Remote intelligent control method and system for water purifier based on Internet of Things

Abstract

The invention relates to the technical field of smart home, and discloses a water purifier remote intelligent control method and system based on the Internet of Things, and the method comprises the steps: obtaining real-time water consumption data and water purifier state data; carrying out load prediction in combination with a preset long-short-term memory network model to obtain a predicted load sequence; performing dynamic adjustment according to the predicted load sequence and the water purifier state data to obtain an adjustment sequence; performing control instruction optimization according to the real-time operation state data to obtain optimized energy-saving adjustment parameters; filter element efficiency data are obtained according to the energy-saving adjustment parameters, load prediction value updating is carried out based on the filter element efficiency data, and a water consumption behavior mode and an updated load prediction value are obtained; and performing time scheduling and power optimization to obtain a final control scheme. According to the method, the dynamic change of the water consumption behavior of the user can be adapted, and energy-saving and efficient self-adaptive intelligent control is carried out.

Description

Technical Field

[0001] This invention relates to the field of smart home technology, and in particular to a remote intelligent control method and system for a water purifier based on the Internet of Things. Background Technology

[0002] Currently, with the rapid development of smart home and Internet of Things technologies, the demand for intelligent and energy-saving features in household water purifiers, as important equipment to ensure drinking water safety, is becoming increasingly urgent.

[0003] In existing technologies, current water purifier control methods often rely on simple timer switches or fixed power operation, which are difficult to adapt to the dynamic changes in users' water usage behavior. They cannot automatically adjust the operating status according to the actual water consumption of the household, nor do they have in-depth analysis of user behavior patterns. When dealing with complex water usage scenarios, they are unable to achieve precise operation optimization, and the overall control characteristics are static and extensive.

[0004] In summary, existing technologies are ill-suited to adapting to the dynamic changes in users' water usage behavior and are unable to achieve energy-saving and efficient adaptive intelligent control. Summary of the Invention

[0005] This invention provides a remote intelligent control method and system for water purifiers based on the Internet of Things, so as to achieve energy-saving and efficient adaptive intelligent control by adapting to the dynamic changes in users' water use behavior.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a remote intelligent control method for a water purifier based on the Internet of Things, comprising: Obtain real-time water consumption data and water purifier status data; Based on the real-time water consumption data and the water purifier status data, load prediction is performed using a preset long short-term memory network model to obtain a predicted load sequence. The adjustment sequence is obtained by dynamically adjusting the predicted load sequence and the water purifier status data. Real-time operating status data is obtained based on the adjustment sequence, and control commands are optimized based on the real-time operating status data to obtain optimized energy-saving adjustment parameters; The filter efficiency data is obtained based on the energy-saving adjustment parameters, and the load prediction value is updated based on the filter efficiency data to obtain the water use behavior pattern and the updated load prediction value. Based on the water usage behavior pattern and the load forecast, time scheduling and power optimization are performed to obtain the final control scheme.

[0007] In one optional implementation, the step of performing load prediction based on the real-time water consumption data and the water purifier status data, combined with a preset long short-term memory network model, to obtain a predicted load sequence includes: The water consumption data and the water purifier status data are smoothed to obtain a smoothed data sequence. Anomaly detection is performed on the smoothed data sequence and the water purifier status data to obtain the water usage behavior status; The water load is predicted by combining the long short-term memory network model with the water use behavior state to obtain the predicted load sequence.

[0008] In one optional implementation, the step of dynamically adjusting based on the predicted load sequence and the water purifier status data to obtain an adjustment sequence includes: The predicted load sequence is subjected to a threshold judgment. If the load level in the predicted load sequence is less than a preset load threshold, it is marked as a low-peak state, and a low-peak state sequence is obtained. The reduction in equipment operating frequency is calculated based on the off-peak state sequence to obtain the initial adjustment sequence; Extract the filter cartridge pressure status data from the water purifier status data, classify the risk levels, and determine the filter cartridge clogging risk level; The initial adjustment sequence is adapted to the filter element clogging risk level to obtain the adjustment sequence.

[0009] In one optional implementation, the step of optimizing control commands based on the real-time operating status data to obtain optimized energy-saving adjustment parameters includes: Acquire historical operational data, and perform risk assessment by comparing the real-time operational status data with the historical operational data to obtain the real-time risk status; The state error is obtained by combining the real-time risk status and the historical operation data; Based on the state error, the adjustment sequence is optimized using a gradient descent algorithm to obtain the optimized energy-saving adjustment parameters.

[0010] In one optional implementation, the step of obtaining filter efficiency data based on the energy-saving adjustment parameters and updating the load prediction value based on the filter efficiency data to obtain the water usage behavior pattern and the updated load prediction value includes: The water quality parameters and filter cartridge usage time are obtained based on the energy-saving adjustment parameters. The efficiency is determined by comprehensively judging the efficiency based on the water quality parameters and the filter cartridge usage time. If the filter efficiency is lower than a preset efficiency threshold, the water use behavior is correlated with the filter efficiency to obtain the water use behavior pattern. The load prediction value is obtained by performing load prediction calculations based on the water usage behavior pattern.

[0011] In one optional implementation, the step of performing time scheduling and power optimization based on the water usage behavior pattern and the load forecast to obtain the final control scheme includes: Extract the deviation factor of the water use behavior pattern; Based on the deviation factor and the load forecast value, the time scheduling coefficient is adjusted using a linear regression algorithm to obtain an optimized time scheduling plan; According to the time scheduling plan, power output data is obtained. If the power output data is greater than the preset power output threshold, the power parameters are adjusted to obtain the power output scheme. The final control scheme is obtained by combining the power output scheme and the time scheduling plan.

[0012] In an optional implementation, after performing time scheduling and power optimization based on the water usage behavior pattern and the load forecast to obtain the final control scheme, the method further includes: Based on the final control scheme, obtain the deviation correction factor; The parameter weights of the Long Short-Term Memory network model are updated based on the bias correction factor.

[0013] Secondly, the present invention provides a remote intelligent control system for a water purifier based on the Internet of Things, comprising: The data acquisition module is used to acquire real-time water consumption data and water purifier status data; The load prediction module is used to predict the load sequence by combining the real-time water consumption data and the water purifier status data with a preset long short-term memory network model. The dynamic adjustment module is used to dynamically adjust the system based on the predicted load sequence and the water purifier status data to obtain an adjustment sequence. The parameter optimization module is used to obtain real-time operating status data according to the adjustment sequence, and to optimize the control instructions according to the real-time operating status data to obtain optimized energy-saving adjustment parameters. The update and optimization module is used to obtain filter efficiency data according to the energy-saving adjustment parameters, and update the load prediction value based on the filter efficiency data to obtain the water use behavior pattern and the updated load prediction value. The scheme determination module is used to perform time scheduling and power optimization based on the water usage behavior pattern and the load prediction value to obtain the final control scheme.

[0014] Thirdly, the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the IoT-based remote intelligent control method for water purifiers described in any one of the above.

[0015] Fourthly, the present invention also provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the IoT-based remote intelligent control method for water purifier described in any one of the above-mentioned methods.

[0016] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention updates the load prediction value through filter efficiency data to obtain the water use behavior pattern, performs time scheduling and power optimization, and combines the load prediction value to achieve more accurate time scheduling and power allocation, which can realize energy-saving and efficient adaptive intelligent control.

[0017] (2) This invention uses a long short-term memory network (LSTM) model to predict load, effectively extract the dynamic change characteristics of user water use behavior, accurately perceive the dynamic water use behavior of users, and predict water load in advance.

[0018] (3) This invention obtains energy-saving adjustment parameters by combining real-time operating status data with optimized control instructions, predicts user water demand in advance, avoids energy waste caused by fixed power operation, provides feedback on current operating efficiency and corrects adjustment strategies, and realizes dynamic optimization of operating status and energy-saving operation according to water demand. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of a remote intelligent control method for a water purifier based on the Internet of Things, provided in the first embodiment of the present invention. Figure 2 This is a schematic diagram of a remote intelligent control system for a water purifier based on the Internet of Things, provided in the second embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Reference Figure 1 The first embodiment of the present invention provides a remote intelligent control method for a water purifier based on the Internet of Things, comprising the following steps: S11, obtain real-time water consumption data and water purifier status data; S12, Based on the real-time water consumption data and the water purifier status data, load prediction is performed using a preset long short-term memory network model to obtain a predicted load sequence. S13, dynamically adjust according to the predicted load sequence and the water purifier status data to obtain the adjustment sequence; S14, obtain real-time operating status data according to the adjustment sequence, and optimize the control command according to the real-time operating status data to obtain the optimized energy-saving adjustment parameters; S15, obtain filter efficiency data according to the energy-saving adjustment parameters, and update the load prediction value based on the filter efficiency data to obtain the water use behavior pattern and the updated load prediction value. S16, Time scheduling and power optimization are performed based on the water usage behavior pattern and the load prediction value to obtain the final control scheme.

[0022] In step S11, real-time water consumption data and water purifier status data are acquired.

[0023] It should be noted that this step serves to acquire and process data. Real-time water consumption data is collected through IoT embedded flow sensors, such as Hall effect sensors, measuring the amount of water flowing through the water purifier per unit time to quantify the dynamic characteristics of user water consumption behavior. Water purifier status data includes real-time parameters such as filter pressure status, motor operating power, and water temperature, used to comprehensively assess the health status of the equipment. This data can be obtained through pressure sensors, current sensors, voltage sensors, and thermocouple sensors. For example, the smart water purifier collects water consumption and filter pressure data once per second through built-in flow and pressure sensors, generating time-series data containing timestamps, flow values, and pressure values, such as [2025-09-09-12:30:00, 2L / min, 0.3MPa].

[0024] In step S12, based on the real-time water consumption data and the water purifier status data, and combined with a preset long short-term memory network model, load prediction is performed to obtain a predicted load sequence, including: The water consumption data and the water purifier status data are smoothed to obtain a smoothed data sequence. Anomaly detection is performed on the smoothed data sequence and the water purifier status data to obtain the water usage behavior status; The water load is predicted by combining the long short-term memory network model with the water use behavior state to obtain the predicted load sequence.

[0025] It should be noted that this step uses a Long Short-Term Memory (LSTM) network model to transform real-time water consumption data into a quantitative prediction of future water demand, providing a basis for dynamic energy-saving control and solving the energy waste problem caused by static control in traditional water purifiers. The smoothing process involves using a sliding window averaging method to process the real-time water consumption data, eliminating instantaneous fluctuation noise. For example, the sliding window averaging method is used to smooth both water consumption data and water purifier status data, with a time window size of 5 minutes and a sliding step size of 1. The arithmetic mean is calculated for each data point slid. If missing values ​​exist, linear interpolation is used to fill them in. In boundary processing, for the beginning of the sequence, such as the first 5 minutes after system startup, forward padding is used, filling with the initial stable value; for the end of the sequence, such as less than 5 minutes before data acquisition stops, backward padding is used, filling with the last valid value, ensuring the temporal continuity of the smoothed data sequence. After completing the sliding window averaging calculation, the average values ​​obtained from each calculation are combined to form a new data sequence, which is the smoothed data sequence after smoothing.

[0026] In this embodiment, anomaly detection for the water consumption data sequence and the water purifier status data is performed using a standard deviation threshold. The mean and standard deviation of the set of historical data with the same time attributes as the current data point, such as day of the week or time interval, are calculated to obtain the standard deviation threshold. If the corresponding data in the smoothed data sequence exceeds the standard deviation threshold, it is determined to be an anomaly. For example, if the current data point is Wednesday 14:00-14:30, then the same time period includes all Wednesdays from 14:00-14:30 in the past 30 days. For instance, if the mean of all data from Monday mornings 7:00-7:30 in the past month is 4 MPa and the standard deviation is 0.8 MPa, then the standard deviation threshold should be twice the standard deviation, i.e., 4 + 2 × 0.8 = 5.6. In the smoothed data sequence, the average filter pressure from Monday mornings 7:00-7:05 is 4.5 MPa, which is less than the standard deviation threshold, so the status is normal; if it exceeds the standard deviation threshold, it needs to be marked as an anomaly. In addition, if a status is marked as abnormal, the system will determine the cause of the abnormality based on the abnormal data. For example, if the filter pressure is abnormal, the system will determine that the filter is potentially clogged; if the water flow rate suddenly increases abnormally, the system will determine that the water pipe is potentially ruptured. These abnormal statuses will be visualized for users via mobile devices.

[0027] Furthermore, water usage behavior states are different status labels or codes that divide the data, such as normal water usage, abnormal water usage, and standby. For example, water usage > 2.5 L / min needs to be labeled as high-frequency water usage; water usage < 0.5 L / min needs to be labeled as low-flow standby. Transforming the raw data into semantic water usage behavior states provides more easily learnable features for the LSTM model.

[0028] In this embodiment, the predicted load sequence is a quantitative prediction of the filter cartridge load (such as pressure difference and processing capacity occupancy) over a future period by a Long Short-Term Memory (LSTM) network model based on historical and real-time data. When the preset LSTM network model performs load prediction, it takes the water usage behavior status of the past 24 hours as input, solves the sequence dependency problem through a gating mechanism (input gate, forget gate, output gate), and outputs the predicted load sequence for the next 24 hours.

[0029] For example, during the training of this Long Short-Term Memory (LSTM) network model, the input layer receives water purifier operation data from the past 24 hours, including water usage behavior. The hidden layer captures long-term dependencies in the time series through gating mechanisms (input gate, forget gate, output gate), and the output layer is configured with 64 neurons, corresponding to load predictions for the next 24 hours. Since the load prediction values ​​are continuous, a linear activation function is used.

[0030] In addition, the historical operating data of the water purifier was divided into training and validation sets in an 8:2 ratio according to time sequence. Mean squared error (RMSE) was used to measure the deviation between the predicted load and the actual load. The Adam optimizer was selected, with a batch size of 32 and a maximum training epoch of 500 epochs. Performance was evaluated on the validation set every 10 epochs. Model convergence was determined when the root mean squared error (RMSE) of the validation set showed no significant fluctuation for 10 consecutive epochs, with a fluctuation range <0.01% and an RMSE <2.0%. In this embodiment, the validation set RMSE converged to 1.85% at epoch 350, and the RMSE stabilized in the 1.82%-1.87% range in subsequent epochs.

[0031] In step S13, dynamic adjustments are made based on the predicted load sequence and the water purifier status data to obtain an adjustment sequence, including: The predicted load sequence is subjected to a threshold judgment. If the load level in the predicted load sequence is less than a preset load threshold, it is marked as a low-peak state, and a low-peak state sequence is obtained. The reduction in equipment operating frequency is calculated based on the off-peak state sequence to obtain the initial adjustment sequence; Extract the filter cartridge pressure status data from the water purifier status data, classify the risk levels, and determine the filter cartridge clogging risk level; The initial adjustment sequence is adapted to the filter element clogging risk level to obtain the adjustment sequence.

[0032] It should be noted that this step generates operating frequency adjustment commands through real-time collaborative analysis of load forecasting and equipment status, solving the problem of high energy consumption caused by static strategies in traditional control methods. The preset load threshold is a critical value set based on historical water usage data. A value below this threshold indicates low water demand, determined by the mean and standard deviation of historical water usage data. For example, in the past 60 days of historical water usage data for a water purifier, the load mean for all Mondays from 7:00 to 9:00 was 3, and the standard deviation was 0.8. Therefore, the off-peak threshold is 3 - 0.8 = 2.2. The preset load threshold is not fixed but dynamically adjusted through online learning; that is, every seven days, the new mean and standard deviation for each time period are recalculated and updated. If the water consumption level in the predicted load sequence from 7:00 to 7:30 is lower than the preset load threshold, it is determined to be an off-peak period. The off-peak period status sequence can be a Boolean sequence marking whether future time periods belong to off-peak periods, where 1 = off-peak period, 0 = non-off-peak period.

[0033] In this embodiment, during off-peak periods, the operating frequency and power of the water purifier need to be reduced to achieve energy savings. The reduction percentage is the proportion of the equipment's operating frequency and power being lowered (e.g., from 100% to 70%), which can be calculated using linear interpolation. Taking frequency adjustment as an example, the initial adjustment sequence is a set of instructions generated to reduce the equipment's operating frequency when the predicted load sequence is below a preset threshold, recording the time-series data of frequency adjustment values ​​for each period. Essentially, energy savings are achieved by reducing unnecessary operations (such as reducing backwashing frequency and shortening pump running time) while ensuring that the basic functions of the equipment are not affected. The calculation can be performed using the following method: Decrease magnitude = (1− )×Maximum decrease The maximum reduction is set at 40% to adapt to the system's operating status and avoid equipment downtime; the predicted load is the water consumption recorded in the predicted load sequence.

[0034] For example, if the predicted load for a certain period is 4L / min, the reduction amplitude is calculated to be (1−4 / 5)×40%=8%. If the maximum filter flushing frequency is set to 20 times / hour, then the flushing frequency should be reduced by 20×8%=2 times / hour.

[0035] Furthermore, the filter element clogging risk level reflects the probability of filter element clogging; the higher the risk level, the greater the probability of clogging. Filter element pressure status data is the pressure difference between the filter element's inlet and outlet monitored by a pressure sensor, i.e., the filter element pressure value. Risk level classification is performed through risk threshold comparison to determine the filter element clogging risk level, including low risk, medium risk, and high risk. The risk threshold is experimentally determined, and the relationship between filter element clogging status and filter element pressure value is verified through multiple experiments. For example, if the filter element pressure value is less than 5 MPa, it is classified as low risk; between 5 and 8 MPa, it is classified as medium risk; and greater than 8 MPa, it is classified as high risk.

[0036] Specifically, the adjustment sequence is an optimization strategy that combines filter element risk assessment. It dynamically adjusts the reduction rate to avoid excessive energy saving that could damage the equipment. For example, different adjustment rates are set for different risk levels in the initial adjustment sequence: 0% for low risk, 20% for medium risk, and 40% for high risk. A potential risk warning is also issued for high risk, and this warning is visualized on the user's mobile device. For example, if the initial adjustment sequence requires reducing the flushing frequency by 2 times per hour, and the filter element clogging risk level is medium, then the adjustment rate is increased by 20%, resulting in a flushing frequency of 20 × 20% = 4 times per hour. Therefore, the final output adjustment sequence requires an increased flushing frequency of 2 times per hour.

[0037] In step S14, real-time operating status data is obtained according to the adjustment sequence, and control commands are optimized based on the real-time operating status data to obtain optimized energy-saving adjustment parameters.

[0038] It should be noted that this step, through real-time risk perception and parameter optimization, re-verifies the adjustment sequence of S13 and transforms it into refined energy-saving parameters. The real-time operating status data refers to the current operating parameters of the water purifier (water flow rate, filter pressure, water temperature), sourced from the execution result of the adjustment sequence output in S13. This means the water purifier parameters are re-collected after the control parameters indicated by the adjustment sequence have been executed. During the data collection process, smoothing is performed as in step S12, using a sliding window averaging method to process the real-time water consumption data and eliminate instantaneous fluctuation noise. For example, the sliding window averaging method is used to smooth the water consumption data and the water purifier status data, with a time window size of 5 minutes and a sliding step size of 1. The arithmetic mean is calculated once for each data point slid. If missing values ​​exist, linear interpolation is used to fill in the missing values. In boundary processing, for the start of the sequence, such as the first 5 minutes after system startup, forward padding is used, filling with the initial stable value; for the end of the sequence, such as less than 5 minutes before data acquisition stops, backward padding is used, filling with the last valid value, ensuring the temporal continuity of the smoothed data sequence. After calculating the sliding window averaging method, the average values ​​obtained from each calculation are combined to form a new data sequence, which is the smoothed data sequence after smoothing. Control command optimization involves a three-level architecture of risk perception, error quantification, and dynamic optimization to re-optimize energy saving and obtain optimized energy-saving adjustment parameters.

[0039] In one implementation, the step of optimizing control commands based on the real-time operating status data to obtain optimized energy-saving adjustment parameters includes: Acquire historical operational data, and perform risk assessment by comparing the real-time operational status data with the historical operational data to obtain the real-time risk status; The state error is obtained by combining the real-time risk status and the historical operation data; Based on the state error, the adjustment sequence is optimized using a gradient descent algorithm to obtain the optimized energy-saving adjustment parameters.

[0040] It should be noted that historical operating data refers to the statistical values ​​of equipment operation over the past 30-60 days, used to establish a baseline risk model. Real-time risk status is an identifier of potential anomalies or danger levels identified by comparing the current operating status of the water purifier (such as differential pressure, flow rate, and load rate) with historical normal status. For example, the historical data threshold is calculated by extracting the mean and standard deviation of historical data for the same period; the mean plus the standard deviation equals the historical threshold. The judgment method in this step is the same as in step S12. For instance, if the average water consumption for the four Monday mornings from 7:00 to 7:30 over the past 30 days is 4 L / min with a standard deviation of 0.8 L / min, then the historical data threshold is 1 standard deviation, i.e., 4 ± 0.8 L / min. In the real-time operating status data obtained from 7:00 to 7:05 on Monday mornings, the water consumption is 2 L / min, less than 3.2 L / min, exceeding the historical data threshold. Therefore, it is determined to be a risky state, requiring more precise state error adjustments for corresponding energy-saving adjustments.

[0041] Furthermore, if a risky state exists, a state error value is calculated. When calculating the deviation between real-time operating status data and historical operating data, a relative deviation quantification model is used to accurately assess the degree of risk. State error is the quantified deviation between the equipment's operating state and its actual real-time state, providing optimization weights for the gradient descent algorithm; the larger the error value, the higher the risk weight. The formula for calculating state error is as follows: 𝛥=(|∞−𝜇| / 𝜎)×100% Where, 𝜇 is the mean of the historical data threshold; 𝜎 is the standard deviation of the historical data threshold; 𝛥𝑥 is the state error; and 𝑥 is the real-time data value.

[0042] In this embodiment, the energy-saving adjustment parameters are the result of optimizing the initial adjustment sequence using a gradient descent algorithm. For example, the weight coefficients of the risk term in the objective function are set based on the state error value. β_dynamic = min(0.6, 0.3 × β)

[0043] Where, β 动态 It is the risk value weight; 𝛥𝑥 is the state error.

[0044] For example, when the state error is 100%, the risk item weight should be 0.3; when the state error is 200%, the risk item weight should be 0.6. This breaks through the limitations of traditional fixed weights and enables priority handling of high-risk events.

[0045] For example, the adaptive controller learning rate φ is set based on the state error value.

[0047] Furthermore, the objective function model is constructed as follows: 𝑓(𝑥)=𝛼⋅𝑃(𝑥)+β 动态 ⋅𝑅(𝑥)

[0048] 𝛼=1−𝛽 动态

[0049] Where 𝛼 represents the weight of the adjustment parameter. 𝑃(𝑥) is the energy consumption item; 𝑥 is the adjustment parameter, such as motor power; 𝑅(𝑥) is the state error; 𝛽 动态 It is the risk value weight.

[0050] Specifically, the partial derivative calculation model and iterative formula are as follows:

[0051]

[0052] The function converges when the rate of change of the function value is less than the learning rate φ. The resulting adjustment parameter value is the optimized energy-saving adjustment parameter.

[0053] In step S15, filter efficiency data is obtained according to the energy-saving adjustment parameters, and the load prediction value is updated based on the filter efficiency data to obtain the water usage behavior pattern and the updated load prediction value, including: The water quality parameters and filter cartridge usage time are obtained based on the energy-saving adjustment parameters. The efficiency is determined by comprehensively judging the efficiency based on the water quality parameters and the filter cartridge usage time. If the filter efficiency is lower than a preset efficiency threshold, the water use behavior is correlated with the filter efficiency to obtain the water use behavior pattern. The load prediction value is obtained by performing load prediction calculations based on the water usage behavior pattern.

[0054] It should be noted that this step addresses the technical blind spot in traditional water purifier control—where filter performance degradation is not quantified and fed back into load prediction—by establishing a dynamic coupling mechanism between filter health status and water usage behavior. This is achieved through real-time iteration of multi-dimensional efficiency assessment and prediction models. Specifically, water quality parameters are a set of physicochemical indicators of the water body monitored in real time after the implementation of energy-saving adjustments, including turbidity (NTU), total dissolved solids (TDS), and conductivity (μS / cm), used to quantify filter efficiency. Filter usage time is the total working time of the filter since its activation, automatically counted by the device's timer and used in conjunction with water quality parameters to determine the filter's lifespan.

[0055] In this embodiment, a multi-parameter weighted method is used to transform discrete parameters into a unified index, namely filter efficiency, when making a comprehensive efficiency judgment. For example, the calculation formula is: 𝜂=𝛼⋅ +𝛽⋅ +𝜔⋅

[0056] Among them, 𝜂 is the filter efficiency, which quantifies the degree of filter performance degradation; exceeding 100% indicates performance deterioration; 𝐴 is the real-time turbidity value, collected by an optical sensor, reflecting the concentration of suspended solids in the water; T A This is the dynamic threshold for turbidity, calculated from the historical mean and standard deviation. Here, we take the historical mean of turbidity plus twice the historical standard deviation of turbidity, T. A =μ A +2σ A B represents the total dissolved solids in real time; the electrochemical sensor measures the total amount of dissolved inorganic matter. T B It is the dynamic threshold for total dissolved solids, derived from the historical average μ. B With 2 times the standard deviation σ B Calculations show that T B =μ B +2σ B C represents the cumulative usage time of the filter element; T C This refers to the filter element's design lifespan (in hours), which is preset to 500 hours. α, β, and ω are weighting coefficients, calibrated through orthogonal experiments and analysis of variance; here, α = 0.4, β = 0.4, and ω = 0.2. The experimental scheme can be set up as an example as follows.

[0057] The purpose of this experiment was to determine the significance of the effects of turbidity factor, total dissolved solids factor, and usage time factor on filter cartridge efficiency decay, and to quantify their weighting ratios. A three-factor, three-level orthogonal array was used, with the following factor and level settings: turbidity ratios of 0.8, 1.2, and 1.6, simulating water quality ranging from clean to heavily polluted; total dissolved solids ratios of 0.8, 1.2, and 1.6, simulating dissolved solids content ranging from low to high; and usage time ratios of 0.3, 0.6, and 0.9, simulating the state from new filter cartridge to the end of its lifespan. A water purifier test platform was built in a controlled laboratory environment, equipped with standard sensors. Nine experimental combinations were arranged according to the selected orthogonal array (and the orthogonal array values ​​mentioned above). For each combination, the water purifier system was run until stable, and stable sensor readings and known filter cartridge usage time were recorded. In this experiment, the process of running the water purifier system until stable can be exemplarily set as a flow rate fluctuation rate of less than ±2% for 5 consecutive minutes, considered stable.

[0058] After each experiment, the filter cartridge was disassembled, and its actual purified water flow rate decay rate (or pressure difference growth rate) was measured on a standard testing platform. This rate was used as the gold standard for the true performance decay of the filter cartridge, denoted as Y_true. The data from each experiment were substituted into the efficiency model to calculate α. Y_true was used as the response variable, and variance analysis was performed on each factor. The purpose was to determine whether the three factors significantly affected the filter cartridge performance decay and to calculate the contribution rate ρ of each factor, i.e., α:β:ω = ρα:ρβ:ρω. After initially calculating the weights based on the contribution rates obtained from the variance analysis, fine-tuning was performed in conjunction with physical mechanisms (such as the physical mechanism of filter cartridge clogging) and model validation results to finally determine a stable set of weight coefficients. If the variance analysis results showed that the contribution rate of a certain factor was extremely low (e.g., ρ < 5%), its weight could be further reduced during fine-tuning to simplify the model. For example, experimental results showed that turbidity and TDS significantly affected short-term efficiency fluctuations, while usage time was the main cause of long-term decay, thus yielding α = 0.4, β = 0.4, w = 0.2. The calibrated model is applied to a set of validation data that were not used in the experiment, and the correlation coefficient (R²) between the predictive efficiency and the true Y is calculated. An R² > 0.85 is required to confirm the model's effectiveness. This validation process ensures that the example weights given in the documentation (α=0.4, β=0.4, ω=0.2) are scientifically validated and reliable values, not arbitrary assumptions.

[0059] For example, the current turbidity, total dissolved solids (TDS), and filter cartridge usage time are compiled into a real-time data stream. A water purifier records a current turbidity of 2.5 NTU, a TDS of 150 ppm, and a filter cartridge usage time of 400 hours. By combining historical turbidity and historical TDS mean and standard deviation data with the calculation formula for the overall efficiency assessment mentioned above, the final result is a filter cartridge efficiency of 25%.

[0060] Specifically, the preset efficiency threshold can be determined through orthogonal experiments. It represents the critical failure point where the filter cartridge performance degrades to an abnormal state; here, it is set to 70%. For example, if the filter cartridge efficiency falls below the preset efficiency threshold, a water usage behavior correlation operation is triggered to classify different water usage scenarios. Water usage behavior refers to the frequency and volume of water used by the user within a certain time period, obtained by capturing pulse waveforms through a flow sensor and recording steady-state high pressure through a pressure sensor. For example, it can be classified as high-frequency short-duration type and continuous high-flow type. [High-frequency, short-duration type, characteristic range, single water usage duration < 30 seconds, daily frequency > 100 times] [Sustained high flow rate, characteristic range, single flow rate > 5L / min, duration > 1 minute, daily average > 30 times] In addition, a balanced type of water use behavior is introduced, whose characteristic range is not in the high-frequency short-time type or the continuous high-flow type.

[0061] Furthermore, the efficiency and behavior coupling model,

[0063] This model dynamically correlates filter efficiency η with behavioral patterns, quantifying the impact of water usage behavior on the filter. For example, the behavior correction factor η is the acceleration coefficient of water usage behavior on filter wear. Through cluster analysis, water usage behavior is divided into three patterns, each corresponding to a specific wear coefficient. User habits are quantified as a correction factor for the filter efficiency decline rate, used to correct the filter efficiency η. 行为 =η×Y. For example, if the filter efficiency η = 50% and Y = 1.5, then the filter efficiency η is corrected. 行为 =75%.

[0064] In this embodiment, the load prediction value is a quantitative prediction of the filter element load over a future period. Load prediction calculations are performed specifically for the filter element load, requiring the classification of high-load and low-load states. A high-load state is defined as a predicted water consumption > 3 L / min or a turbidity deviation ΔA ≥ 0.4; a low-load state is defined as a predicted water consumption deviation < 1 L / min or a turbidity deviation ΔA ≤ 0.1; and all other states are classified as balanced states. Correspondingly, the base load for the high-load state is set to 80%, the base load for the low-load state is set to 30%, and the base load for the balanced state is set to 50%. Combined with user behavior compensation,

[0066] Among them, P 决策树 This refers to the base load corresponding to high load, low load, and balanced states; P f𝑖𝑛𝑎l This is the final predicted load.

[0067] In step S16, time scheduling and power optimization are performed based on the water usage behavior pattern and the load forecast value to obtain the final control scheme, including: Extract the deviation factor of the water use behavior pattern; Based on the deviation factor and the load forecast value, the time scheduling coefficient is adjusted using a linear regression algorithm to obtain an optimized time scheduling plan; According to the time scheduling plan, power output data is obtained. If the power output data is greater than the preset power output threshold, the power parameters are adjusted to obtain the power output scheme. The final control scheme is obtained by combining the power output scheme and the time scheduling plan.

[0068] It should be noted that this step integrates water usage behavior characteristics with load prediction results to dynamically optimize equipment operating time and power allocation, minimizing energy consumption while ensuring water supply demand. The deviation factor (δ) is a statistical measure quantifying the degree of deviation between water usage behavior patterns and historical benchmarks, reflecting the dynamic impact of user water usage habits on filter cartridge load. The calculation formula is as follows:

[0070] Among them, A i B represents the real-time turbidity value for the i-th time period; i μ represents the real-time TDS value for the i-th time period. A The average turbidity for the same period in history; μ B The mean TDS for the same period in history; σ A σ represents the historical standard deviation of turbidity. B is the historical standard deviation of TDS; n is the number of data points within the statistical period.

[0071] By calculating the deviation factor, the impact of current water usage behavior on the filter cartridge is identified. For example, the real-time data for a user during the afternoon peak period (14:00-16:00) is as follows: [Turbidity data, 2.5 NTU, 2.6 NTU, ..., 2.8 NTU]; [TDS data, 160ppm, 162ppm, ..., 165ppm]; The historical average turbidity for the same period last week (14:00-16:00) was 2.0 NTU, with a standard deviation of 0.3 NTU; the historical average TDS for the same period was 150 ppm, with a standard deviation of 10 ppm; therefore, the deviation factor δ could be δ= [(1.67+1.0)+(1.73+1.2)+...+(2.00+1.5)]= ×180=3.0.

[0072] Furthermore, the optimized time scheduling sequence is a dynamic operation plan generated by adjusting the time scheduling coefficients based on water usage behavior patterns (quantified by the deviation factor δ) and load forecasts using linear regression. Its core is to quantify the impact of user water usage habits on filter cartridge load and, combined with future load demands, adjust parameters such as backwashing frequency and operating periods. When using a linear regression algorithm to adjust the time scheduling coefficients, linear regression is a statistical method that fits the linear relationship between the input variables (deviation factor, load forecasts, and output variables) and the time scheduling coefficients. Its regression model is as follows: K t =ζ⋅δ+φ⋅P f𝑖𝑛𝑎l +ϵ

[0073] Among them, K tζ is the time scheduling coefficient, the adjustment factor of the original time scheduling plan, used to measure the combined impact of current water usage behavior and load demand on the scheduling strategy. Its value directly determines the direction of the scheduling plan adjustment; ζ is the weighting coefficient of the deviation factor; φ is the weighting coefficient of the load forecast value; δ is the deviation factor; P f𝑖𝑛𝑎l ϵ represents the final load prediction value; ϵ is the random error term, which is set to 0 for a simple example. In particular, the weighting coefficients ζ and φ need to be obtained through model training to minimize the error between the model's predicted time scheduling coefficients and actual needs (such as filter cartridge wear control and water supply stability).

[0074] During training, the input features are the deviation factor calculation results for each time period of each day over the past 3 months, the corresponding load prediction values, and the manually labeled historical best scheduling coefficients.

[0075] Here, the mean squared error (MSE) between the predicted and true values ​​is minimized for model training validation.

[0076] Where m is the number of historical data samples; δ is the actual time scheduling coefficient required on day j; j P is the deviation factor for day j; j This represents the load forecast for day j.

[0077] The training method uses the least squares approach to find the optimal weights, with the goal of minimizing the prediction error.

[0078] Cross-validation was used to evaluate the model's generalization ability. Historical data was divided into a training set (70%) and a test set (30%), and the prediction error (MSE) on the test set was calculated. If the error on the test set was too large, the model parameters were adjusted, and the model was retrained until the error met the requirement of MSE ≤ 0.1. The final results showed that ζ = 0.6, φ = 0.4, which means that for every 1 unit increase in the bias factor, the time scheduling coefficient increases by 0.6 units, indicating that water quality fluctuations have a greater impact on scheduling; and for every 1 unit increase in the load prediction value, the time scheduling coefficient increases by 0.4 units.

[0079] For example, the time scheduling coefficient can be obtained from the regression model described above, K t =2.0 indicates that the backwashing frequency needs to be significantly increased. Adjust the backwashing frequency from once every 2 hours to once every 1 hour (extend the operating time to 3 hours) to cover the water demand during periods of high deviation and high load.

[0080] Furthermore, power output data refers to the electrical power that the equipment is expected to consume in each time period according to the time scheduling plan. Essentially, it is a quantitative mapping between the time scheduling plan and the equipment's operating characteristics. The preset power output threshold is the maximum safe power set during equipment design. Power parameter adjustment is used when the power output data exceeds the preset power output threshold, by reducing operating power, adjusting backwashing frequency, etc., to limit the power within a safe range. The power output scheme generates a power allocation strategy for each time period based on the time scheduling plan and equipment operating characteristics, ensuring that the equipment meets water demand while operating within a safe power range and achieving energy consumption optimization. For example, assuming the power output is 120W and the preset power output threshold is 100W, if the power output exceeds the preset power output threshold, the operating frequency will be gradually reduced to decrease the power output, such as reducing the backwashing frequency from 14 times / hour to 12 times / hour, thereby reducing the power output to 90W, within the preset power output threshold. Finally, combining the time scheduling plan and the power output scheme forms a synergistically optimized final control scheme.

[0081] In one implementation, after performing time scheduling and power optimization based on the water usage behavior pattern and the load forecast to obtain the final control scheme, the method further includes: Based on the final control scheme, obtain the deviation correction factor; The parameter weights of the Long Short-Term Memory network model are updated based on the bias correction factor.

[0082] It should be noted that this additional step is used to achieve closed-loop optimization. By obtaining the deviation correction factor after the final control scheme is executed, the parameter weights of the Long Short-Term Memory (LSTM) network model are updated in reverse, enabling the model to continuously adapt to water use behavior patterns and load changes. The deviation correction factor is a statistical measure that quantifies the deviation between the final control scheme's execution effect and the expected target, reflecting the dynamic deviation between water use behavior patterns, load predictions, and actual operation. Essentially, it identifies the errors in load prediction or behavior pattern recognition of the LSTM model by comparing the model's predicted values ​​with actual operating data, providing a supervisory signal for model parameter updates. The deviation correction factor is obtained by calculating the load prediction error and the filter efficiency error using the actual data after the final control scheme is executed and then fusing them.

[0083] Furthermore, when updating the parameter weights of the Long Short-Term Memory (LSTM) network model, the model inputs are historical water consumption data (flow rate, water usage duration), water purifier status data (pressure difference, temperature), and historical load sequences, and the output is a load prediction sequence for the next 24 hours. The core parameters include input gate weights, forget gate weights, output gate weights, and candidate cell state weights.

[0084] When updating parameters, the error between the LSTM model's predicted load and the actual load is minimized by adjusting the aforementioned weights, thereby improving the accuracy of filter efficiency prediction. A supervisory signal is constructed based on the bias correction factor. The gradient of the LSTM model parameters is calculated and the weights are updated using the backpropagation algorithm and gradient descent optimizer, making the model prediction closer to actual operating data. The loss function is in the form of...

[0085] Where Reg(W) is the weight regularization term (L2 norm) to prevent overfitting; δ is the regularization coefficient; and Δθ is the bias correction factor.

[0086] Taking the partial derivatives of the weights of each LSTM gate, the parameters are updated using gradient descent.

[0087] Where Ω is the learning rate, set to 0.001; Let W be the gradient of the loss function with respect to the weights W.

[0088] By continuously optimizing LSTM parameters as water usage behavior patterns change, the static model avoids lagging behind in adapting to dynamic demands.

[0089] In summary, this invention discloses a remote intelligent control method for water purifiers based on the Internet of Things (IoT), comprising: acquiring real-time water consumption data and water purifier status data; performing load prediction based on the real-time water consumption data and the water purifier status data, combined with a preset long short-term memory network model, to obtain a predicted load sequence; dynamically adjusting the load based on the predicted load sequence and the water purifier status data to obtain an adjustment sequence; acquiring real-time operating status data based on the adjustment sequence, and optimizing control commands based on the real-time operating status data to obtain optimized energy-saving adjustment parameters; acquiring filter efficiency data based on the energy-saving adjustment parameters, and updating the load prediction value based on the filter efficiency data to obtain a water consumption behavior pattern and an updated load prediction value; and performing time scheduling and power optimization based on the water consumption behavior pattern and the load prediction value to obtain a final control scheme. This invention uses a Long Short-Term Memory (LSTM) network model for load prediction, effectively extracting dynamic changes in user water usage behavior, accurately sensing dynamic user water usage behavior, and predicting water load in advance. By combining real-time operating status data with optimized control commands, energy-saving adjustment parameters are obtained, allowing for advance prediction of user water demand and avoiding energy waste caused by fixed-power operation. The current operating efficiency is fed back and the adjustment strategy is corrected, achieving dynamic optimization of operating status and energy-saving operation based on water demand. By updating the load prediction value with filter efficiency data, the water usage behavior pattern is obtained, enabling time scheduling and power optimization. Combining the load prediction value with the actual load prediction value achieves more accurate time scheduling and power allocation, realizing long-term adaptive control based on user behavior patterns and maintaining continuous and efficient operation.

[0090] Reference Figure 2 The second embodiment of the present invention provides a remote intelligent control system for a water purifier based on the Internet of Things, comprising: The data acquisition module is used to acquire real-time water consumption data and water purifier status data; The load prediction module is used to predict the load sequence by combining the real-time water consumption data and the water purifier status data with a preset long short-term memory network model. The dynamic adjustment module is used to dynamically adjust the system based on the predicted load sequence and the water purifier status data to obtain an adjustment sequence. The parameter optimization module is used to obtain real-time operating status data according to the adjustment sequence, and to optimize the control instructions according to the real-time operating status data to obtain optimized energy-saving adjustment parameters. The update and optimization module is used to obtain filter efficiency data according to the energy-saving adjustment parameters, and update the load prediction value based on the filter efficiency data to obtain the water use behavior pattern and the updated load prediction value. The scheme determination module is used to perform time scheduling and power optimization based on the water usage behavior pattern and the load prediction value to obtain the final control scheme.

[0091] It should be noted that the IoT-based remote intelligent control system for water purifiers provided in this embodiment of the invention is used to execute all the process steps of the IoT-based remote intelligent control method for water purifiers described in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0092] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as an update and optimization program. When the processor executes the computer program, it implements the steps described in the various embodiments of the IoT-based remote intelligent control method for water purifiers, for example... Figure 1 The step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described device embodiments, such as the update and optimization module.

[0093] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0094] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0095] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0096] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0097] Wherein, if the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0098] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0099] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A remote intelligent control method for a water purifier based on the Internet of Things, characterized in that, include: Obtain real-time water consumption data and water purifier status data; Based on the real-time water consumption data and the water purifier status data, load prediction is performed using a preset long short-term memory network model to obtain a predicted load sequence. The adjustment sequence is obtained by dynamically adjusting the predicted load sequence and the water purifier status data. Real-time operating status data is obtained based on the adjustment sequence, and control commands are optimized based on the real-time operating status data to obtain optimized energy-saving adjustment parameters; The filter efficiency data is obtained based on the energy-saving adjustment parameters, and the load prediction value is updated based on the filter efficiency data to obtain the water use behavior pattern and the updated load prediction value. Based on the water usage behavior pattern and the load forecast, time scheduling and power optimization are performed to obtain the final control scheme.

2. The remote intelligent control method for a water purifier based on the Internet of Things according to claim 1, characterized in that, The step of predicting the load sequence based on the real-time water consumption data and the water purifier status data, combined with a preset long short-term memory network model, includes: The water consumption data and the water purifier status data are smoothed to obtain a smoothed data sequence. Anomaly detection is performed on the smoothed data sequence and the water purifier status data to obtain the water usage behavior status; The water load is predicted by combining the long short-term memory network model with the water use behavior state to obtain the predicted load sequence.

3. The remote intelligent control method for a water purifier based on the Internet of Things according to claim 1, characterized in that, The step of dynamically adjusting based on the predicted load sequence and the water purifier status data to obtain an adjustment sequence includes: The predicted load sequence is subjected to a threshold judgment. If the load level in the predicted load sequence is less than a preset load threshold, it is marked as a low-peak state, and a low-peak state sequence is obtained. The reduction in equipment operating frequency is calculated based on the off-peak state sequence to obtain the initial adjustment sequence; Extract the filter cartridge pressure status data from the water purifier status data, classify the risk levels, and determine the filter cartridge clogging risk level; The initial adjustment sequence is adapted to the filter element clogging risk level to obtain the adjustment sequence.

4. The remote intelligent control method for a water purifier based on the Internet of Things according to claim 1, characterized in that, The process involves optimizing control commands based on the real-time operating status data to obtain optimized energy-saving adjustment parameters, including: Acquire historical operational data, and perform risk assessment by comparing the real-time operational status data with the historical operational data to obtain the real-time risk status; The state error is obtained by combining the real-time risk status and the historical operation data; Based on the state error, the adjustment sequence is optimized using a gradient descent algorithm to obtain the optimized energy-saving adjustment parameters.

5. The remote intelligent control method for a water purifier based on the Internet of Things according to claim 1, characterized in that, The step of obtaining filter efficiency data based on the energy-saving adjustment parameters and updating the load prediction value based on the filter efficiency data to obtain the water usage behavior pattern and the updated load prediction value includes: The water quality parameters and filter cartridge usage time are obtained based on the energy-saving adjustment parameters. The efficiency is determined by comprehensively judging the efficiency based on the water quality parameters and the filter cartridge usage time. If the filter efficiency is lower than a preset efficiency threshold, the water use behavior is correlated with the filter efficiency to obtain the water use behavior pattern. The load prediction value is obtained by performing load prediction calculations based on the water usage behavior pattern.

6. The remote intelligent control method for a water purifier based on the Internet of Things according to claim 1, characterized in that, The step of performing time scheduling and power optimization based on the water usage behavior pattern and the load forecast to obtain the final control scheme includes: Extract the deviation factor of the water use behavior pattern; Based on the deviation factor and the load forecast value, the time scheduling coefficient is adjusted using a linear regression algorithm to obtain an optimized time scheduling plan; According to the time scheduling plan, power output data is obtained. If the power output data is greater than the preset power output threshold, the power parameters are adjusted to obtain the power output scheme. The final control scheme is obtained by combining the power output scheme and the time scheduling plan.

7. The remote intelligent control method for a water purifier based on the Internet of Things according to claim 1, characterized in that, After performing time scheduling and power optimization based on the water usage behavior pattern and the load forecast to obtain the final control scheme, the method further includes: Based on the final control scheme, obtain the deviation correction factor; The parameter weights of the Long Short-Term Memory network model are updated based on the bias correction factor.

8. A remote intelligent control system for a water purifier based on the Internet of Things, characterized in that, include: The data acquisition module is used to acquire real-time water consumption data and water purifier status data; The load prediction module is used to predict the load sequence by combining the real-time water consumption data and the water purifier status data with a preset long short-term memory network model. The dynamic adjustment module is used to dynamically adjust the system based on the predicted load sequence and the water purifier status data to obtain an adjustment sequence. The parameter optimization module is used to obtain real-time operating status data according to the adjustment sequence, and to optimize the control instructions according to the real-time operating status data to obtain optimized energy-saving adjustment parameters. The update and optimization module is used to obtain filter efficiency data according to the energy-saving adjustment parameters, and update the load prediction value based on the filter efficiency data to obtain the water use behavior pattern and the updated load prediction value. The scheme determination module is used to perform time scheduling and power optimization based on the water usage behavior pattern and the load prediction value to obtain the final control scheme.

Images