Campus direct drinking water energy consumption optimization method and system based on energy-saving control

Abstract

The invention provides a campus direct drinking water energy consumption optimization method and system based on energy-saving control. The method comprises the following steps: collecting key data of campus direct drinking water to construct a water demand prediction model based on a multilayer long short-term memory network; according to a prediction result of the water demand prediction model, generating a dynamic equipment scheduling strategy and executing the dynamic equipment scheduling strategy on the campus direct drinking water equipment; monitoring the data after the dynamic equipment scheduling strategy is executed in real time, and calculating the power deviation between the dynamic equipment scheduling strategy and the actual power of the real-time monitoring equipment; adaptively optimizing the corrected dynamic equipment scheduling strategy according to the actual feedback data for executing the corrected dynamic equipment scheduling strategy; and performing dynamic closed-loop feedback adjustment according to an equipment scheduling strategy in combination with power deviation and water consumption deviation in actual operation. The intelligent energy-saving system effectively overcomes the defects that in the prior art, equipment operation is not matched with the water demand, and water leakage is not detected in time, and intelligent energy-saving optimization of the campus direct drinking water system is achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of energy-saving control, and in particular relates to a method and system for optimizing energy consumption of direct drinking water on campus based on energy-saving control. Background Art

[0002] With the acceleration of urbanization, campus infrastructure is becoming more and more modernized, especially the direct drinking water system, which has become an important part of public places such as universities and office buildings. Campus direct drinking water systems usually include multiple water treatment equipment, such as filters, heaters, disinfectors and water pumps, to ensure the provision of safe and clean drinking water. However, the operation of these devices is often continuous, or controlled according to simple time period preset programs, without considering the actual changes in water demand, resulting in a large amount of energy waste. In addition, the water demand for direct drinking water on campus has significant time fluctuations, such as the peak before class in the morning, concentrated water use during breaks and lunch breaks, and increased demand after students return to the dormitory at night. These fluctuation characteristics require water treatment equipment to be able to flexibly respond to actual needs. However, in the prior art, most direct drinking water systems lack the ability to dynamically perceive water demand and usually adopt a fixed operating mode. This method results in insufficient equipment load and poor user experience during peak water use periods, and excessive equipment operation during off-peak periods, resulting in serious energy waste.

[0003] Existing optimization methods mainly rely on schedules or manual adjustments. This static management mode ignores the complexity of direct drinking water systems and the variability of demand. Some more advanced systems may use simple sensors to monitor parameters such as water temperature and water pressure to adjust the operating status of some equipment, but these systems are usually limited to single-point control and cannot globally optimize the energy consumption of the entire system. On the other hand, pipeline leakage is also one of the common problems in direct drinking water systems, especially in complex pipe network systems in large campuses. Leakage not only wastes a lot of water resources, but also increases the energy consumption of the system, because water pumps and other equipment require higher power to compensate for the loss of water pressure. However, the existing technology for monitoring pipeline leakage usually relies on manual inspections or regular water meter inspections, which cannot achieve real-time and efficient leakage detection, resulting in leakage problems often being delayed, resulting in long-term energy waste and increased maintenance costs. Based on this, the existing technology has obvious deficiencies in energy-saving optimization and pipeline monitoring: on the one hand, there is a lack of dynamic scheduling mechanism for changes in user demand, and on the other hand, there is a lack of efficient pipeline monitoring means to avoid energy waste caused by leakage.

[0004] In view of these problems, it is particularly important to propose an intelligent and energy-saving campus drinking water energy consumption optimization system. In order to effectively reduce system energy consumption and improve user experience, there is an urgent need for a comprehensive energy-saving solution that can dynamically regulate and monitor in real time based on the actual water demand and pipeline status on campus. Summary of the invention

[0005] The purpose of the present invention is to design a campus direct drinking water energy consumption optimization method and system based on energy-saving control, which solves the problems of energy waste, untimely response and inefficient pipeline leakage monitoring in the prior art. By introducing AI technology to accurately predict campus water demand and combining it with real-time monitoring of pipeline status, the present invention realizes a more efficient and intelligent energy-saving optimization solution through various innovations.

[0006] In order to achieve the above object, the present invention provides a first aspect of a method for optimizing energy consumption of direct drinking water in campus based on energy-saving control, the method comprising the following steps: S1. Collect key data of campus drinking water to build a water demand prediction model based on a multi-layer long short-term memory network. The key data includes historical water consumption, weather data and student activity data, including: Decompose the historical water consumption in time series, extract the trend, seasonality and residual features, combine the trend, seasonality, residual features, weather data and student activity data into a multi-feature input matrix, build a water demand forecasting model based on a multi-layer long short-term memory network based on the multi-feature input matrix as input, and use the water demand forecasting model to predict future periods to water demand; S2. Based on the prediction results of the water demand prediction model and the energy consumption and power of the current equipment, a dynamic equipment scheduling strategy is generated and executed on the campus direct drinking water equipment; the dynamic equipment scheduling strategy includes a peak period preheating strategy and a valley period energy saving strategy; S3, real-time monitoring of the data after the execution of the dynamic equipment scheduling strategy, calculating the power deviation between the dynamic equipment scheduling strategy and the actual power of the real-time monitoring equipment, and the water consumption deviation between the real-time actual water consumption and the predicted water consumption. If the power deviation or water consumption deviation exceeds the preset threshold, the adaptive scheduling mechanism is triggered to correct the dynamic equipment scheduling strategy; S4, adaptively optimizing the revised dynamic device scheduling strategy according to the actual feedback data of executing the revised dynamic device scheduling strategy, obtaining the device scheduling strategy, and ensuring the optimization of energy saving effect; S5. Perform dynamic closed-loop feedback adjustment based on the equipment scheduling strategy and the power deviation and water consumption deviation in actual operation. At the same time, design an adaptive dynamic learning mechanism in the dynamic closed-loop feedback adjustment. By combining the power deviation and demand deviation with the adaptive adjustment of the model, the long-term prediction capability of the water demand prediction model can be improved.

[0007] Furthermore, the calculation process of the multi-layer long short-term memory network is expressed as follows: ; in, For the moment The LSTM hidden state represents the memory information of the network; is the weight matrix of the LSTM network, is bias; is the activation function, which is used to output the hidden state at the next moment; The water demand forecasting model is used to predict the future period to Water demand , which is expressed as follows: ; in, are the network parameters obtained through historical data training, It is a multi-layer long short-term memory network; The loss function of the water demand prediction model introduces an offset regularization term to limit the deviation of the multi-layer long short-term memory network when the demand changes sharply. It is expressed as follows: ; in, For the predicted water consumption, is the actual water consumption, is the offset coefficient, i.e., the offset regularization term, which indicates the rate of change of water consumption from the previous moment to the current moment. is the regularization coefficient, which is used to control the weight of the offset term. n represents the total duration, and i represents the current moment.

[0008] Furthermore, the dynamic device scheduling strategy execution rule is expressed as: If an upcoming peak water usage is predicted, the device will enter the preheating state in advance and execute the peak preheating strategy, as shown below: ; in, is the threshold coefficient for peak period judgment; It is the historical average water consumption, a benchmark used to identify fluctuations in demand; Indicates the power level of the equipment preheating in advance to ensure that the equipment is in good condition when the peak water consumption period arrives; Indicates that the device is in normal operation; If future water demand is forecast to be low, When the power level drops below a certain threshold, the device automatically enters energy-saving mode and executes the low-peak energy-saving strategy, as shown below: ; in, is the threshold coefficient of the trough period, Indicates that the device is in standby or off state. Indicates that the equipment is running under low load; Output the dynamic device scheduling strategy that is finally executed ; Said S2 further comprises: Design energy efficiency control formula, combined with equipment efficiency and energy cap , optimize the device power without exceeding the energy consumption limit, as shown below: ; in, For the current time period The power at which the equipment actually operates; is the equipment efficiency coefficient, which fluctuates with load changes and indicates the energy efficiency of the equipment; To limit the system's energy consumption, ensure that the device does not exceed the maximum power at any time.

[0009] Furthermore, the adaptive scheduling mechanism is to modify the scheduling strategy, which is expressed as follows: ; in, is the revised scheduling strategy; and are adjustment coefficients, which are used to control the impact of power deviation and water consumption deviation on the scheduling strategy; is the device power deviation, The water consumption deviation.

[0010] Furthermore, the S3 also includes performing a pipeline leakage detection on the campus direct drinking water equipment, specifically: Water leakage detection identifies pipeline anomalies by calculating the deviation of water pressure or water flow, as shown below: ; in, is the normal water pressure or water flow of the pipeline in historical data, It is the difference between the current pipeline water pressure / flow and the historical normal state; like Exceeding the threshold , it is judged that there is a water leakage problem, an alarm signal is triggered, and the abnormal situation is recorded for maintenance; At the same time, the power smoothing regularization term is designed , the power variation range of the control equipment is expressed as follows: ; in, is the power smoothing regularization term, which is used to reduce the drastic changes in power; is the actual power of the equipment in the previous period; It is the adjustment coefficient, which is used to control the penalty intensity of power variation.

[0011] Furthermore, the adaptively optimizing the modified dynamic device scheduling strategy according to actual feedback data of executing the modified dynamic device scheduling strategy specifically includes: Power deviation according to feedback and demand deviation , adaptively adjust the device scheduling strategy , which is expressed as follows: ; in, and are adjustment coefficients, which respectively control the impact of power deviation and demand deviation on the scheduling strategy; The device power deviation indicates the deviation between the current device power and the expected value. is the water consumption deviation, which indicates the deviation between the current actual water consumption and the predicted value. During the execution of the modified dynamic device scheduling strategy, an energy efficiency regularization term is added , which is used to constrain the power variation of the equipment, is expressed as follows: ; in, is the energy efficiency regularization term, used to smooth power fluctuations; The actual power of the equipment in the previous period; To adjust the parameters, control the penalty intensity of power changes in the current period; It is a cumulative adjustment parameter used to penalize large power fluctuations over a long period of time; Indicates the past The total amount of power fluctuation within a time period is used to smooth the power changes in multiple time periods.

[0012] Furthermore, the adaptive optimization specifically includes: ; in, is the water demand prediction model parameter at the current moment; is the learning rate, which controls the effect of the bias on the update of the water demand forecasting model; is the regularized learning rate, which controls the impact of energy efficiency regularization on the parameter update of the water demand prediction model; and They are power deviation and water consumption deviation, respectively, and are used to adjust the accuracy of the water demand prediction model.

[0013] Furthermore, the dynamic closed-loop feedback adjustment is expressed as follows: ; in, The updated scheduling strategy will be used for the operation control of the equipment in the next period; and is the adjustment coefficient, used to control the power deviation and demand deviation Impact on scheduling strategies; It is a power smoothing regularization term to prevent power fluctuations caused by frequent start and stop of equipment; is the regularization coefficient, which is used to adjust the weight of the smoothing regularization term in the scheduling strategy; The adaptive dynamic learning mechanism is expressed as follows: ; in, The updated water demand forecasting model parameters are used for demand forecasting in the next period; is the learning rate, controlling the influence of power and demand deviation on the updating of water demand forecasting model; is the weight of the regularization term, ensuring that the water demand forecasting model considers smoothing regularization during the update process; is a power smoothing regularization term, which prevents the negative impact of power fluctuations on the learning of the water demand prediction model.

[0014] Furthermore, the S5 also includes designing a power fluctuation suppression mechanism in the campus direct drinking water equipment to reduce equipment power fluctuations by further optimizing the scheduling strategy, thereby ensuring that the system can maintain a stable energy consumption level for a long time, as shown below: ; in, is the total energy consumption smoothing regularization term, which is used to adjust the power changes and demand fluctuations in multiple periods; To adjust the coefficient, control the impact of power changes on the equipment in each time period; and They are the equipment power of the current period and the previous period respectively; It is the cumulative demand smoothing coefficient, which controls the impact of long-term demand fluctuations on power regulation.

[0015] In a second aspect of the present invention, a campus direct drinking water energy consumption optimization system based on energy-saving control is provided, the system comprising: The campus data collection unit is used to collect key data of campus drinking water to build a water demand prediction model based on a multi-layer long short-term memory network. The key data includes historical water consumption, weather data, and student activity data, including: Decompose the historical water consumption in time series, extract the trend, seasonality and residual features, combine the trend, seasonality, residual features, weather data and student activity data into a multi-feature input matrix, build a water demand forecasting model based on a multi-layer long short-term memory network based on the multi-feature input matrix as input, and use the water demand forecasting model to predict future periods to water demand; The direct drinking water prediction unit is used to generate a dynamic equipment scheduling strategy and execute it on the campus direct drinking water equipment according to the prediction results of the water demand prediction model and the energy consumption and power of the current equipment; the dynamic equipment scheduling strategy includes a peak period preheating strategy and a valley period energy saving strategy; The direct drinking water scheduling unit is used to monitor the data after executing the dynamic equipment scheduling strategy in real time, calculate the power deviation between the dynamic equipment scheduling strategy and the actual power of the real-time monitoring equipment, and the water consumption deviation between the real-time actual water consumption and the predicted water consumption. If the power deviation or water consumption deviation exceeds the preset threshold, the adaptive scheduling mechanism is triggered to correct the dynamic equipment scheduling strategy; A scheduling strategy optimization unit is used to adaptively optimize the modified dynamic device scheduling strategy according to actual feedback data of executing the modified dynamic device scheduling strategy to obtain the device scheduling strategy to ensure the optimization of energy saving effect; The equipment closed-loop optimization unit is used to perform dynamic closed-loop feedback adjustment based on the equipment scheduling strategy combined with the power deviation and water consumption deviation in actual operation. At the same time, an adaptive dynamic learning mechanism is designed in the dynamic closed-loop feedback adjustment. By combining the power deviation and demand deviation with the adaptive adjustment of the model, the long-term prediction capability of the water demand prediction model is improved.

[0016] The beneficial technical effects of the present invention are at least as follows: In view of the problem that the existing technology cannot cope with the fluctuation of water demand, the present invention adopts a water demand prediction model based on deep learning, which can predict future water demand in advance based on historical water use data, weather conditions, students' work and rest time and other multi-dimensional data. The system dynamically schedules the operating time and intensity of the direct drinking water equipment according to the prediction results. Before the peak water use period, the system will automatically preheat the water source and improve the equipment operation efficiency, and reduce or shut down the operation of some equipment during the off-peak period, thereby effectively avoiding excessive operation of the equipment during unnecessary time periods and solving the energy waste problem caused by the traditional fixed scheduling mode.

[0017] In response to the water leakage problem of the complex pipeline system in the campus, the present invention introduces a pipeline monitoring system based on the compressed sensing algorithm, which realizes the state perception of the entire pipeline system through a small amount of sensor data. This technology can detect the water pressure changes in the pipeline in real time and find the possible leakage location in time, solving the inefficiency problem of relying on manual inspection and water meter comparison in the existing technology. Through this real-time monitoring, the system can remind maintenance personnel to perform maintenance in time, avoiding the waste of energy and increased maintenance costs caused by long-term water leakage problems.

[0018] The present invention also designs a closed-loop energy consumption optimization feedback mechanism. The system not only dynamically adjusts the working status of the equipment according to AI predictions, but also continuously optimizes the prediction model and equipment scheduling strategy according to the real-time monitoring of water consumption and pipeline status. This feedback mechanism can adaptively adjust the operating mode to cope with changes in the external environment, further improving the energy-saving effect of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The present invention is further described using the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative work.

[0020] Figure 1 This is a flow chart of the method for optimizing energy consumption of direct drinking water on campus based on energy-saving control of the present invention. DETAILED DESCRIPTION

[0021] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0022] In one or more embodiments, Figure 1 As shown, a method for optimizing campus direct drinking water energy consumption based on energy-saving control is disclosed, and the method comprises the following steps: S1. Collect key data of campus drinking water to build a water demand prediction model based on a multi-layer long short-term memory network. The key data includes historical water consumption, weather data and student activity data, including: Decompose the historical water consumption in time series, extract the trend, seasonality and residual features, combine the trend, seasonality, residual features, weather data and student activity data into a multi-feature input matrix, build a water demand forecasting model based on a multi-layer long short-term memory network based on the multi-feature input matrix as input, and use the water demand forecasting model to predict future periods to water demand.

[0023] Specifically, the present invention combines the most representative input data of the campus direct drinking water system. The specific input includes: Historical water consumption : Record past time periods The actual water consumption is assumed here to be recorded hourly. It reflects the actual demand fluctuations of the system in different periods of time and can provide trend information on future demand.

[0024] Weather data :The weather has a significant impact on campus water use, such as when temperatures rise, water use increases, while rainy days may reduce demand. Specific variables include temperature, humidity, and precipitation.

[0025] Student activity data : Indicates time periods based on campus schedules and activities (such as exam weeks, holidays, assemblies, etc.) This data has a direct impact on water demand. Use 0 / 1 coding to represent the intensity of activities, for example: 1 during exam weeks and assemblies, and 0 at other times.

[0026] Furthermore, since the water consumption of the campus is not only a linear growth or decline, but also has periodicity, trend and uncertainty, the present invention performs a series of feature engineering processing on the data, and the specific steps are as follows: Historical water consumption Decompose the time series and extract its trend, seasonality (periodicity) and residual parts, which are expressed as (trend), (seasonal) and (Residual). This step helps the present invention to better identify the long-term and short-term volatility of campus water use.

[0027] ; Among them, the trend item Indicates changes in water demand over the long term. It reflects daily periodic fluctuations, such as peak water consumption in the morning before classes, during lunch time, and after school. Means capturing unpredictable fluctuations, such as sudden events or weather anomalies.

[0028] Furthermore, in order to ensure that the prediction model can comprehensively consider the impact of weather and student activities on water consumption, the present invention combines multiple features into an input matrix : ; The matrix captures the combined impact of trends, cyclicality, weather and activities on water use.

[0029] Specifically, including Represents trend data for the past two hours. Represents seasonal data for the past two hours. Indicates current weather information (such as temperature, humidity, precipitation). Represents the current density of student activities (such as exams, assemblies, etc.). This combination of multiple features ensures that the model can fully understand the complex dependencies between history and the future.

[0030] Furthermore, a multi-layer long short-term memory network (LSTM) was used to predict the water demand of the campus direct drinking water system. This is because LSTM can capture long dependencies in time series data and is particularly suitable for predicting water use data with periodicity and trend.

[0031] LSTM network structure: input matrix Calculate the future period through LSTM network to Water demand The core calculation process of the LSTM network is: ; in, For the moment The LSTM hidden state represents the memory information of the network. is the weight matrix of the LSTM network, For bias. is the activation function, which is used to output the hidden state at the next moment.

[0032] The final water demand forecast results are: ; in are the network parameters obtained through training with historical data.

[0033] Furthermore, in the patent scenario, sudden and drastic changes in campus water consumption may occur due to factors such as seasons and student group activities. In order to make the model better adapt to these unstable changes, an offset regularization term is introduced in the loss function to limit the deviation of the model when demand changes sharply: ; in, For the predicted water consumption, The actual water consumption. It is an additional designed offset coefficient, which indicates the rate of change of water consumption from the previous moment to the current moment (if the rate of change is large, large, and vice versa), thus ensuring that the model is more sensitive to the prediction of drastic changes. is the regularization coefficient, which is used to control the weight of the offset term. The design of this offset regularization term is specifically optimized for unpredictable sudden water changes in the campus direct drinking water system, such as campus collective activities or sudden weather conditions, which can greatly improve the prediction accuracy of the model in these situations.

[0034] Furthermore, the weight parameters of the LSTM network The training is performed using historical data, and the loss function combining mean square error (MSE) and offset regularization is used for optimization. The specific process is as follows: During the model training process, the weights in the LSTM network are adjusted by minimizing the above loss function. and bias .

[0035] After each training iteration, according to the residual term Adjust the learning rate of the network to further improve the model's prediction accuracy for future water demand.

[0036] The final output of this step is the water demand forecast for the next 24 hours. This output provides the demand at different time periods in the future, especially identifying the peak and valley periods of water use. This forecast value will be directly input into the next step for equipment scheduling decisions.

[0037] S2. Based on the prediction results of the water demand prediction model and combined with the energy consumption and power of the current equipment, a dynamic equipment scheduling strategy is generated and executed on the campus direct drinking water equipment; the dynamic equipment scheduling strategy includes a peak period preheating strategy and a valley period energy-saving strategy.

[0038] Specifically, the specific input is as follows: :Forecast water demand in the next 24 hours and provide basic demand information for equipment scheduling.

[0039] : Current time The actual energy consumption of the device, which represents the current power demand of the system.

[0040] :Efficiency coefficient of the equipment, which indicates the energy utilization efficiency of the equipment under different loads. Usually changes with load.

[0041] : System energy consumption upper limit, which means that in each time period, the total energy consumption of all devices cannot exceed the threshold value to ensure that the device will not be overloaded.

[0042] Furthermore, based on water demand forecasts , combining historical water consumption and current equipment power to generate dynamic equipment scheduling strategies The core goal of scheduling is to dynamically adjust the power of equipment through demand forecasting, so as to meet the demand during peak periods and avoid wasting energy during off-peak periods.

[0043] Peak preheating strategy: When the upcoming water consumption peak is predicted, the equipment will enter the preheating state in advance to ensure that the equipment is in the best operating state during the peak period to avoid users waiting. The formula is as follows: ; in, is the threshold coefficient for peak period judgment. It is the historical average water usage and is used as a benchmark to identify fluctuations in demand. Indicates the power level at which the equipment is preheated in advance to ensure that the equipment is in good condition when the peak water usage period arrives. Indicates that the device is in normal operation.

[0044] Energy saving strategy during off-peak period: If the system predicts that the future water demand is low ( If the power consumption is below a certain threshold), the device will automatically enter energy-saving mode, reduce power or even shut down some devices to reduce energy consumption. The formula is as follows: ; in, is the threshold coefficient of the trough period. Indicates that the device has entered standby or shutdown mode to further reduce energy consumption. Indicates that the equipment is running at low load, reducing energy consumption.

[0045] Furthermore, the energy efficiency of the equipment and the energy consumption limit of the system need to be fully considered in the scheduling strategy to ensure that the equipment operates efficiently and does not exceed the upper limit of the system energy consumption. and energy consumption cap , to optimize device power.

[0046] ; in, For the current time period The actual operating power of the device. is the equipment efficiency coefficient, which fluctuates with load changes and indicates the energy efficiency of the equipment. The system energy consumption limit ensures that the device does not exceed the maximum power at any time. Through this device energy efficiency optimization, it is ensured that the device can work with high energy efficiency under different loads and will not exceed the energy consumption limit.

[0047] Furthermore, in actual operation, frequent start and stop of equipment will lead to energy waste and equipment loss. Therefore, the present invention designs a regularization term specifically used to reduce the fluctuation of equipment in power regulation, smooth the power curve of equipment, and ensure the stable working state of equipment: ; in, is the power fluctuation regularization term, which indicates the fluctuation amplitude of the device power in a continuous time period. is the power of the device in the previous time period. is a tuning parameter used to balance the penalty for power fluctuations. By adding this regularization term, the equipment will not start and stop drastically when the load changes, reducing equipment wear and the volatility of power consumption.

[0048] Furthermore, the scheduling strategy The output contains the device status and power control information for each time period, ensuring efficient operation of the system and saving energy: Device power-on status: whether it is in preheating or off mode.

[0049] Equipment power level: Power that is dynamically adjusted based on demand forecasts.

[0050] Power smoothing control: Smooth the power curve through regularization terms to reduce device fluctuations.

[0051] S3. Real-time monitoring of the data after the execution of the dynamic equipment scheduling strategy, calculating the power deviation between the dynamic equipment scheduling strategy and the actual power of the real-time monitoring equipment, and the water consumption deviation between the real-time actual water consumption and the predicted water consumption. If the power deviation or water consumption deviation exceeds the preset threshold, the adaptive scheduling mechanism is triggered to correct the dynamic equipment scheduling strategy.

[0052] Specifically, the primary task of this step is to monitor the equipment operating status and water consumption in real time and compare it with the equipment scheduling strategy in step 2. and the predicted water consumption in step 1 Compare, generate feedback signals, and adjust subsequent equipment scheduling strategies. The core operations include the following steps: Real-time monitoring of the actual power of the device and compare it with the target power in the scheduling strategy Compare and calculate the power deviation : ; like Exceeding the set threshold , the system determines that the equipment power operation deviation is large, there may be equipment abnormality or failure, and an adjustment signal needs to be triggered. This step compares the actual power of the equipment with the power target in the scheduling strategy to ensure that the equipment operates normally according to the scheduling strategy. If the power deviation exceeds the reasonable range, it means that the equipment is not working as planned, which may be caused by system failure, equipment aging, or external factors.

[0053] The actual water consumption in real time and forecast water consumption Compare and calculate the demand deviation : ; like Exceeding the threshold , it indicates that there is a large error in the water consumption prediction, and the system should trigger the model adaptive adjustment mechanism; water consumption deviation detection can timely capture the deviation of the prediction model by comparing the actual water consumption with the predicted water consumption. When thresholds are exceeded, the system automatically adjusts the forecast model to better adapt to current water demand fluctuations.

[0054] Furthermore, in response to the energy-saving requirements in the patent scenario, an adaptive scheduling mechanism based on deviation correction is designed. Deviation from water consumption If a power deviation or water consumption deviation is detected, the system will adjust the scheduling strategy. Make corrections. The correction formula is: ; in, is the revised scheduling strategy. and are adjustment coefficients, which are used to control the impact of power deviation and water consumption deviation on the scheduling strategy. is the device power deviation, The adaptive scheduling mechanism can automatically adjust the equipment scheduling strategy according to the real-time deviation information, thereby improving the energy efficiency of the system. When the power deviation is large, the equipment operation status in the scheduling strategy will be reduced or shut down to avoid energy waste; when the water consumption deviation is large, the system will increase or decrease the equipment load to optimize the water supply.

[0055] Furthermore, in the campus direct drinking water system, water leakage often causes unnecessary energy consumption. Therefore, this step combines the real-time monitoring of pipeline water pressure or flow data , an innovative water leakage detection mechanism was designed. This mechanism detects potential water leakage problems in pipes by comparing the current water flow status with the historical normal status.

[0056] Among them, water leakage detection identifies pipeline abnormalities by calculating the deviation of water pressure or water flow. The specific formula is: ; in, It is the normal water pressure or water flow in the pipeline in historical data. is the difference between the current pipeline water pressure / flow and the historical normal state. Exceeding the threshold , the system determines that there may be a water leakage problem, triggers an alarm signal, and records the abnormal situation for maintenance. By monitoring the water pressure or water flow of the pipeline in real time, the system can quickly detect pipeline leakage problems and prevent the waste of water resources and extra energy consumption of equipment caused by long-term undetected leakage.

[0057] Furthermore, frequent start and stop of equipment will lead to excessive power fluctuations, increase energy consumption and equipment wear. Therefore, in order to reduce the volatility of equipment power, a power smoothing regularization term is designed , used to regulate the power variation of the equipment.

[0058] Power smoothing formula: Smooth the operating status of the device by limiting the power change in the current time period: ; in, is a power smoothing regularization term, which is used to reduce the drastic changes in power. It is the actual power of the equipment in the previous period. The power smoothing regularization term is used to control the penalty for power variation. The system can avoid frequent startup and shutdown of equipment through the power smoothing regularization term, thereby reducing power fluctuations and improving the operating stability and life of the equipment. This term is particularly important when dealing with drastic power changes.

[0059] S4. Adaptively optimize the modified dynamic device scheduling strategy according to the actual feedback data of executing the modified dynamic device scheduling strategy to obtain the device scheduling strategy to ensure the optimization of energy saving effect.

[0060] Specifically, according to the power deviation fed back in step 3 and demand deviation , adaptively adjust the device scheduling strategy The purpose is to reduce the gap between equipment operation and water demand and ensure that the power level of the equipment is adapted to the demand, avoiding energy waste or insufficient equipment load. The adaptive scheduling optimization formula is as follows: ; in, It is the optimized scheduling strategy for the next period. and are adjustment coefficients, which respectively control the impact of power deviation and demand deviation on the scheduling strategy. The device power deviation indicates the deviation between the current device power and the expected value. is the water consumption deviation, which indicates the deviation between the current actual water consumption and the predicted value. This formula dynamically adjusts the equipment scheduling strategy based on the feedback power and demand deviation to ensure that the system operation is closer to the actual demand. For example, if the power deviation If the water consumption is larger, the system will adjust the equipment running time or power to prevent excessive energy consumption; If the water consumption is large, the system will increase or decrease the equipment operating load accordingly to better match the current water demand.

[0061] Furthermore, in order to prevent power fluctuations and equipment wear caused by frequent start and stop of equipment, an innovative energy efficiency regularization term is designed , which is used to constrain the power variation of the device, keep the device running under a stable load, and avoid sudden power fluctuations. The energy efficiency regularization formula is: ; in, is the energy efficiency regularization term, which is used to smooth power fluctuations. It is the actual power of the device in the previous period. To adjust the parameters, control the penalty intensity of power changes in the current period. It is a cumulative adjustment parameter used to penalize large power fluctuations over a long period of time. Indicates the past The total amount of power fluctuation in a period is used to smooth the power changes in multiple periods. Control power fluctuations in a single period, and Used to penalize cumulative power fluctuations, the system can better balance energy efficiency and load fluctuations of the equipment and avoid overly frequent starts and stops. This innovative energy efficiency regularization term ensures that the equipment operates under smooth loads, extending equipment life and further improving energy efficiency.

[0062] Furthermore, in order to cope with changes in the external environment, weather, and user needs in different time periods, the system has designed a dynamic learning mechanism based on energy efficiency feedback. By learning from historical data, the prediction model and equipment scheduling strategy can be adaptively adjusted according to real-time energy efficiency feedback, allowing the entire system to respond to demand changes more accurately and intelligently. The dynamic learning formula is: ; in, is the prediction model parameter at the current moment. is the learning rate, which controls the effect of the bias on model updates. is the regularized learning rate, which controls the impact of energy-efficient regularization on model parameter updates. and They are power deviation and water consumption deviation, which are used to adjust the accuracy of the model. Dynamic adjustment, the system can continuously optimize the prediction model based on historical feedback to ensure future scheduling strategies More accurate. Learning rate and The influence of bias and regularization terms on model learning is controlled, thereby achieving adaptive learning and improving the long-term optimization capability of the system.

[0063] Finally, the output of this step includes the optimized scheduling strategy , updated model parameters , and the power smoothing regularization term These outputs will directly affect the scheduling decisions in subsequent time periods, ensuring that the system can continuously optimize itself based on real-time feedback and further reduce energy consumption.

[0064] : The updated scheduling strategy is used for equipment scheduling execution in the next period.

[0065] : Updated prediction model parameters to improve the accuracy of subsequent water consumption predictions.

[0066] Power optimization report: records the power fluctuations of the equipment, water consumption deviations, and energy efficiency feedback results for subsequent scheduling and system monitoring.

[0067] This step is closely related to step 3. It adjusts the equipment scheduling strategy based on feedback, and further achieves continuous optimization through energy efficiency regularization terms and adaptive learning mechanisms. By combining real-time feedback data and historical learning, the system makes the equipment operating status respond to water demand more accurately, while reducing unnecessary power fluctuations and extending the service life of the equipment.

[0068] As a closed-loop optimization link of the entire system, this step ensures that the energy efficiency of the campus direct drinking water system continues to improve during long-term operation by combining energy consumption feedback, adaptive scheduling and energy efficiency regularization. The optimization results of each scheduling cycle will be input into the next round of scheduling execution to ensure that the system can continue to adapt to changing needs and environments and achieve energy-saving goals.

[0069] S5. Perform dynamic closed-loop feedback adjustment based on the equipment scheduling strategy and the power deviation and water consumption deviation in actual operation. At the same time, design an adaptive dynamic learning mechanism in the dynamic closed-loop feedback adjustment. By combining the power deviation and demand deviation with the adaptive adjustment of the model, the long-term prediction capability of the water demand prediction model can be improved.

[0070] Specifically, combined with the power deviation in actual operation and water consumption deviation , and make dynamic feedback adjustments. The scheduling strategy adjusted through feedback can provide a more accurate reference for the operation of the equipment in the next period. The closed-loop feedback formula of the scheduling strategy is: ; in, Indicates the updated scheduling strategy, which will be used for the operation control of the equipment in the next period. and Indicates the adjustment coefficient, which is used to control the power deviation separately and demand deviation Impact on scheduling strategies. Represents the power smoothing regularization term, which prevents power fluctuations caused by frequent start and stop of the equipment. represents the regularization coefficient, which is used to adjust the weight of the smoothing regularization term in the scheduling strategy. This formula adjusts the equipment scheduling strategy through real-time feedback to ensure that the strategy can be adaptively adjusted when the equipment power and actual water demand fluctuate. At the same time, by introducing the power smoothing regularization term , ensuring that the equipment operates efficiently while reducing the impact of frequent start and stop on energy consumption and equipment life.

[0071] Furthermore, in order to further improve the accuracy of the prediction model, the system needs to continuously update the model parameters based on real-time feedback. In order to cope with the impact of external environment (such as seasonal changes, student group activities, etc.) on water consumption, an adaptive dynamic learning mechanism is designed. By combining power deviation and demand deviation with the adaptive adjustment of the model, the long-term prediction ability of the system is improved. The model adaptive update formula is: ; in, Represents the updated model parameters, which are used for demand forecasting in the next period. Denotes the effects of learning rate, control power, and demand bias on model updates. Represents the weight of the regularization term, ensuring that the model takes smoothing regularization into account during the update process. Represents the power smoothing regularization term to prevent the negative impact of power fluctuations on model learning. This model adaptive update formula dynamically adjusts the prediction model parameters by considering power and demand deviations. , thereby improving the accuracy of water consumption prediction in future periods. It can ensure that the model will not produce excessive errors due to power fluctuations during the parameter update process, thereby ensuring the long-term prediction stability of the system.

[0072] Furthermore, in the energy-saving optimization of the direct drinking water system, frequent equipment start and stop will not only cause energy waste, but also shorten the service life of the equipment. Therefore, this step introduces an innovative power fluctuation suppression mechanism, which further optimizes the scheduling strategy to reduce equipment power fluctuations and ensure that the system can maintain a stable energy consumption level for a long time. The comprehensive energy consumption smoothing formula is: ; in, It represents the total energy consumption smoothing regularization term, which is used to adjust the power changes and demand fluctuations in multiple periods. Represents the regulation coefficient, which controls the impact of power changes on the system in each time period. and Indicates the device power in the current period and the previous period. represents the cumulative demand smoothing coefficient, which controls the impact of long-term demand fluctuations on power adjustment. This formula achieves comprehensive smooth optimization of equipment scheduling strategies by suppressing power fluctuations and demand fluctuations. In practical applications, this smoothing mechanism can reduce the increase in energy consumption caused by frequent start-up and shutdown of equipment, ensuring smooth and efficient system operation.

[0073] Furthermore, in closed-loop control, the system not only adjusts the equipment scheduling strategy through feedback information, but also continuously optimizes future demand forecasts through adaptive learning of the prediction model. The equipment operation results of each period will be recorded in real time and input into the next round of prediction and control to achieve closed-loop self-optimization. Through multiple rounds of feedback iteration, the system ensures long-term stability and maximizes energy efficiency.

[0074] The output includes the following parts: Updated Scheduling Strategy : The scheduling strategy adjusted through feedback and regularization guides the operation of equipment in the next period.

[0075] Updated model parameters : The adjusted forecast model parameters are used for demand forecasting in subsequent periods.

[0076] Energy consumption optimization report: records in detail the equipment power fluctuations, demand deviations and changes in scheduling strategies for system optimization.

[0077] This step continuously optimizes the equipment scheduling strategy through iterations of closed-loop control and real-time feedback, and combines adaptive prediction model updates to ensure that the system can self-adjust and adapt to environmental changes during long-term operation. At the same time, the innovative power fluctuation smoothing mechanism greatly reduces the system's energy consumption fluctuations and further extends the service life of the equipment. The closed-loop energy control module forms a complete system adaptive optimization solution by combining feedback adjustment, energy consumption smoothing and prediction model updates. This module is a key link in the continuous optimization of the entire system, ensuring that the equipment is intelligently adjusted according to changes in demand and achieving long-term energy-saving effects.

[0078] In one or more embodiments, a second aspect of the present invention discloses a campus direct drinking water energy consumption optimization system based on energy-saving control, the system comprising: The campus data collection unit is used to collect key data of campus drinking water to build a water demand prediction model based on a multi-layer long short-term memory network. The key data includes historical water consumption, weather data, and student activity data, including: Decompose the historical water consumption in time series, extract the trend, seasonality and residual features, combine the trend, seasonality, residual features, weather data and student activity data into a multi-feature input matrix, build a water demand forecasting model based on a multi-layer long short-term memory network based on the multi-feature input matrix as input, and use the water demand forecasting model to predict future periods to water demand; The direct drinking water prediction unit is used to generate a dynamic equipment scheduling strategy and execute it on the campus direct drinking water equipment according to the prediction results of the water demand prediction model and the energy consumption and power of the current equipment; the dynamic equipment scheduling strategy includes a peak period preheating strategy and a valley period energy saving strategy; The direct drinking water scheduling unit is used to monitor the data after executing the dynamic equipment scheduling strategy in real time, calculate the power deviation between the dynamic equipment scheduling strategy and the actual power of the real-time monitoring equipment, and the water consumption deviation between the real-time actual water consumption and the predicted water consumption. If the power deviation or water consumption deviation exceeds the preset threshold, the adaptive scheduling mechanism is triggered to correct the dynamic equipment scheduling strategy; A scheduling strategy optimization unit is used to adaptively optimize the modified dynamic device scheduling strategy according to actual feedback data of executing the modified dynamic device scheduling strategy to obtain the device scheduling strategy to ensure the optimization of energy saving effect; The equipment closed-loop optimization unit is used to perform dynamic closed-loop feedback adjustment based on the equipment scheduling strategy combined with the power deviation and water consumption deviation in actual operation. At the same time, an adaptive dynamic learning mechanism is designed in the dynamic closed-loop feedback adjustment. By combining the power deviation and demand deviation with the adaptive adjustment of the model, the long-term prediction capability of the water demand prediction model is improved.

[0079] The above disclosures are only some preferred embodiments of the present invention, which certainly cannot be used to limit the scope of rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of the above embodiments and equivalent changes made according to the claims of the present invention are still within the scope of the invention.

Claims

1. A campus direct drinking water energy consumption optimization method based on energy-saving control, characterized in that: The method comprises the following steps: S1. Collect key data of campus drinking water to build a water demand prediction model based on a multi-layer long short-term memory network. The key data includes historical water consumption, weather data and student activity data, including: Decompose the historical water consumption in time series, extract the trend, seasonality and residual features, combine the trend, seasonality, residual features, weather data and student activity data into a multi-feature input matrix, build a water demand forecasting model based on a multi-layer long short-term memory network based on the multi-feature input matrix as input, and use the water demand forecasting model to predict future periods to water demand; S2. Based on the prediction results of the water demand prediction model and the energy consumption and power of the current equipment, a dynamic equipment scheduling strategy is generated and executed on the campus direct drinking water equipment; the dynamic equipment scheduling strategy includes a peak period preheating strategy and a valley period energy saving strategy; S3, real-time monitoring of the data after the execution of the dynamic equipment scheduling strategy, calculating the power deviation between the dynamic equipment scheduling strategy and the actual power of the real-time monitoring equipment, and the water consumption deviation between the real-time actual water consumption and the predicted water consumption. If the power deviation or water consumption deviation exceeds the preset threshold, the adaptive scheduling mechanism is triggered to correct the dynamic equipment scheduling strategy; S4, adaptively optimizing the modified dynamic device scheduling strategy according to the actual feedback data of executing the modified dynamic device scheduling strategy, obtaining the device scheduling strategy, and ensuring the optimization of energy saving effect; S5. Perform dynamic closed-loop feedback adjustment based on the equipment scheduling strategy and the power deviation and water consumption deviation in actual operation. At the same time, design an adaptive dynamic learning mechanism in the dynamic closed-loop feedback adjustment. By combining the power deviation and demand deviation with the adaptive adjustment of the model, the long-term prediction capability of the water demand prediction model can be improved.

2. The campus direct drinking water energy consumption optimization method based on energy-saving control according to claim 1 is characterized in that: The calculation process of the multi-layer long short-term memory network is expressed as follows: ； in, For the moment The LSTM hidden state represents the memory information of the network; is the weight matrix of the LSTM network, is bias; is the activation function, which is used to output the hidden state at the next moment; The water demand forecasting model is used to predict the future period to Water demand , which is expressed as follows: ； in, are the network parameters obtained through historical data training, It is a multi-layer long short-term memory network; The loss function of the water demand prediction model introduces an offset regularization term to limit the deviation of the multi-layer long short-term memory network when the demand changes sharply. It is expressed as follows: ； in, For the predicted water consumption, is the actual water consumption, is the offset coefficient, i.e., the offset regularization term, which indicates the rate of change of water consumption from the previous moment to the current moment. is the regularization coefficient, which is used to control the weight of the offset term. n represents the total duration, and i represents the current moment.

3. The campus direct drinking water energy consumption optimization method based on energy-saving control according to claim 1 is characterized in that: The dynamic device scheduling strategy execution rule is expressed as: If an upcoming peak water usage is predicted, the device will enter the preheating state in advance and execute the peak preheating strategy, as shown below: ； in, is the threshold coefficient for peak period judgment; It is the historical average water consumption, a benchmark used to identify fluctuations in demand; Indicates the power level of the equipment preheating in advance to ensure that the equipment is in good condition when the peak water consumption period arrives; Indicates that the device is in normal operation; If future water demand is forecast to be low, When the power level drops below a certain threshold, the device automatically enters energy-saving mode and executes the low-peak energy-saving strategy, as shown below: ； in, is the threshold coefficient of the trough period, Indicates that the device is in standby or off state. Indicates that the equipment is running under low load; Output the dynamic device scheduling strategy that is finally executed ; Said S2 further comprises: Design energy efficiency control formula, combined with equipment efficiency and energy consumption cap , optimize the device power without exceeding the energy consumption limit, as shown below: ； in, For the current time period The power at which the equipment actually operates; is the equipment efficiency coefficient, which fluctuates with load changes and indicates the energy efficiency of the equipment; To limit the system's energy consumption, ensure that the device does not exceed the maximum power at any time.

4. The campus direct drinking water energy consumption optimization method based on energy-saving control according to claim 1 is characterized in that: The adaptive scheduling mechanism is to modify the scheduling strategy, which is expressed as follows: ； in, is the revised scheduling strategy; and are adjustment coefficients, which are used to control the impact of power deviation and water consumption deviation on the scheduling strategy; is the device power deviation, The water consumption deviation.

5. The campus direct drinking water energy consumption optimization method based on energy-saving control according to claim 1 is characterized in that: The S3 also includes performing a pipeline leakage detection on the campus direct drinking water equipment, specifically: Water leakage detection identifies pipeline anomalies by calculating the deviation of water pressure or water flow, as shown below: ； in, is the normal water pressure or water flow of the pipeline in historical data, It is the difference between the current pipeline water pressure / flow and the historical normal state; like Exceeding the threshold , it is judged that there is a water leakage problem, an alarm signal is triggered, and the abnormal situation is recorded for maintenance; At the same time, the power smoothing regularization term is designed , the power variation range of the control equipment is expressed as follows: ； in, is the power smoothing regularization term, which is used to reduce the drastic changes in power; is the actual power of the equipment in the previous period; It is the adjustment coefficient, which is used to control the penalty intensity of power variation.

6. The campus direct drinking water energy consumption optimization method based on energy-saving control according to claim 1 is characterized in that: The step of adaptively optimizing the modified dynamic device scheduling strategy according to actual feedback data of executing the modified dynamic device scheduling strategy specifically includes: Power deviation according to feedback and demand deviation , adaptively adjust the device scheduling strategy , which is expressed as follows: ； in, and are adjustment coefficients, which respectively control the impact of power deviation and demand deviation on the scheduling strategy; The device power deviation indicates the deviation between the current device power and the expected value. is the water consumption deviation, which indicates the deviation between the current actual water consumption and the predicted value. During the execution of the modified dynamic device scheduling strategy, an energy efficiency regularization term is added , which is used to constrain the power variation of the equipment, is expressed as follows: ； in, is the energy efficiency regularization term, used to smooth power fluctuations; The actual power of the equipment in the previous period; To adjust the parameters, control the penalty intensity of power changes in the current period; It is a cumulative adjustment parameter used to penalize large power fluctuations over a long period of time; Indicates the past The total amount of power fluctuation within a time period is used to smooth the power changes in multiple time periods.

7. The campus direct drinking water energy consumption optimization method based on energy-saving control according to claim 6 is characterized in that: The adaptive optimization specifically includes: ； in, is the water demand prediction model parameter at the current moment; is the learning rate, which controls the effect of the bias on the update of the water demand forecasting model; is the regularized learning rate, which controls the impact of energy efficiency regularization on the parameter update of the water demand prediction model; and They are power deviation and water consumption deviation, respectively, and are used to adjust the accuracy of the water demand prediction model.

8. The campus direct drinking water energy consumption optimization method based on energy-saving control according to claim 6 is characterized in that: The dynamic closed-loop feedback adjustment is expressed as follows: ； in, The updated scheduling strategy will be used for the operation control of the equipment in the next period; and is the adjustment coefficient, used to control the power deviation and demand deviation Impact on scheduling strategies; It is a power smoothing regularization term to prevent power fluctuations caused by frequent start and stop of equipment; is the regularization coefficient, which is used to adjust the weight of the smoothing regularization term in the scheduling strategy; The adaptive dynamic learning mechanism is expressed as follows: ； in, The updated water demand forecasting model parameters are used for demand forecasting in the next period; is the learning rate, controlling the influence of power and demand deviation on the updating of water demand forecasting model; is the weight of the regularization term, ensuring that the water demand forecasting model considers smoothing regularization during the update process; is a power smoothing regularization term, which prevents the negative impact of power fluctuations on the learning of the water demand prediction model.

9. The campus direct drinking water energy consumption optimization method based on energy-saving control according to claim 8 is characterized in that: The S5 also includes designing a power fluctuation suppression mechanism in the campus direct drinking water equipment to further optimize the scheduling strategy, reduce the power fluctuation of the equipment, and ensure that the system can maintain a stable energy consumption level for a long time, as shown below: ； in, is the total energy consumption smoothing regularization term, which is used to adjust the power changes and demand fluctuations in multiple periods; To adjust the coefficient, control the impact of power changes on the equipment in each time period; and They are the equipment power of the current period and the previous period respectively; It is the cumulative demand smoothing coefficient, which controls the impact of long-term demand fluctuations on power regulation.

10. Campus direct drinking water energy consumption optimization system based on energy-saving control, characterized in that: The system comprises: The campus data collection unit is used to collect key data of campus drinking water to build a water demand prediction model based on a multi-layer long short-term memory network. The key data includes historical water consumption, weather data, and student activity data, including: Decompose the historical water consumption in time series, extract the trend, seasonality and residual features, combine the trend, seasonality, residual features, weather data and student activity data into a multi-feature input matrix, build a water demand forecasting model based on a multi-layer long short-term memory network based on the multi-feature input matrix as input, and use the water demand forecasting model to predict future periods to water demand; The direct drinking water prediction unit is used to generate a dynamic equipment scheduling strategy and execute it on the campus direct drinking water equipment according to the prediction results of the water demand prediction model and the energy consumption and power of the current equipment; the dynamic equipment scheduling strategy includes a peak period preheating strategy and a valley period energy saving strategy; The direct drinking water scheduling unit is used to monitor the data after executing the dynamic equipment scheduling strategy in real time, calculate the power deviation between the dynamic equipment scheduling strategy and the actual power of the real-time monitoring equipment, and the water consumption deviation between the real-time actual water consumption and the predicted water consumption. If the power deviation or water consumption deviation exceeds the preset threshold, the adaptive scheduling mechanism is triggered to correct the dynamic equipment scheduling strategy; A scheduling strategy optimization unit is used to adaptively optimize the modified dynamic device scheduling strategy according to actual feedback data of executing the modified dynamic device scheduling strategy to obtain the device scheduling strategy to ensure the optimization of energy saving effect; The equipment closed-loop optimization unit is used to perform dynamic closed-loop feedback adjustment based on the equipment scheduling strategy combined with the power deviation and water consumption deviation in actual operation. At the same time, an adaptive dynamic learning mechanism is designed in the dynamic closed-loop feedback adjustment. By combining the power deviation and demand deviation with the adaptive adjustment of the model, the long-term prediction capability of the water demand prediction model is improved.

Images