by the instantaneous flow rate of incoming water, the quality of raw water, the quality of process water and the quality of outgoing water collected in real time at the current moment. This method can better adapt to the overall scheduling process of the water plant. At the same time, water volume data no longer relies on sensor meters, and the calculation of dosing amount no longer uses a fixed formula model, which improves the accuracy of dosing and stabilizes the outgoing water quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0011] Figure 1 A schematic flow chart of an adaptive intelligent digital dosing method based on water demand prediction provided in one embodiment of the present application. DETAILED DESCRIPTION

[0012] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0013] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0014] In an exemplary embodiment, Figure 1 As shown, the embodiment of the present application provides an adaptive intelligent digital dosing method based on water demand prediction, including the following steps 101 to 109. Among them:

[0015] Step 101: Create a dosing scenario model, and receive historical data of operating indicators and dosing amounts of the water plant during an actual water treatment process through the dosing scenario model.

[0016] Step 102: perform a correlation analysis on the operating indicators and the dosage, select modeling indicators from the operating indicators based on the correlation analysis results, and form a sample set with the historical data of the modeling indicators, the statistical characteristics of the modeling indicators and the historical data of the dosage; the modeling indicators at least include the instantaneous flow rate of the inlet water, the quality of the raw water, the quality of the process water and the quality of the outlet water.

[0017] Step 103: Use the sample set to train a neural network model to obtain a medication model.

[0018] Step 104: Determine the water demand of the forecast period through the water demand forecasting time series model; the water demand of the forecast period includes the instantaneous outflow water flow at each forecasting moment in the future period of time.

[0019] Step 105: Generate a water intake plan for the forecast period according to the water demand for the forecast period and the target liquid level setting of the reservoir; the water intake plan for the forecast period includes the water flow rate of the plant at each forecast time in the future period.

[0020] Step 106: Input the modeling indicator data including at least the water intake plan of the forecast period, the average value of raw water quality, the process water quality set value and the factory water quality set value together with the statistical characteristics of the modeling indicators into the dosing model to obtain the dosing amount of the forecast period; the dosing amount of the forecast period is the dosing amount at each forecast moment in the future period.

[0021] Step 107: Determine a dosing control plan for the forecast period based on the dosing amount for the forecast period and the control rule of the dosing equipment, and generate a scheduling plan for the dosing equipment for the water plant based on the dosing control plan for the forecast period.

[0022] Step 108: Receive modeling index data in real time through the dosing scenario model, and input the modeling index data including at least the instantaneous flow rate of inlet water, raw water quality, process water quality and outlet water quality collected in real time together with the statistical characteristics of the modeling index into the dosing model to obtain a real-time predicted dosing amount.

[0023] Step 109: Determine the real-time dosage according to the real-time predicted dosage and the control rule of the dosage device, and send the real-time dosage to the dosage device based on the scheduling plan of the dosage device of the water plant to adjust the dosage.

[0024] By implementing the above steps 101 to 109, the water demand in the future can be predicted through the water demand prediction time series model. The dosing model can be used to adaptively predict the dosing amount in the future. The dosing amount at the current moment can be adaptively predicted through the instantaneous flow rate of the incoming water, the raw water quality, the process water quality and the outgoing water quality collected in real time at the current moment. This can better adapt to the overall scheduling process of the water plant. At the same time, the water volume data no longer depends on the sensor meter, and the calculation of the dosing amount no longer uses a fixed formula model, which improves the accuracy of the dosing amount and stabilizes the outgoing water quality.

[0025] In another exemplary embodiment of the present application, the creation of the dosing scenario model in the above step 101 can be replaced by the following steps 201 to 203:

[0026] Step 201: Select a dosing scenario on the model building platform.

[0027] Enter the scene name on the model building platform, such as "PAC (Poly aluminumChloride) dosing scene in the flocculation process of a certain water plant", and the scene introduction, construction process, usage algorithm, scene requirements, etc. of the smart dosing scene will be displayed. After fully understanding it, you can click OK to create the smart dosing scene.

[0028] Step 202: configuring the modeled drug in the drug addition scenario.

[0029] Step 203: Based on the modeled agent, the operating indicators and the associated agent dosing equipment are bound in the dosing scenario, the expected target of dosing is set, and the data collection frequency, data length, and data volume verification of the operating indicators are set, and the pump frequency limit, usage status, usage priority and / or pump flow limit of the associated agent dosing equipment are set to obtain a dosing scenario model; the associated agent dosing equipment supports configuration of dosing strategies including over-limit strategy, pump frequency recommendation strategy and / or pump flow recommendation strategy; the over-limit strategy is to limit the frequency allocated to the pump to not exceed the upper limit of the pump frequency, or the maximum flow of the pump to not exceed the upper limit of the pump flow; the pump frequency recommendation strategy refers to determining the start and stop combination of the pump and the frequency allocated to each pump based on the over-limit strategy; the pump flow recommendation strategy refers to determining the start and stop combination of the pump and the flow allocated to each pump based on the over-limit strategy.

[0030] Configure modeling agents (such as flocculants), select IoT devices related to the modeling agents (an inherent attribute required in the modeling process). The IoT devices are derived from the IoT platform independently developed by the inventive subject, and key indicators are bound, associated pump equipment is bound, the expected goals of intelligent dosing are set, key indicators are verified in terms of data collection frequency, data length, data volume, etc., and a verification analysis is issued.

[0031] In another exemplary embodiment of the present application, the above step 101 of receiving the historical data of the operating indicators and the historical data of the dosage of the water plant in the actual water treatment process through the dosing scenario model can be replaced by the following steps 301 to 304:

[0032] Step 301: Considering the delay difference of different process operations in the water treatment process, configure the delay coefficient.

[0033] The delay coefficient is the delay of the correlation between the data of different process sections. Because the water in the water plant has a certain hydraulic retention time from water inlet to water outlet, different data correspond to different time dimensions. For example, the water quality data of the outlet section at the same time should be benchmarked against the water quality data of the water inlet several hours or even more than ten hours ago. Therefore, it is necessary to configure the delay coefficient and reasonably plan the time of reagent addition to ensure that the analysis of the reagent dosage (dosage) and related parameters are in the same time dimension. The system automatically aligns the historical data delay, or manually sets the correction delay coefficient.

[0034] Step 302: Receive historical data of operating indicators and historical data of cumulative dosage of reagents in the actual water treatment process of the water plant through the dosing scenario model.

[0035] Step 303: Subtract the historical data of the cumulative dosage of the drug at adjacent receiving moments to obtain the historical data of the dosage at each receiving moment.

[0036] Step 304: According to the delay coefficient, the historical data of the operation index and the historical data of the dosage are aligned with each other.

[0037] In another exemplary embodiment of the present application, in order to screen out the key indicators that affect the dosage from multiple operating indicators, the operating indicators are statistically analyzed for maximum, minimum, mean, variance and other information, and then the relationship between various indicators and the dosage of the agent is analyzed with one click, and the indicator correlation ranking and correlation are calculated; support batch configuration of modeling indicators according to the correlation analysis results, including feedforward indicators (such as inlet flow, raw water quality, etc.), feedback indicators (such as outlet water quality at the dosing point, factory water quality, etc.); feedback indicators support configuration of feedback strategies, including whether to enable, and the configuration method supports interface configuration and manual configuration. Then the above step 102 can be replaced by the following steps 401 to 404:

[0038] Step 401: Determine the statistical characteristics of the operation index; the statistical characteristics at least include the maximum value, the minimum value, the mean value and the variance.

[0039] Step 402: According to the historical data of the operating index, the statistical characteristics of the operating index and the historical data of the dosage, a correlation analysis is performed on the operating index and the dosage to obtain the correlation coefficient of the operating index.

[0040] Step 403: Select an operating indicator whose correlation coefficient is greater than a preset coefficient threshold as a modeling indicator.

[0041] The modeling indicators include at least the instantaneous flow rate of inlet water, raw water quality, process water quality and outlet water quality. The indicators of raw water quality include raw water turbidity, pH, water temperature, conductivity, etc.

[0042] Step 404: After data cleaning, the historical data of the modeling indicators, the statistical characteristics of the modeling indicators and the historical data of the dosage are used to form a sample set.

[0043] Built-in data cleaning process customized for intelligent dosing scenarios, automatically identifies abnormal data and missing data. Users can clean data with one click, thus effectively ensuring the quality of data samples. Manual cleaning is also supported. Supports viewing statistical information, abnormal information and data details of data before and after processing, and supports exporting data.

[0044] Data anomaly identification uses methods such as BoxplotMethod and DBSCAN to calculate abnormal outlier data, and missing data is filtered by conditional filtering. After identification, the mean filter method is used to backfill missing or abnormal data. The sample set is usually an Excel file.

[0045] For example, the correlation analysis uses the fit and score methods of LinearRegression to calculate R 2 The coefficient of determination is the correlation coefficient, and the formula is R 2 =1-(RSS / TSS), where RSS represents the variance not explained by the regression model, and TSS represents the total variance of the dependent variable. i is the actual observed value, is the predicted value, is the average of all actual observations.

[0046] In another exemplary embodiment of the present application, the neural network model may be a multi-layer feedforward neural network (BackPropagation, BP).

[0047] In another exemplary embodiment of the present application, when the neural network model is trained in the above step 103, the corresponding number of hidden layers, batch size, training rounds, loss selection, learning rate, maximum training time, maximum number of training models, etc. will be built in, and the training status, current training rounds, and current training model loss value will be displayed during the training process; retraining is supported; and viewing of training logs is supported. After the training is completed, the model is evaluated from the perspectives of model performance, model stability, and model accuracy; the comparison curves between the prediction and actual data for historical data are displayed from two aspects: water quality and dosage; and the fitting effect of the intelligent dosing model is evaluated from the perspectives of mean square error, mean absolute error, root mean square error, and determination coefficient.

[0048] In another exemplary embodiment of the present application, in order to improve the adaptability of the dosing model to the latest situation, after the above step 103, the method may further include the following steps 501 to 504. Among them:

[0049] Step 501: Configure self-learning trigger conditions and self-learning end conditions.

[0050] Step 502: When the self-learning trigger condition is triggered, incremental learning of the dosing model is performed according to the newly generated time series data of the operation indicators and dosing amount of the water plant during the actual water treatment process.

[0051] Step 503: If the medication dosing model after incremental learning meets the self-learning end condition, the medication dosing model is switched to the medication dosing model after incremental learning; the switching includes automatic switching and manual switching.

[0052] Step 504: If the dosing model after incremental learning does not meet the self-learning end condition, return to the step of "performing incremental learning of the dosing model according to the newly generated time series data of the water plant's operating indicators and dosing amounts during the actual water treatment process".

[0053] Continuous incremental learning and iteration are carried out based on real-time data after operation to improve the adaptability of the model to the latest situation. It supports the configuration of self-learning trigger conditions, self-learning end conditions, model iteration replacement methods, etc., and displays iteration history records. You can switch historical models for re-publishing and re-learning.

[0054] Incremental learning has self-learning trigger conditions. When the conditions are triggered, the system starts the self-learning code function to perform a round of model training. When the round of training is completed, if the self-learning end conditions are not met, retraining will be performed until the end conditions are met. The model iteration replacement method can be selected as immediate replacement or manual execution. If it is replaced immediately, the system will replace and publish the trained model that meets the end conditions. If it is executed manually, you need to manually click Publish to replace the original published model.

[0055] In another exemplary embodiment of the present application, the above-mentioned step 104 predicts water demand through a water demand prediction time series model, which may include: inputting seasonal data and weather data corresponding to the prediction period into the water demand prediction time series model to obtain the water demand of the prediction period; the seasonal data at least includes holidays, weeks, months and quarters; the weather data at least includes weather and temperature.

[0056] By using information such as weather, temperature, holidays, weeks, months and quarters, the water demand forecasting model is linked to the water volume of the process section and the water volume leaving the factory, and the changes in water volume in the future can be predicted.

[0057] In another exemplary embodiment of the present application, the water demand of the forecast period obtained in the above step 104 is the outflow of water from the water plant, and the dosage is related to the inflow of water, so in step 105, it is also necessary to generate a water intake plan for the forecast period according to the water demand of the forecast period and the target liquid level setting of the reservoir. The water intake plan for the forecast period includes the inflow of water at each forecast time in a future period, where the future period can refer to the next 24 hours or a time period of other lengths in the future.

[0058] Exemplarily, according to the water demand of the forecast period and the target liquid level of the reservoir, an optimization algorithm (e.g., genetic algorithm) is used to generate a water intake plan for the forecast period. The water intake plan refers to a plan in which less water is produced between 6 and 24 o'clock, more water is produced between 0 and 6 o'clock, and the liquid level of the clear water tank of the target water plant does not bottom out and cause a low-level alarm, so that the preferential electricity price during off-peak hours (such as 0-6 o'clock) can be fully utilized to increase the water production, and the water production can be reduced during non-off-peak hours (such as 6-24 o'clock), while ensuring that the liquid level of the clear water tank does not bottom out and cause a low-level alarm, thereby achieving the purpose of reducing the water production cost of the water plant.

[0059] In another exemplary embodiment of the present application, the average value of the raw water quality in the above step 106 is obtained by averaging the raw water quality over a recent period of time.

[0060] In another exemplary embodiment of the present application, the determination of the dosing control scheme of the prediction cycle in step 107 can be replaced by the following steps 601 to 604:

[0061] Step 601: Setting control rules for the drug dosing device; there are at least two control rules for the drug dosing device, one is that the drug dosing device controls the dosage by setting the dosing pump frequency, and the other is that the drug dosing device controls the dosage by setting the dosing pump flow rate.

[0062] Step 602: When the control rule of the drug dosing device is that the drug dosing device controls the dosage by setting the dosing pump frequency, a linear relationship between the dosing pump frequency and the dosage is fitted.

[0063] The linear relationship between the dosing pump frequency and the dosage is fitted, which facilitates the subsequent generation of corresponding pump frequency control suggestions based on the dosage suggestions.

[0064] Step 603: According to the dosage of the predicted period, based on the linear relationship between the dosage pump frequency and the dosage, the dosage pump frequency of the predicted period is obtained and determined as the dosage control scheme.

[0065] Step 604: When the control rule of the drug dosing equipment is that the drug dosing equipment controls the dosage by setting the dosing pump flow rate, the flow rate of each pump is directly set according to the dosage of the prediction period, and the flow rate of each pump in the prediction period is determined as the dosing control scheme; wherein the total of all pump flows in the prediction period is the total dosage of the prediction period.

[0066] In another exemplary embodiment of the present application, in order to continuously adapt to the latest operating status of the dosing device, after step 603, the method may further include the following steps 605 to 606. Among them:

[0067] Step 605: Set the fitting period.

[0068] Step 606: In each fitting cycle, the linear relationship between the dosing pump frequency and the dosing amount is refitted according to the latest operating status of the dosing equipment and the dosing pump.

[0069] The scheduling plan for the water plant's chemical dosing equipment can include the selection of dosing pumps for a period of time in the future. If a dosing pump fails, it will no longer be used when adjusting the dosage.

[0070] In another exemplary embodiment of the present application, since the actual control process is often different from the prediction, a dosage compensation mechanism and a feedback mechanism are set. After the above step 109, the method may also include: according to the actual operating state in the dosing control process, calculating the compensation dosage required for the next prediction cycle through the dosage compensation mechanism, and / or calculating the feedback dosage required for the next prediction cycle through the feedback mechanism.

[0071] Among them, the dosage compensation mechanism is: if the dosage of the predicted period is inconsistent with the actual dosage in the dosage control process, the difference between the dosage of the predicted period and the actual dosage in the dosage control process is calculated as the compensation dosage, and the dosage is supplemented or reversely compensated according to the compensation dosage in the next prediction period.

[0072] The feedback mechanism is: if the process water quality does not meet expectations, and / or the factory water quality does not meet expectations, a protective dosage feedback adjustment is performed to deal with water quality abnormalities. The protective dosage feedback adjustment refers to determining the feedback dosage required for the next forecast period based on the real-time data of the operating indicators during the dosing control process and the relationship between the operating indicator data and the dosage.

[0073] The relationship between operating index data and dosage varies with the situation of each water plant and is determined based on the specific actual situation of each water plant.

[0074] The feedback mechanism mainly performs multi-level feedback adjustment calculations based on the feedback rules of related indicators, and obtains the feedback dosage by integrating all feedback results. The total predicted dosage is then converted into a dosing control plan for distribution.

[0075] In another exemplary embodiment of the present application, the method may further include: determining the data refresh cycle based on checking the data collection frequency of the operating indicators. For example, if the original collected data is one item every 5 minutes, then the interval between the next prediction period and the previous prediction period can be greater than or equal to 5 minutes. The specific setting needs to be set in combination with process requirements and human regulations. Usually the time period is consistent with the collection frequency, or the default cycle is hourly. For example, one piece of data is predicted at each hour, 24 items a day, or one piece of data is predicted at each minute point that is a multiple of 5.

[0076] This application creates an intelligent dosing scenario model through a model building platform, and configures key input and output points and control point information, analyzes the timing relationship between each data indicator and configures the delay coefficient, analyzes the correlation between each data indicator and the dosage, determines all input and output indicators based on the correlation ranking and configures a general feedback logic, retrieves historical data based on the selected input and output indicators for data cleaning, screening, and identification, and then combines the processed sample data with the configured model hyperparameters to start batch training models, select the best performing model to view the model evaluation results and the predicted effects on recent water quality results and dosing results, select a deployment method, configure self-learning conditions, and select the current or historical training model to publish to the corresponding deployment method. Then establish a water demand forecasting time series model, configure the forecasting cycle, and the dosing model performs regular adaptive dosage forecasting based on the forecasting cycle and the results of the water demand forecasting model, and sends the forecasting results to the device end for control and execution of forecasting instructions.

[0077] The data in this application is based on a comprehensive correlation analysis of the overall production data, and then the model is built. The factors considered are more comprehensive. After the model is built, samples can be continuously accumulated, and the model can be trained and iterated independently to ensure the availability and timeliness of the model.

[0078] The water demand data of this application is the water consumption for a period of time in the future calculated according to the water demand model. Based on the water demand forecast, it can better adapt to the overall scheduling process of the water plant.

[0079] The advantages of this application method are as follows:

[0080] ① High modeling efficiency, able to realize agile modeling (most existing technologies have long modeling cycles and are difficult to implement).

[0081] ② The modeling reference indicators are comprehensive (associated with real IoT device points and analyzed for historical data correlation).

[0082] ③ Match water usage patterns and the overall scheduling strategy of the water plant (water demand forecast).

[0083] ④Learn independently for a long time and maintain a high accuracy rate (autonomous learning mechanism).

[0084] ⑤ Dynamic compensation of dosage (dosage compensation mechanism and feedback mechanism).

[0085] Based on the same inventive concept, the embodiment of the present application also provides an adaptive intelligent digital dosing system based on water demand prediction for implementing the adaptive intelligent digital dosing method based on water demand prediction involved above. The implementation scheme for solving the problem provided by the system is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more embodiments of the adaptive intelligent digital dosing system based on water demand prediction provided below can refer to the limitations of the adaptive intelligent digital dosing system method based on water demand prediction above, and will not be repeated here.

[0086] In an exemplary embodiment, an adaptive intelligent digital dosing system based on water demand prediction is provided, and the adaptive intelligent digital dosing system based on water demand prediction supports local deployment and cloud deployment; the adaptive intelligent digital dosing system based on water demand prediction executes the above-mentioned adaptive intelligent digital dosing method based on water demand prediction.

[0087] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0088] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. An adaptive intelligent digital dosing method based on water demand prediction, characterized in that: The adaptive intelligent digital dosing method based on water demand prediction includes: Creating a dosing scenario model, and receiving historical data of operating indicators and dosing amounts of the water plant during an actual water treatment process through the dosing scenario model; Conducting a correlation analysis on the operation index and the dosage, selecting a modeling index from the operation index according to the correlation analysis result, and forming a sample set with the historical data of the modeling index, the statistical characteristics of the modeling index and the historical data of the dosage; the modeling index at least includes the instantaneous flow rate of the inlet water, the quality of the raw water, the quality of the process water and the quality of the outlet water; Using the sample set to train a neural network model to obtain a dosing model; Determine the water demand of the forecast period through the water demand forecasting time series model; the water demand of the forecast period includes the instantaneous outflow water flow at each forecasting moment in the future period; Generate a water intake plan for the forecast period according to the water demand for the forecast period and the target liquid level setting of the water reservoir; the water intake plan for the forecast period includes the water flow rate of the inlet plant at each forecast time in the future period; Inputting the modeling index data including at least the water intake scheme, the average value of raw water quality, the set value of process water quality and the set value of factory water quality, together with the statistical characteristics of the modeling index, into the dosing model to obtain the dosing amount of the forecast period; the dosing amount of the forecast period is the dosing amount at each forecast time in a future period of time; According to the dosage of the predicted period and the control rules of the agent dosing equipment, determine the dosing control plan of the predicted period, and generate the dispatching plan of the water plant agent dosing equipment according to the dosing control plan of the predicted period; The dosing scenario model receives modeling index data in real time, and inputs the modeling index data including at least the instantaneous flow rate of inlet water, raw water quality, process water quality and outlet water quality collected in real time, together with the statistical characteristics of the modeling index, into the dosing model to obtain a real-time predicted dosing amount; According to the real-time predicted dosage and the control rules of the chemical dosing equipment, the real-time dosage is determined, and based on the scheduling plan of the chemical dosing equipment in the water plant, the real-time dosage is sent to the chemical dosing equipment to adjust the dosage.

2. The adaptive intelligent digital dosing method based on water demand prediction according to claim 1 is characterized in that: Create a dosing scenario model, including: Select the dosing scenario on the model building platform; configuring a modeled agent in the dosing scenario; According to the modeled agent, the operating indicators and the associated agent dosing equipment are bound in the dosing scenario, the expected target of dosing is set, and the data collection frequency, data length, and data volume verification of the operating indicators are set, and the pump frequency limit, usage status, usage priority and / or pump flow limit of the associated agent dosing equipment are set to obtain a dosing scenario model; the associated agent dosing equipment supports configuration of dosing strategies including over-limit strategy, pump frequency recommendation strategy and / or pump flow recommendation strategy; the over-limit strategy is to limit the frequency allocated to the pump to not exceed the upper limit of the pump frequency, or the maximum flow of the pump to not exceed the upper limit of the pump flow; the pump frequency recommendation strategy refers to determining the start and stop combination of the pump and the frequency allocated to each pump based on the over-limit strategy; the pump flow recommendation strategy refers to determining the start and stop combination of the pump and the flow allocated to each pump based on the over-limit strategy.

3. The adaptive intelligent digital dosing method based on water demand prediction according to claim 1 is characterized in that: The historical data of the operation indicators and the historical data of the dosage of the water plant in the actual water treatment process are received through the dosing scenario model, specifically including: Consider the time delay difference of different process operations in the water treatment process and configure the time delay coefficient; Receive historical data of the water plant's operating indicators and historical data of the cumulative amount of reagent addition during the actual water treatment process through the dosing scenario model; Subtract the historical data of the cumulative amount of the medicine added at adjacent receiving moments to obtain the historical data of the amount of medicine added at each receiving moment; According to the delay coefficient, the historical data of the operation index and the historical data of the dosage are aligned with each other in a time delay.

4. The adaptive intelligent digital dosing method based on water demand prediction according to claim 1 is characterized in that: A correlation analysis is performed on the operation index and the dosage, and a modeling index is selected from the operation index according to the correlation analysis result, and the historical data of the modeling index, the statistical characteristics of the modeling index and the historical data of the dosage form a sample set, specifically including: Determine the statistical characteristics of the operation index; the statistical characteristics include at least maximum value, minimum value, mean value and variance; According to the historical data of the operating indicators, the statistical characteristics of the operating indicators and the historical data of the dosage, the correlation analysis between the operating indicators and the dosage is performed to obtain the correlation coefficient of the operating indicators; Selecting the operating indicators whose correlation coefficients are greater than the preset coefficient threshold as modeling indicators; After data cleaning of historical data of modeling indicators, statistical characteristics of modeling indicators and historical data of dosage, a sample set is formed.

5. The adaptive intelligent digital dosing method based on water demand prediction according to claim 1 is characterized in that: The sample set is used to train a neural network model to obtain a dosing model, and then the following steps are further included: Configure the self-learning trigger conditions and self-learning end conditions; When the self-learning trigger condition is triggered, incremental learning of the dosing model is performed according to the newly generated time series data of the operation indicators and dosing amount of the water plant during the actual water treatment process; If the dosing model after incremental learning meets the self-learning end condition, the dosing model is switched to the dosing model after incremental learning; the switching includes automatic switching and manual switching; If the dosing model after incremental learning does not meet the self-learning end condition, return to step "perform incremental learning of the dosing model according to the newly generated time series data of the water plant's operating indicators and dosing amount during the actual water treatment process".

6. The adaptive intelligent digital dosing method based on water demand prediction according to claim 1 is characterized in that: The water demand in the forecast period is determined by the water demand forecasting time series model, including: The seasonal data and weather data corresponding to the forecast period are input into the water demand forecasting time series model to obtain the water demand of the forecast period; the seasonal data at least includes holidays, weeks, months and quarters; the weather data at least includes weather and temperature.

7. The adaptive intelligent digital dosing method based on water demand prediction according to claim 1 is characterized in that: According to the dosage of the predicted period and the control rules of the agent dosing equipment, determine the dosing control plan for the predicted period, including: Setting control rules for the drug dosing device; there are at least two control rules for the drug dosing device, one is that the drug dosing device controls the drug dosing amount by setting the dosing pump frequency, and the other is that the drug dosing device controls the drug dosing amount by setting the dosing pump flow rate; When the control rule of the drug dosing equipment is that the drug dosing equipment controls the dosage by setting the dosing pump frequency, a linear relationship between the dosing pump frequency and the dosage is fitted; According to the dosage of the predicted period, based on the linear relationship between the dosage pump frequency and the dosage, the dosage pump frequency of the predicted period is obtained and determined as the dosage control scheme; When the control rule of the drug dosing equipment is that the drug dosing equipment controls the dosage by setting the dosing pump flow rate, the flow rate of each pump is directly set according to the dosage of the prediction period, and the flow rate of each pump in the prediction period is determined as the dosing control plan; wherein, the total of all pump flows in the prediction period is the total dosage of the prediction period.

8. The adaptive intelligent digital dosing method based on water demand prediction according to claim 7 is characterized in that: According to the dosage of the predicted period, based on the linear relationship between the dosage pump frequency and the dosage, the dosage pump frequency of the predicted period is obtained and determined as the dosage control scheme, which also includes: Set the fitting period; In each fitting cycle, the linear relationship between the dosing pump frequency and the dosing amount is refitted according to the latest operating status of the dosing equipment and the dosing pump.

9. The adaptive intelligent digital dosing method based on water demand prediction according to claim 1 is characterized in that: According to the real-time predicted dosage and the control rules of the dosing equipment, the real-time dosage is determined, and based on the scheduling plan of the water plant's dosing equipment, the real-time dosage is sent to the dosing equipment to adjust the dosage, and then it also includes: According to the actual operation status in the dosing control process, the compensation dosage required for the next prediction cycle is calculated through the dosage compensation mechanism, and / or the feedback dosage required for the next prediction cycle is calculated through the feedback mechanism; The dosage compensation mechanism is as follows: if the dosage of the predicted period is inconsistent with the actual dosage during the dosage control process, the difference between the dosage of the predicted period and the actual dosage during the dosage control process is calculated as the compensation dosage, and the dosage is supplemented or reversely compensated according to the compensation dosage in the next predicted period; The feedback mechanism is: if the process water quality does not meet expectations, and / or the factory water quality does not meet expectations, a protective dosage feedback adjustment is performed to cope with the abnormal water quality; the protective dosage feedback adjustment refers to determining the feedback dosage required for the next prediction cycle based on the real-time data of the operating indicators during the dosing control process and the relationship between the operating indicator data and the dosage.

10. An adaptive intelligent digital dosing system based on water demand prediction, characterized in that: The adaptive intelligent digital dosing system based on water demand prediction supports local deployment and cloud deployment; the adaptive intelligent digital dosing system based on water demand prediction executes the adaptive intelligent digital dosing method based on water demand prediction described in any one of claims 1-9.

Images