Low-carbon operation and maintenance method for sewage treatment plant during spring festival transportation based on mobile phone signaling

Abstract

The present application relates to a kind of based on mobile phone signaling during Spring Festival travel gradient shutdown type low-carbon operation and maintenance method, belong to sewage treatment intelligent control and energy-saving optimization technical field, including: the multi-source data base of fusing mobile phone signaling, sewage plant basic operation data, weather and time sequence characteristics is constructed;Signaling data is converted into high spatial and temporal resolution grid dynamic pollution load map;Introduce the long short-term memory network model of feature-time double attention mechanism, predict future inflow load;According to the prediction result, formulate and execute gradient shutdown strategy that is adaptively matched with load fluctuation;Based on energy consumption and material consumption quantitative evaluation reduces carbon effect.The present application realizes the high-precision quantitative tracking of dynamic pollution load by the deep integration of multi-source data fusion and intelligent decision-making technology, significantly improves the capture ability and prediction accuracy of the model on the special load law during Spring Festival travel, and provides a reliable technical path for realizing the low-carbon and efficient operation of sewage plant in special period.

Description

Technical Field

[0001] This invention relates to a low-carbon operation and maintenance method for wastewater treatment plants during the Spring Festival travel rush based on mobile phone signaling, belonging to the field of intelligent control and energy-saving optimization technology for wastewater treatment. Background Technology

[0002] During periods of large-scale population migration, such as the Spring Festival travel rush, wastewater treatment plants often operate under high-carbon emission conditions for extended periods due to a severe mismatch between dynamic influent load and static operating modes. The root cause lies in the fact that current operational control systems generally rely on static population data and historical load patterns, failing to accurately perceive and respond to real-time load changes caused by short-term population movements. This results in key energy-consuming units such as aeration and booster pumps operating outside their efficient operating range for extended periods, leading not only to energy waste but also directly increasing carbon emission intensity.

[0003] Traditional wastewater treatment optimization methods primarily focus on process improvements and equipment upgrades. However, when dealing with short-term large-scale population migration scenarios such as the Spring Festival travel rush, they lack both high-resolution spatiotemporal dynamic population monitoring methods and corresponding flexible control strategies. In practice, some wastewater treatment plants have attempted to cope with sudden load drops using crude control methods such as "complete plant shutdown" or "single unit shutdown." However, these strategies rely heavily on manual experience and lack scientific data support and forward-looking predictions, often resulting in inappropriate timing of control measures: shutting down facilities too early may lead to overflow pollution due to insufficient treatment capacity, affecting effluent quality standards; shutting down too late may result in operating units remaining at low loads for extended periods before shutdown, leading to low energy efficiency. Furthermore, due to the lack of accurate prediction of load recovery timing, facility restarts are often delayed or premature, further exacerbating system instability and carbon emission intensity. Although existing research indicates that optimizing influent load allocation can improve system energy efficiency, the lack of precise quantification methods for dynamic pollutant-generating populations has prevented the development of systematic and scalable technical solutions for corresponding control strategies. Summary of the Invention

[0004] This invention aims to address the problem that during periods of large-scale and drastic population fluctuations, such as the Spring Festival travel rush, wastewater treatment plants cannot detect changes in influent load in real time, leading to rigid operating modes and long-term inefficient operation of key equipment, resulting in serious energy waste and excessive carbon emissions. The invention proposes a low-carbon operation and maintenance method for wastewater treatment plants during the Spring Festival travel rush based on mobile phone signaling, which involves a gradual shutdown.

[0005] The technical solution of the present invention:

[0006] A low-carbon operation and maintenance method for wastewater treatment plants during the Spring Festival travel rush, based on mobile phone signaling, includes: S1. Construct a multi-source data foundation for low-carbon operation and maintenance of sewage treatment plants during the Spring Festival travel rush: integrate multi-source data including mobile phone signaling data, sewage treatment plant basic data, sewage treatment plant monitoring data, external environmental data, and time-series characteristics; S2. Based on mobile phone signaling, perform gridded dynamic population quantification during the Spring Festival travel rush and calculate the gridded dynamic sewage output. S3. Construct a wastewater treatment plant influent prediction model during the Spring Festival travel rush by introducing an attention mechanism: Utilize a long short-term memory network model that integrates feature-time dual attention mechanism, using the multi-source data from step S1 and the dynamic wastewater production obtained in step S2 as input, to predict the influent volume at a specified time step in the future. S4. Formulate and implement a low-carbon operation and maintenance strategy for the gradual shutdown of the wastewater treatment plant during the Spring Festival travel season: Based on the predicted influent volume obtained in step S3, calculate the real-time or predicted average influent load rate, and based on the preset shutdown threshold range, make decisions and execute the orderly shutdown or restart of the corresponding group of treatment facilities according to the shutdown decision function, so that the number of operating facilities is adaptively matched with the influent load. S5. Conduct carbon reduction effect assessment: Based on the actual energy and material consumption data before and after the implementation of the low-carbon operation and maintenance strategy of the wastewater treatment plant during the Spring Festival travel season, the total carbon emissions are quantitatively assessed using the carbon emission factor method.

[0007] Specifically, in step S1,

[0008] Mobile signaling data: includes population dynamics data such as user ID, timestamp, base station latitude and longitude, and statistics on the city's permanent resident population;

[0009] Basic data of wastewater treatment plants: including the designed treatment capacity of the wastewater treatment plant, the service area boundary delineated by GIS technology, the type of core biological treatment process, and the number of treatment facilities operating in parallel;

[0010] Wastewater treatment plant monitoring data includes historical and real-time hourly influent flow rates, as well as at least one key water quality indicator from chemical oxygen demand, ammonia nitrogen, total nitrogen, total phosphorus, and suspended solids.

[0011] External environmental data: including meteorological observation data such as temperature and rainfall;

[0012] Temporal features: including month, date, day of the week, hour, and day of the Spring Festival travel rush.

[0013] Specifically, step S2 includes the following process:

[0014] S2.1. Perform noise reduction and user deduplication on the multi-source data, and filter the number of unique users in the target age group;

[0015] S2.2. Expand the population sample based on mobile phone penetration rate and data provider market share to obtain a full-quantitative dynamic population estimate.

[0016] S2.3. Introduce a pollution-generating path allocation mechanism, combining base station spatial coverage weight and user activity weight, to dynamically allocate the population to each grid, thus obtaining a gridded pollution-generating population.

[0017] S2.4 Divide the study area into regular grids and calculate the dynamic wastewater production of each grid.

[0018] Specifically, in step S3,

[0019] The prediction model is a long short-term memory network model that incorporates a feature-time dual attention mechanism; wherein, the feature attention mechanism is used to dynamically assign weights to the input features, focusing on dynamic sewage production; and the time attention mechanism is used to capture key fluctuation patterns in historical sequences that are related to the Spring Festival travel cycle. The long short-term memory network model with the feature-time dual attention mechanism adopts an encoder-decoder sequence-to-sequence architecture. The input features include at least wastewater treatment plant monitoring data, external environmental data, time-series features, and dynamic wastewater production. The output is the predicted influent volume for multiple future time steps.

[0020] Specifically, in step S4, the process of determining and executing the orderly shutdown or restart of the corresponding group of processing facilities based on a preset shutdown threshold range and a shutdown decision function includes:

[0021] Real-time calculation or prediction of the average influent load rate of a wastewater treatment plant with N sets of parallel and consistent treatment facilities;

[0022] A set of shutdown thresholds corresponding to the number of facility groups is preset;

[0023] When the predicted future inflow load rate falls into a certain threshold range, the system automatically triggers the shutdown of the corresponding number of k groups of facilities, so that the load rate of the remaining (Nk) groups of facilities is increased to the high-efficiency operating range;

[0024] When the load rate is detected to rise above the set recovery threshold, the shut-down facilities are started up in sequence to achieve a gradual recovery of processing capacity.

[0025] Specifically, the shutdown decision is implemented through the following function:

[0026]

[0027] in, This represents the status of the k-th facility group, where 1 indicates operation and 0 indicates shutdown. The minimum operating load for a single facility; t is time; The forecast time window is N; N is the total number of facility groups. Let i be the load factor at time i; This is the load fluctuation correction factor, used to account for load forecasting errors and operational uncertainties; its value range is... .

[0028] Specifically, in step S5, the total carbon emissions... Calculated using the following formula:

[0029]

[0030] in, To assess the carbon dioxide equivalent generated by the wastewater treatment plant during the assessment period. Total emissions; This represents the power consumption of the i-th main power-consuming unit during this period, expressed in kWh. For corresponding The carbon emission factor of the electricity consumed, in units of ; This represents the consumption of the j-th externally applied chemical agent during this period, expressed in kg. Let be the carbon emission factor of the j-th agent during its life cycle, in units of . .

[0031] Specifically, the mobile signaling data is aggregated statistical data that has been legally authorized and anonymized, and does not contain any personal information that can identify a specific natural person.

[0032] The beneficial effects of this invention are:

[0033] This invention, through the deep integration of multi-source data fusion and intelligent decision-making technology, has revolutionized the operation mode of wastewater treatment systems under special population mobility scenarios, achieving significant low-carbon optimization results, specifically reflected in the following aspects:

[0034] (1) High-precision quantitative tracking of dynamic pollution load has been achieved. Traditional methods rely on outdated statistical population data, which cannot effectively capture the instantaneous and dramatic changes in sewage load caused by large-scale population migration during the Spring Festival travel rush. Based on mobile phone signaling data, this invention establishes a complete dynamic population analysis and sewage production estimation process. Through dynamic population expansion, spatial rasterization mapping, and dynamic correction of the pollution production coefficient that comprehensively considers seasonal and weather factors, it achieves a precise spatiotemporal dynamic characterization of sewage production in the service area, providing unprecedented real-time data support for operation and control.

[0035] (2) An intelligent water inflow load prediction system for special periods was developed. This invention constructs a long short-term memory network model that integrates an attention mechanism. This model can autonomously identify key fluctuation patterns in historical water inflow sequences and assign higher prediction weights to these characteristic periods. This mechanism significantly improves the model's ability to capture and predict the unique load patterns during the Spring Festival travel rush, enabling reliable forecasting of water inflow volume and quality for the next 24 hours, and providing crucial forward-looking information for the early deployment of operational strategies.

[0036] (3) A gradient shutdown operation strategy adaptively matched with load fluctuations was established. This invention abandons the traditional experience-based control method and designs a gradient shutdown decision function based on load rate range discrimination. The system automatically triggers the orderly shutdown of facilities by comparing the predicted load with the preset threshold in real time, ensuring that the number of operating facilities is always in an optimal matching state with the influent load, thereby stably guiding the treatment units from the inefficient operation range to the efficient operation range. This strategy directly and significantly reduces the system's energy and material consumption through operation optimization, and establishes a complete closed loop for evaluating carbon reduction effects, providing a reliable technical path for achieving low-carbon and efficient operation of wastewater treatment plants during special periods. Attached Figure Description

[0037] Figure 1 This is a technical roadmap for the present invention;

[0038] Figure 2 A process for creating a rasterized population map based on mobile signaling;

[0039] Figure 3 The process of constructing a wastewater influent prediction model for the Spring Festival travel rush by introducing an attention mechanism;

[0040] Figure 4 Prediction results of a wastewater treatment plant influent prediction model during the Spring Festival travel rush, incorporating an attention mechanism;

[0041] Figure 5 This is a schematic diagram illustrating the tiered shutdown strategy during the Spring Festival travel rush.

[0042] Figure 6 This study outlines strategies for shutting down and restarting wastewater treatment plants and their corresponding carbon reduction effects. Detailed Implementation

[0043] All mobile signaling data used in this invention originates from anonymized aggregated statistical data provided by legally qualified third-party data service providers and does not contain any personally identifiable information. All data processing is based on regional-level statistical results and does not involve any individual-level data analysis; therefore, there is no infringement on citizens' privacy rights.

[0044] Example 1:

[0045] like Figure 1 As shown, a low-carbon operation and maintenance method for wastewater treatment plants during the Spring Festival travel rush, based on mobile phone signaling, includes:

[0046] S1. Building a data foundation for low-carbon operation and maintenance of wastewater treatment plants during the Spring Festival travel rush:

[0047] The construction of the data foundation is the basis for the implementation of this invention. This foundation integrates four main categories of data: First, population dynamic data, with mobile phone signaling data containing user IDs, timestamps, and base station latitude and longitude as the core, supplemented by urban resident population statistics; second, basic data of wastewater treatment plants, including the designed treatment capacity of each plant, the service area boundaries precisely delineated through GIS technology, the type of core biological treatment process adopted, and the exact number of parallel-operating treatment facilities; third, wastewater treatment plant monitoring data, covering historical and real-time hourly influent flow, as well as key water quality indicators such as chemical oxygen demand, ammonia nitrogen, total nitrogen, total phosphorus, and suspended solids; and finally, external environmental data, including meteorological observation data such as temperature and rainfall, and temporal context features such as month, date, day of the week, hour, and the day of the Spring Festival travel rush provided for model construction. All data underwent missing value imputation, outlier handling, coordinate system unification, and timestamp standardization before being entered into the database.

[0048] S2. Gridded dynamic population quantification during the Spring Festival travel rush based on mobile phone signaling:

[0049] This step aims to transform mobile signaling data into a dynamic pollution load map with high spatiotemporal resolution.

[0050] First, the raw signaling data is preprocessed. A distributed computing framework is used to reduce noise and filter invalid records; user deduplication is performed to ensure individual uniqueness; and an age inference model based on behavioral characteristics is applied to identify the number of unique users in the primary pollution-generating age group of 16-55 years old.

[0051]

[0052] Where t represents time; c represents the base station; Let t be the effective user base of base station c; This represents the original number of signaling users. This is the deduplication factor (considering the case of one person having multiple SIM cards); This is the age screening coefficient (percentage of those aged 16-55).

[0053] Subsequently, precise calculation and spatial allocation of the polluting population are performed. The following two formulas are used to extrapolate from the sample to the population and to address the issue of polluting location distribution caused by user movement:

[0054]

[0055] in, For a moment Base stations Full-scale population estimate within the coverage area; The mobile phone penetration rate for the 16-55 age group is used to expand the number of mobile phone users to the actual population in that age group. The market share of the data provider in the target region is used to further expand the sample population to the entire population;

[0056] Next, a pollution-generating pathway allocation mechanism is introduced to assign the population to spatial grids where they are likely to generate wastewater:

[0057]

[0058] in, For time t, estimate the polluting population within the geographic grid g; The spatial coverage weight of base station c on grid g; Let C be the user activity weight at time t for base station c; C is the total number of base stations.

[0059] Finally, dynamic wastewater production was calculated. The study area was divided into a 1km × 1km regular grid. The aforementioned wastewater-generating population data was rasterized, and the hourly wastewater generation coefficient was applied to calculate the wastewater generation of each grid cell.

[0060]

[0061] in, Let g be the wastewater output of grid g at time t; The baseline hourly average pollution per capita coefficient; This is a seasonal correction factor; This is a weather correction factor.

[0062] S3. Construction of a wastewater influent prediction model for the Spring Festival travel rush, incorporating an attention mechanism:

[0063] To accurately capture the abnormal fluctuations in water load during the Spring Festival travel rush, this invention proposes an innovative Long Short-Term Memory Network (FTA-LSTM) prediction model with a dual feature-temporal attention mechanism. The innovation of this model lies in its dynamic weight allocation of input variables through a feature attention mechanism, enabling the model to autonomously focus on key load features derived from dynamic population quantification steps. Simultaneously, it accurately captures key fluctuation patterns related to the Spring Festival travel rush cycle in historical sequences through a temporal attention mechanism, thus achieving deep perception and memorization of load patterns during special periods within the model. The model is built on the TensorFlow 2.20.0 deep learning framework, with the training and test sets divided temporally in a 7:3 ratio to ensure the continuity of the time series is not disrupted. The model adopts an encoder-decoder sequence-to-sequence (Seq2Seq) architecture, with an input time window of 24 time steps and an output prediction window of 24 time steps, achieving multi-step rolling prediction.

[0064] The core architecture of the model consists of three key components: the encoder LSTM network adopts a two-layer stacked structure with 128 LSTM units per layer, the activation function is tanh, and the dropout rate is set to 0.2 to prevent overfitting; the decoder LSTM network also adopts a two-layer 128-unit structure, designed symmetrically with the encoder; the feature attention module is located at the encoder input, and adaptively calculates the weight allocation of each input feature through 8 attention heads, focusing on the rasterized pollution load features derived from the dynamic population quantization step; the temporal attention module connects the encoder and decoder, and identifies the most critical time step in the historical sequence for the current prediction through 4 attention heads, effectively capturing the load impact pattern of the same period of the Spring Festival travel rush in previous years.

[0065] The model training uses the Adam optimizer with an initial learning rate of 0.001 and an exponential decay strategy, with a decay rate of 0.9 every 10 epochs. The gradient clipping threshold is set to 5.0 to prevent gradient explosion. The loss function is a weighted combination of mean squared error (MSE) and mean absolute error (MAE) (weight ratio 7:3). The batch size is set to 32, the maximum number of training epochs is 200, and an early stopping mechanism (patience=15) is used to prevent overfitting. The weights are initialized using the Xavier uniform initialization method.

[0066] The attention mechanism is implemented through adaptive weight allocation based on the model's internal states. First, the correlation between the encoder's hidden states at each time step and the decoder's current state is calculated. Then, the correlation score is converted into attention weights using a softmax function. The specific formula is as follows:

[0067]

[0068] in, Let t be the attention weight of time t to historical time i; Let softmax() be the attention score at time t for historical time i, and let softmax() be the normalization function that converts the attention score into a probability distribution. The current hidden state; The hidden state of historical moment i; Let j be the hidden state at historical time j; score() is the attention scoring function.

[0069] Based on the calculated attention weights, a context vector containing key historical information is generated:

[0070]

[0071] in, is the context vector at time t; T is the historical time step.

[0072] This vector is essentially a weighted summary of historical sequences, highlighting the historical periods that are most valuable for current predictions.

[0073] The current state is combined with the context vector to generate the final predicted value.

[0074]

[0075] in, f() is the predicted inflow load for the next time step; W and b are the output layer weights and biases, respectively. This is a concatenation of the hidden state and the context vector.

[0076] The model's input feature system includes historical inflow sequence, daily average population statistics, rainfall, temperature, and other meteorological indicators, as well as feature dimensions generated through feature engineering, such as 7-day lag features, 14-day moving average features, seasonal features, and time-series interaction features. The output is a predicted inflow value for the next 12 time steps, providing a reliable basis for subsequent gradient shutdown decisions. This step ensures that the prediction results consider both the current system state and fully integrate historical experience.

[0077] S4. Formulation of a low-carbon operation and maintenance strategy for wastewater treatment plants during the Spring Festival travel rush:

[0078] Based on the dynamic wastewater production and influent load forecasts obtained from the preceding steps, a refined "gradual shutdown" strategy is formulated. The implementation process of this strategy is as follows: For a design scale of... For a wastewater treatment plant with N sets of parallel and identical treatment facilities, the average influent load rate should first be calculated or predicted in real time:

[0079]

[0080] in, Let be the load factor at time t; Let be the total inflow at time t; To design the processing scale.

[0081] To optimize the operational efficiency of multiple processing facilities, a set of shutdown thresholds is preset. (For example, for four sets of facilities, it is possible to set up) , , When the cumulative inflow load for the next 24 hours is predicted to fall within a certain range, the system will shut down the corresponding group of facilities in a preset sequence, thus reducing the number of remaining operating facilities. Matching the current inflow load, the system guides facilities from inefficient to efficient operating ranges. As the population returns during the later stages of the Spring Festival travel rush, the inflow load gradually increases. Through real-time monitoring and prediction, when the load rate exceeds the set recovery threshold, the system will sequentially activate the shut-down facilities to achieve a gradual recovery of processing capacity, ensuring that the system always operates within the efficient load range.

[0082] The shutdown decision function is defined as follows:

[0083]

[0084] in, The status of the kth group of facilities is (1 in operation, 0 out of operation). This represents the minimum operating load for a single facility. is the prediction time window; N is the total number of facility groups.

[0085] S5. Carbon Reduction Effect Assessment:

[0086] To quantify the carbon emission reduction after the implementation of this method, a carbon emission factor method based on actual energy and material consumption is used for assessment. The total carbon emissions of a wastewater treatment plant within a specific assessment period are then calculated. Calculated using the following formula:

[0087]

[0088] in, The carbon dioxide equivalent generated by the wastewater treatment plant during the assessment period. Total emissions; This represents the power consumption of the i-th main power-consuming unit (such as the influent pump, aerator, sludge dewatering machine, etc.) during this period, expressed in kilowatt-hours (kWh). For corresponding The carbon emission factor of the electricity consumed, in units of ; This represents the consumption of the j-th type of externally applied chemical agent (such as a carbon source, phosphorus removal agent, etc.) during this cycle, expressed in kilograms (kg). Let be the carbon emission factor of the j-th agent during its life cycle, in units of . .

[0089] Example 2:

[0090] This embodiment uses actual operational data from a wastewater treatment plant in Shenzhen, Guangdong Province, during the 2024 Spring Festival travel rush to provide a detailed description of the invention. This embodiment fully implements the entire process from dynamic population quantification, influent load prediction, tiered shutdown decisions to carbon reduction effect evaluation.

[0091] (1) Construction of dynamic raster map of polluting population

[0092] like Figure 2 As shown, mobile signaling data from January to February 2024 was collected within the service area of ​​the plant, including user IDs, timestamps, and base station location information. The Spark distributed computing framework was used to denoise and deduplicate the raw data, and an age inference model was used to filter out effective users in the main pollution-generating age group of 16-55 years old. Population expansion was performed based on regional mobile phone penetration rate and data provider market share to obtain comprehensive dynamic population data. Subsequently, a mapping relationship between base stations and geographic rasters was established, and spatial allocation was performed by combining user activity weights to generate a 1km×1km resolution rasterized pollution-generating population map. Finally, dynamic pollution generation coefficients were applied to calculate the wastewater output of each raster, forming an hourly pollution load distribution map. This process fully demonstrates the transformation from raw signaling data to pollution load space.

[0093] (2) Construction and validation of LSTM prediction model based on attention mechanism

[0094] like Figure 3 As shown, to address the abnormal fluctuations in water inflow during the Spring Festival travel rush, a long short-term memory neural network prediction model incorporating an attention mechanism was constructed. This model takes historical water inflow sequences, meteorological data, and dynamic population characteristics as inputs. An attention layer automatically identifies key periods in the historical sequences that significantly influence future load predictions and assigns them higher weights. The model employs an encoder-decoder structure. The encoder extracts the temporal dependencies of the input features, while the attention mechanism establishes a direct connection between the encoder and decoder, enabling the model to focus on historical information most relevant to the current prediction.

[0095] like Figure 4As shown, this model was applied to predict the inflow load of the plant during the 2024 Spring Festival travel rush, with the inflow rate over the next 24 hours as the prediction target. The model training used mean squared error as the loss function, and the Adam optimizer was used for parameter optimization. Validation results show that the model achieved excellent predictive performance on the test set, with a coefficient of determination R² of 0.9551, a mean absolute error (MAE) of 0.1896, a root mean square error (RMSE) of 0.2441, and a mean absolute percentage error (MAPE) of only 1.46%, demonstrating its accurate ability to capture the special load patterns of the Spring Festival travel rush.

[0096] (3) Implementation parameters of gradient shutdown strategy

[0097] like Figure 5 As shown, a phased shutdown strategy is implemented based on accurate load forecasting results. The plant has a design capacity of 200,000 tons / day and is equipped with four parallel and identical processing facilities. Shutdown thresholds are set. , , When the predicted average load rate for the next 24 hours falls within a certain range, the corresponding group of facilities is shut down. By monitoring load rate changes in real time, the system automatically executes shutdown or restart commands to ensure that the number of operating facilities always matches the actual processing demand.

[0098] (4) Evaluation of carbon reduction effect

[0099] like Figure 6 As shown, to evaluate the decarbonization benefits of the tiered shutdown strategy, this embodiment systematically compares and analyzes the number of operating facilities, power load distribution, and chemical usage based on population dynamics, load forecasts, and operational records during the Spring Festival travel rush. Under the traditional model, the plant maintained all four sets of facilities running at full capacity before and after the Spring Festival travel rush, making it difficult to adapt to the decrease in influent caused by the rapid outflow of population. Major energy-consuming units such as aeration and booster pumps operated at low loads for several days, and the chemical dosage was difficult to adjust in a timely manner according to load changes. After implementing the tiered shutdown strategy, the number of operating facilities was adjusted step by step according to the load, gradually decreasing from four sets to one to three sets during the low-load period from January 17th to January 24th, and synchronously resuming operation according to the forecast results during the load recovery period. This kept the load level of the operating units within a more reasonable range, significantly reducing energy waste under low load conditions.

[0100] Quantitative results show that during the 41-day Spring Festival travel rush, the cumulative operating scale under the traditional model was 4 groups / day × 41 days, totaling 164 groups / day, while the actual operating scale under the gradient shutdown strategy was 97 groups / day, a reduction of 40.9% in operating units. During the lowest load phase, maintaining four groups of operation resulted in a single group load of only about 0.16–0.60, while after implementing gradient shutdown, the single group load increased to about 0.50–0.97, improving load utilization efficiency by 35–70%. With reduced ineffective operating time, improved load matching, and optimized simultaneous chemical dosing, the normalized carbon emissions during the Spring Festival travel rush decreased by approximately 35–45%. The results demonstrate that this invention can significantly reduce ineffective energy consumption, improve the operating efficiency of key equipment, and achieve substantial pollution reduction and carbon reduction benefits while ensuring stable effluent compliance. Through the above implementation process, this invention successfully achieved refined and low-carbon operation and maintenance of a wastewater treatment plant in Shenzhen during the Spring Festival travel rush, providing reliable technical support for the plant's energy conservation, consumption reduction, and stable operation during this period of special population movement. This invention has strong versatility and applicability, and is applicable to other wastewater treatment plants with similar conditions.

Claims

1. A low-carbon operation and maintenance method for wastewater treatment plants during the Spring Festival travel rush based on mobile phone signaling, characterized in that... include: S1. Construct a multi-source data foundation for low-carbon operation and maintenance of sewage treatment plants during the Spring Festival travel rush: integrate multi-source data including population dynamics data, sewage treatment plant basic data, sewage treatment plant monitoring data, external environmental data, and time-series characteristics; Population dynamics data includes mobile signaling data and statistics on the city's permanent resident population; Mobile signaling data: includes population dynamics data such as user ID, timestamp, and base station latitude and longitude; The mobile signaling data is aggregated statistical data that has been legally authorized and has undergone anonymization and desensitization processing, and does not contain any personal information that can identify a specific natural person; S2. Based on mobile phone signaling, perform gridded dynamic population quantification during the Spring Festival travel rush and calculate the gridded dynamic sewage output. S3. Construct a wastewater treatment plant influent prediction model during the Spring Festival travel rush by introducing an attention mechanism: Utilize a long short-term memory network model that integrates feature-time dual attention mechanism, using the multi-source data from step S1 and the dynamic wastewater production obtained in step S2 as input, to predict the influent volume at a specified time step in the future. S4. Formulate and implement a low-carbon operation and maintenance strategy for the gradual shutdown of the wastewater treatment plant during the Spring Festival travel season: Based on the predicted influent volume obtained in step S3, calculate the real-time or predicted average influent load rate, and based on the preset shutdown threshold range, make decisions and execute the orderly shutdown or restart of the corresponding group of treatment facilities according to the shutdown decision function, so that the number of operating facilities is adaptively matched with the influent load. S5. Conduct carbon reduction effect assessment: Based on the actual energy and material consumption data before and after the implementation of the low-carbon operation and maintenance strategy of the wastewater treatment plant during the Spring Festival travel season, the total carbon emissions are quantitatively assessed using the carbon emission factor method.

2. The method for low-carbon operation and maintenance of sewage treatment plants during the Spring Festival travel rush based on mobile phone signaling as described in claim 1, characterized in that, In step S1, the basic data of the wastewater treatment plant includes: the designed treatment capacity of the wastewater treatment plant, the service area boundary delineated by GIS technology, the type of core biological treatment process, and the number of treatment facility groups operating in parallel. Wastewater treatment plant monitoring data includes historical and real-time hourly influent flow rates, as well as at least one key water quality indicator from chemical oxygen demand, ammonia nitrogen, total nitrogen, total phosphorus, and suspended solids. External environmental data: including meteorological observation data such as temperature and rainfall; Temporal features: including month, date, day of the week, hour, and day of the Spring Festival travel rush.

3. The method for low-carbon operation and maintenance of sewage treatment plants during the Spring Festival travel rush based on mobile phone signaling as described in claim 2, characterized in that, Step S2 specifically includes the following processes: S2.

1. Perform noise reduction and user deduplication on the multi-source data, and filter the number of unique users in the target age group; S2.

2. Expand the population sample based on mobile phone penetration rate and data provider market share to obtain a full-quantitative dynamic population estimate. S2.

3. Introduce a pollution-generating path allocation mechanism, combining base station spatial coverage weight and user activity weight, to dynamically allocate the population to each grid, thus obtaining a gridded pollution-generating population. S2.4 Divide the study area into regular grids and calculate the dynamic wastewater production of each grid.

4. The method for low-carbon operation and maintenance of sewage treatment plants during the Spring Festival travel rush based on mobile phone signaling as described in claim 3, characterized in that, In step S3, The prediction model is a long short-term memory network model that incorporates a feature-time dual attention mechanism; wherein, the feature attention mechanism is used to dynamically assign weights to the input features, focusing on dynamic sewage production; and the time attention mechanism is used to capture key fluctuation patterns in historical sequences that are related to the Spring Festival travel cycle. The long short-term memory network model with the feature-time dual attention mechanism adopts an encoder-decoder sequence-to-sequence architecture. The input features include at least wastewater treatment plant monitoring data, external environmental data, time-series features, and dynamic wastewater production. The output is the predicted influent volume for multiple future time steps.

5. The method for low-carbon operation and maintenance of sewage treatment plants during the Spring Festival travel rush based on mobile phone signaling as described in claim 4, characterized in that, In step S4, the real-time or predicted average influent load rate is calculated. The process of determining and executing the orderly shutdown or restart of the corresponding group of treatment facilities based on the preset shutdown threshold range and the shutdown decision function includes: Real-time calculation or prediction of a person with N The average influent load rate of a wastewater treatment plant with parallel and consistent treatment facilities; A set of shutdown thresholds corresponding to the number of facility groups is preset; When the predicted future inflow load rate falls within a certain threshold range, the system automatically triggers the shutdown of the corresponding number of [unspecified facilities / systems]. k The group's facilities enable the remaining operations ( Nk The load factor of the facilities has been increased to the high-efficiency operating range; When the load rate is detected to rise above the set recovery threshold, the shut-down facilities are started up in sequence to achieve a gradual recovery of processing capacity.

6. The method for low-carbon operation and maintenance of sewage treatment plants during the Spring Festival travel rush based on mobile phone signaling as described in claim 5, characterized in that, The shutdown decision is implemented through the following function: ； in, For the first k The status of the facility group is 1 for operation and 0 for shutdown. The minimum operating load for a single facility; t is time; For the prediction time window; N This represents the total number of facility groups. For a moment i Load rate; .

7. The method for low-carbon operation and maintenance of sewage treatment plants during the Spring Festival travel rush based on mobile phone signaling as described in claim 6, characterized in that, In step S5, the total carbon emissions Calculated using the following formula: ； in, To assess the carbon dioxide equivalent generated by the wastewater treatment plant during the assessment period. Total emissions; For the period, the first i The power consumption of each major power-consuming unit, in kWh; For corresponding Carbon emission factor of electricity consumed, in kg ; For the period, the first j The amount of externally applied chemical agents consumed, in kg; For the first j The carbon emission factor of a pesticide over its life cycle, in kg. .

Images