Optimal control method of central air-conditioning system for building load forecasting

Abstract

The present invention discloses a central air-conditioning system optimization control method for building load prediction. By constructing a building load prediction model based on a long short-term memory network, combined with real-time collected building electricity load, indoor and outdoor temperature, humidity and equipment operating power data, the building cooling load demand in future time periods is accurately predicted. On this basis, the time-of-use electricity price information is used to dynamically adjust the cooling capacity distribution strategy of the central air-conditioning system to maximize the reduction of system energy consumption and operating costs while meeting the load demand. The present invention also constructs an energy efficiency model of key equipment of the central air-conditioning system, and combines genetic algorithms to optimize the hourly operating parameters of the refrigeration unit, water pump and cooling tower to achieve continuous optimization of system energy efficiency. Compared with the prior art, the present invention effectively solves the problems of insufficient load fluctuation response and insufficient energy consumption optimization, and can significantly improve the operating efficiency and economy of the building central air-conditioning system.

Description

Technical Field

[0001] The present invention relates to the technical field of building energy management and optimization control of central air-conditioning systems, and in particular to an optimization control method for central air-conditioning systems oriented to building load prediction. Background Art

[0002] As the scale and complexity of buildings continue to increase, the energy consumption of central air-conditioning systems has become increasingly prominent. In most modern buildings, central air-conditioning systems are one of the main sources of energy consumption, accounting for more than 50% of the total energy consumption of buildings. The volatility of building load demand, the complexity of time-of-use electricity prices, and the energy efficiency issues of various equipment make it a key issue to achieve efficient operation of central air-conditioning systems. To this end, existing building energy efficiency management systems and control solutions have proposed several optimization control methods to reduce energy consumption and improve system operation efficiency through technical means.

[0003] China's public patent CN104571068A proposes a solution to improve the overall operating efficiency by optimizing and controlling the distributed energy system; the solution mainly focuses on the management of distributed energy, and improves the overall energy efficiency of the system by controlling the power generation and distribution of energy suppliers (such as solar energy, wind energy, etc.); however, the limitation of this technical solution is that it is too focused on the management of energy supply, and does not effectively respond to the actual load demand changes of the building; specifically, the solution does not fully combine the dynamic load fluctuation characteristics of the building's central air-conditioning system in actual use scenarios, ignores the fluctuations in the building's cooling load demand in different time periods, and the impact of fluctuations in time-of-use electricity prices on the overall operating cost of the building; therefore, although the control solution of this patent has a certain optimization effect on energy distribution, it still has great limitations in dealing with complex load fluctuations in the central air-conditioning system and optimizing operating costs;

[0004] China's public patent JP2021018055A proposes a building management method for optimizing the operation of a central air-conditioning system by automatically adjusting comfort constraints. The focus of this technology is to meet the user's comfort needs by sensing the environmental parameters inside the building (such as temperature, humidity, etc.) and dynamically adjusting the operating parameters of the air-conditioning system. However, while paying attention to user comfort, this solution fails to fully consider the energy efficiency and operating costs of the system. Especially during peak electricity price periods, the system still needs to maintain a high operating power to ensure comfort, resulting in excessive energy consumption. In addition, the solution lacks accurate prediction of building load demand, and the system optimization is mainly based on the current environmental status, without long-term load prediction capabilities. Based on this, a central air-conditioning system optimization control method for building load prediction is required. Summary of the invention

[0005] To achieve the above object, the present invention provides the following solution: a central air conditioning system optimization control method for building load prediction, comprising the following control steps:

[0006] Step 1: Collect the historical electricity load data L of the building hist (t), indoor and outdoor temperature data T in (t) and T out (t), outdoor relative humidity RH out (t), and the equipment operating power P of the building equipment equip (t), indicating that the above data is used to construct a building load forecasting model based on the long short-term memory network LSTM, and the prediction formula is: L pred (t+1)=f(L hist (t), L hist (t-1), ...., T in (t), T out (t), RH out (t), P equip (t), to predict the hourly load demand in future periods;

[0007] Among them, L pred (t+1) represents the future load forecast value at time t+1. Function f is obtained through LSTM model training. The input includes data on historical power load, indoor and outdoor temperature, outdoor relative humidity, and equipment operating power. The back propagation algorithm is used to optimize the network weights and reduce the mean square error.

[0008] Step 2: Use the predicted load L output by the building load prediction model in step 1 pred (t+1), combined with the time-of-use electricity price P(t) information, formulate the hourly cooling capacity Q cool (t) allocation strategy, which meets the load demand while reducing energy costs, Optimizing the objective function minimizes energy cost;

[0009] Among them, P(t) represents the time-of-use electricity price at time t, E AC (t) represents the energy consumption of the air conditioning system at time t, and the cooling capacity Q cool (t) According to the predicted load L pred (t+1) is dynamically adjusted, and the system energy consumption E AC (t)Through Calculate, where COP(t) is the energy efficiency coefficient of the refrigeration unit;

[0010] Step 3: By building energy efficiency and function models of key equipment in the central air-conditioning system, such as refrigeration units, water pumps and cooling towers, the energy efficiency performance of the equipment under different working conditions is evaluated. The energy efficiency model deduces the energy efficiency coefficient COP(t) of the equipment according to the equipment operation curve and environmental parameters. The calculation formula is: COP(t) = g(T out (t), L pred (t), P equip (t));

[0011] Where g is the external temperature T out (t), predicted load L pred (t) and equipment power to calculate the energy efficiency coefficient COP(t);

[0012] Step 4: Optimize the hourly operating parameters P of the central air-conditioning system through genetic algorithm equip (t), including the load rate of the refrigeration unit, the flow rate of the water pump and the speed of the cooling tower fan, the optimization process is achieved through the following steps:

[0013] The initial population of equipment operating parameters is randomly generated to represent the operating status of the equipment under different working conditions. The fitness value of each population is calculated. The fitness function is the negative value of the total energy consumption:

[0014] F itness (t)=-(E AC (t)+E pump (t)+E fan (t))

[0015] Among them, E AC (t), E pump (t), E fan (t) are the energy consumption of central air conditioner, pump and fan in time period t, respectively. Then, the roulette wheel selection algorithm is used to select individuals with high fitness for crossover, and new individuals are introduced by mutation operation to increase diversity. After multiple iterations, the population is updated and the optimal equipment operation parameters that minimize energy consumption are finally obtained.

[0016] Step 5: Based on load forecast L pred (t+1) and equipment operating power P equip (t), for cooling capacity Q cool (t) and equipment operating power P equip (t) Make dynamic adjustments to cope with fluctuations in building load demand, thereby reducing energy costs during peak periods while ensuring system operating efficiency.

[0017] Furthermore, the historical power load data is collected through multiple data sources, including the building's power meter data, the central air-conditioning system's historical energy consumption records, and the real-time load monitoring data of the building's floors, and the system automatically updates the historical power load data every 15 minutes. The weighted average method is used to fuse the historical data from multiple data sources, and the K-means clustering algorithm is used to group similar historical load patterns. By combining multiple data sources to collect historical load data, the energy consumption history of the building can be more comprehensively reflected, and the data can be updated regularly to ensure the timeliness of load forecasting. The combination of the weighted average method and the K-means clustering algorithm can identify and process similar historical load patterns, improve the adaptability and prediction accuracy of the prediction model under different load fluctuation conditions, and further optimize the operation and control effect of the central air-conditioning system.

[0018] Furthermore, during the training process, the LSTM-based building load prediction model adjusts the learning rate according to the rate of change of the loss function during the training process. When the loss function no longer decreases significantly in multiple training rounds, the model will dynamically reduce the learning rate to ensure that the LSTM model continues to converge when facing complex load changes. The error measurement in the training process uses the mean square error MSE indicator, and the optimization process of the loss function is completed through the Adam optimizer. Dynamic adjustment of the learning rate can effectively prevent the model from overfitting during the training process, while ensuring the stable convergence of the model under complex load conditions. Using the mean square error MSE as a measurement indicator can accurately measure the prediction error, and the Adam optimizer can speed up the convergence speed of the training process, reduce the model training time, and ensure the robustness and efficiency of the model in building load prediction.

[0019] Furthermore, the source of the time-of-use electricity price information is the real-time electricity price information of the electricity market. Through the network interface, it is connected to the electricity market platform in real time to obtain the time-of-use electricity price fluctuations of the day and future periods, compare the time-of-use electricity price information with the historical electricity price data of the building, and use the time series prediction ARIMA model to predict the electricity price fluctuations in future periods, and adjust the cooling capacity Q in advance. cool (t) allocation strategy. By obtaining the time-of-use electricity price information of the electricity market in real time, the system can respond to electricity price fluctuations in real time and avoid high energy consumption during peak hours. Combined with the time series prediction model of historical electricity price data, it can predict future electricity price trends in advance and adjust the cooling capacity allocation strategy of the central air-conditioning accordingly, effectively reducing the energy cost of buildings during peak electricity price periods and improving the economic benefits of the system.

[0020] Furthermore, the cooling capacity allocation strategy assigns priorities according to the temperature comfort requirements and population density of different areas in the building. Areas with higher priorities obtain more cooling capacity allocation when electricity prices are low, and prioritize the minimum necessary cooling demand when electricity prices are high. By setting priorities for the temperature comfort requirements and population density of different areas inside the building, the system can flexibly adjust the cooling capacity allocation strategy of each area. When electricity prices are low, cooling capacity is allocated to high-demand areas first to ensure comfort; during peak electricity price periods, cooling capacity is limited to low-priority areas to reduce energy consumption, thereby achieving refined management of the central air-conditioning system and further improving the overall energy efficiency of the building.

[0021] Furthermore, in the process of establishing the energy efficiency and functional model of the central air-conditioning system, the historical operation data of the refrigeration unit, water pump and cooling tower are regularly collected to generate an equipment aging model, and the actual energy efficiency COP(t) of the equipment is dynamically adjusted in combination with the service life of the equipment, the frequency of daily maintenance and the historical failure rate. The equipment aging model can adjust the energy efficiency parameters of the equipment in real time in combination with the historical data of the refrigeration unit, water pump and cooling tower and the actual service life of the equipment, and reflect the energy efficiency changes of the equipment due to aging and different maintenance frequencies. This dynamic adjustment ensures that the energy efficiency model of the central air-conditioning system can truly reflect the operating status of the equipment, improve the control accuracy of the system, reduce energy waste and extend the service life of the equipment.

[0022] Furthermore, the energy efficiency models of the refrigeration unit, water pump and cooling tower equipment are calibrated through historical operating data and equipment performance curves provided by equipment manufacturers. By real-time monitoring of the inlet and outlet water temperature difference of the equipment and the cooling air temperature environmental parameters, and comparing the real-time data with the equipment performance curve, the load factor of the refrigeration unit and the parameters of the flow curve of the water pump in the energy efficiency model are calibrated. By real-time monitoring of the environmental parameters of the equipment and comparing them with the performance curve provided by the equipment manufacturer, the key parameters in the energy efficiency model are automatically calibrated to reflect the actual working status of the equipment.

[0023] Furthermore, when the genetic algorithm is used to optimize the hourly operating parameters of the central air-conditioning system, an individual selection strategy based on the tournament selection method is adopted. In each generation of the population, several individuals are randomly selected from the current population for comparison, and individuals with higher fitness are subjected to crossover and mutation operations to generate new individuals. By adopting the tournament selection method, individuals with higher fitness are effectively selected in each generation of the population for optimization operations, thereby accelerating the convergence process of the genetic algorithm. This selection method can ensure that the diversity of the population is maintained during the optimization process and avoid falling into a local optimal solution, thereby improving the global search capability of the central air-conditioning system hourly operating parameter optimization and achieving optimization of system performance and energy efficiency.

[0024] Furthermore, the crossover operation in the genetic algorithm uses multi-point crossover, which performs gene exchange at multiple gene points. During the crossover process, each new individual generated is checked for constraints to ensure that the device operating power P equip (t) Always within the safe operating range of the equipment.

[0025] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0026] 1. This application can accurately predict the building cooling load demand in the future period by constructing a building load prediction model based on long short-term memory network. The prediction model can take into account multiple variables such as historical power load data, indoor and outdoor temperature, humidity and equipment operating parameters, so that the system can be pre-adjusted according to future load demand, which not only improves the responsiveness of the system, but also effectively reduces the loss of operating efficiency caused by load fluctuations.

[0027] 2. This application combines time-of-use electricity price information to optimize the allocation of cooling capacity. This solution not only takes into account the user's comfort needs, but also dynamically adjusts the cooling capacity allocation by predicting electricity price fluctuations. When the electricity price is low, the cooling reserve is increased in advance, and when the electricity price is high, the operating power is reduced to reduce the energy consumption cost during peak hours. By combining the optimization control method of load forecasting and electricity price information, it is possible to maximize the reduction of system energy consumption while ensuring comfort and achieve higher economic benefits.

[0028] 3. This application constructs an energy efficiency model of the key equipment of the central air-conditioning system, and optimizes the hourly operating parameters of the system in combination with a genetic algorithm. On the basis of dynamic load prediction and energy efficiency models, the load rate of the refrigeration unit, the flow rate of the water pump and the speed operating parameters of the cooling tower fan are optimized to further reduce the energy consumption of the system. The present invention can take into account both user comfort and minimization of system energy consumption, thereby achieving higher energy utilization efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0030] Figure 1 It is a flow chart of the steps of the present invention. DETAILED DESCRIPTION

[0031] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0032] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0033] Example

[0034] like Figure 1 As shown, the central air conditioning system optimization control method for building load prediction in this embodiment includes the following control steps:

[0035] Step 1: Collect the historical electricity load data L of the building hist (t), indoor and outdoor temperature data T in (t) and T out (t), outdoor relative humidity RH out (t), and the equipment operating power P of the building equipment equip (t), indicating that the above data is used to construct a building load forecasting model based on the long short-term memory network LSTM, and the prediction formula is: L pred (t+1)=f(L hist (t), L hist (t-1), ...., T in (t), T out (t), RH out (t), P equip (t), to predict the hourly load demand in future periods;

[0036] Among them, L pred (t+1) represents the future load forecast value at time t+1. Function f is obtained through LSTM model training. The input includes data on historical power load, indoor and outdoor temperature, outdoor relative humidity, and equipment operating power. The back propagation algorithm is used to optimize the network weights and reduce the mean square error.

[0037] Historical power load data: The historical data obtained from the building's power meter can reflect the building's past power consumption patterns. hist (t) represents the load usage of the building at the past time t, indicating the fluctuation of the building's power demand in different time periods;

[0038] Indoor temperature data: The operation of the central air-conditioning system is closely related to the indoor temperature. The real-time temperature data of each area inside the building is collected through temperature sensors. The temperature fluctuation directly affects the cooling load demand of the air-conditioning system.

[0039] Outdoor temperature data: The ambient temperature outside the building affects the overall energy efficiency of the central air conditioning system. Higher outdoor temperatures will increase the cooling load, causing the central air conditioning system to require higher power to operate. Collecting outdoor temperature data can help the system dynamically adjust its prediction and operation strategies.

[0040] Outdoor relative humidity data: Humidity affects the cooling demand of the air. Higher humidity will increase the dehumidification load of the air conditioning system. Therefore, it is necessary to collect and record outdoor relative humidity data in real time, especially in humid climate conditions. Humidity is an important factor affecting the cooling load.

[0041] Real-time equipment operating power data: The operating power of key equipment in the air-conditioning system is an indicator that directly reflects the current working status of the system. The operating parameters of the refrigeration unit's workload and the power consumption of the water pump can reflect the current energy consumption level of the system.

[0042] LSTM processes dependencies over a longer time frame through internal memory units and gates. In building load forecasting, historical load data is time series data. LSTM can effectively learn the long-term and short-term change characteristics of these data. The input of the model is the historical data sequence, including the historical power load data L hist (t), indoor and outdoor temperature data T in (t) and T out (t), outdoor relative humidity RH out (t), and the equipment operating power P of the building equipment equip (t), the output is the cooling load forecast value L at the future time t+1 pred (t+1);

[0043] In order to enable the model to accurately predict future building load demand, the LSTM model needs to be trained. The model measures the difference between the predicted results and the actual results through the MSE loss function, and adjusts the network weights through the back-propagation algorithm to minimize the loss. The Adam optimizer is widely used in LSTM training. Combined with momentum and adaptive learning rate adjustment, it can accelerate convergence and prevent the model from falling into local optimality. Seasonal fluctuations and long-term trends in building load data will affect future load changes. Therefore, LSTM predicts future loads based on the dynamic changes of historical data through its gating;

[0044] Step 2: Use the predicted load L output by the building load prediction model in step 1 pred(t+1), combined with the time-of-use electricity price P(t) information, formulate the hourly cooling capacity Q cool (t) allocation strategy, which meets the load demand while reducing energy costs, Optimizing the objective function minimizes energy cost;

[0045] Among them, P(t) represents the time-of-use electricity price at time t, E AC (t) represents the energy consumption of the air conditioning system at time t, and the cooling capacity Q cool (t) According to the predicted load L pred (t+1) is dynamically adjusted, and the system energy consumption E AC (t)Through Calculate, where COP(t) is the energy efficiency coefficient of the refrigeration unit;

[0046] Step 3: By building energy efficiency and function models of key equipment in the central air-conditioning system, such as refrigeration units, water pumps and cooling towers, the energy efficiency performance of the equipment under different working conditions is evaluated. The energy efficiency model deduces the energy efficiency coefficient COP(t) of the equipment according to the equipment operation curve and environmental parameters. The calculation formula is: COP(t) = g(T out (t), L pred (t), P equip (t));

[0047] Where g is the external temperature T out (t), predicted load L pred (t) and equipment power to calculate the energy efficiency coefficient COP(t);

[0048] Step 4: Optimize the hourly operating parameters P of the central air-conditioning system through genetic algorithm equip (t), including the load rate of the refrigeration unit, the flow rate of the water pump and the speed of the cooling tower fan, the optimization process is achieved through the following steps:

[0049] The initial population of equipment operating parameters is randomly generated to represent the operating status of the equipment under different working conditions. The fitness value of each population is calculated. The fitness function is the negative value of the total energy consumption:

[0050] F itness (t)=-(E AC (t)+E pump (t)+E fan (t))

[0051] Among them, E AC (t), E pump (t), E fan(t) are the energy consumption of central air conditioner, pump and fan in time period t, respectively. Then, the roulette wheel selection algorithm is used to select individuals with high fitness for crossover, and new individuals are introduced by mutation operation to increase diversity. After multiple iterations, the population is updated and the optimal equipment operation parameters that minimize energy consumption are finally obtained.

[0052] Step 5: Based on load forecast L pred (t+1) and equipment operating power P equip (t), for cooling capacity Q cool (t) and equipment operating power P equip (t) Make dynamic adjustments to cope with fluctuations in building load demand, thereby reducing energy costs during peak periods while ensuring system operating efficiency.

[0053] In this embodiment, specifically, historical power load data is collected through multiple data sources, including the building's power meter data, the historical energy consumption records of the central air-conditioning system, and the real-time load monitoring data of the building floors, and the system automatically updates the historical power load data every 15 minutes. The weighted average method is used to fuse the historical data from multiple data sources, and the K-means clustering algorithm is used to group similar historical load patterns;

[0054] Historical electricity load data is the basis for predicting future load demand of buildings. Through multiple data sources (electricity meters of buildings, historical energy consumption records of central air-conditioning systems, and real-time load monitoring data of building floors), the load information of buildings is collected in real time. The information provided by each data source may be different. Therefore, the weighted average method is used to fuse the historical load data from different data sources. The weighted average method allocates the importance of data according to the weight of each data source to ensure representative data in the load forecasting process. The data is automatically updated every 15 minutes to ensure the timeliness and accuracy of the data. In addition, the K-means clustering algorithm is introduced to process the grouping of historical load patterns. K-means clustering is an unsupervised machine learning algorithm that divides data into different groups (i.e., "clusters") so that the data in each group are as similar as possible in some characteristics, while the data between different groups are quite different. Here, the K-means clustering algorithm is used to cluster the historical load data and group similar historical load patterns, which helps to identify the changing trend and periodic fluctuation of building load demand, thereby improving the prediction accuracy of the model for future loads.

[0055] In this embodiment, specifically, during the training process of the LSTM-based building load prediction model, the learning rate is adjusted according to the rate of change of the loss function during the training process. When the loss function no longer decreases significantly in multiple training rounds, the model will dynamically reduce the learning rate to ensure that the LSTM model continues to converge in the face of complex load changes. The error metric in the training process uses the mean square error MSE indicator, and the optimization process of the loss function is completed by the Adam optimizer;

[0056] LSTM (Long Short-Term Memory) is a deep learning model specifically designed to process time series data and is suitable for building load forecasting because it can capture the time dependency of loads. This dependent claim relates to how to optimize the training process of the LSTM model to ensure that it can efficiently and accurately predict in complex load forecasting scenarios. During the training process of the LSTM model, the learning rate is a key parameter that determines the step size of the model's parameter update in each iteration. If the learning rate is too high, the model may oscillate or even diverge during training; if the learning rate is too low, the model will converge too slowly and may fall into a local optimal solution.

[0057] The error metric uses MSE (mean square error), which is the average of the squared difference between the predicted value and the actual value. The MSE indicator can provide information on the accuracy of the model prediction. By combining the advantages of the MSE indicator and the Adam optimizer, this system can ensure that the LSTM model achieves efficient and accurate convergence under complex load conditions, while improving the robustness and generalization ability of the model.

[0058] The central air-conditioning system of a building is usually one of the largest energy consumers in the building. The time-of-use electricity price can flexibly adjust the energy consumption strategy of the central air-conditioning according to the fluctuation of electricity prices, thereby reducing the operating costs of the building. By obtaining the electricity price information of the electricity market in real time and using the time series prediction model to predict the electricity price fluctuations in future time periods, the cooling capacity allocation strategy is optimized. Then, the network interface is connected to the electricity market platform in real time to automatically obtain the time-of-use electricity price fluctuations in the current and future time periods, and compare these real-time electricity price information with the historical electricity price data of the building. In order to optimize the energy consumption allocation of the central air-conditioning system, the ARIMA time series prediction model is used to predict future electricity price fluctuations. The ARIMA model is a commonly used time series prediction model. By analyzing the past electricity price fluctuation trend, the future electricity price changes are predicted. By predicting the change trend of electricity prices, the cooling capacity allocation strategy of the air conditioner is adjusted in advance. When the electricity price is low, the cooling capacity reserve is appropriately increased to meet the building's cooling demand in advance; when the electricity price is high, the cooling capacity is reduced to minimize the electricity cost during peak hours.

[0059] In this embodiment, the source of the time-of-use electricity price information is the real-time electricity price information of the electricity market. The network interface is connected to the electricity market platform in real time to obtain the time-of-use electricity price fluctuations of the day and future periods. The time-of-use electricity price information is compared with the historical electricity price data of the building, and the electricity price fluctuations of future periods are predicted by using the time series prediction ARIMA model, and the cooling capacity Q is adjusted in advance. cool (t) allocation strategy;

[0060] In a building, the cooling load demand in different areas is usually different, which is affected by multiple factors such as population density, activity type and temperature comfort requirements. Therefore, the cooling capacity allocation needs to consider not only the overall energy consumption optimization, but also ensure that the cooling load demand of high-demand areas is met first in different time periods. This embodiment proposes a priority-based cooling capacity allocation strategy. According to the temperature comfort requirements and population density of different areas of the building, the cooling capacity is dynamically adjusted. During periods with lower electricity prices, more cooling capacity is allocated to high-priority areas to ensure the temperature comfort of these areas. When the electricity price is at a peak period, the cooling load allocation of these high-priority areas is reduced by controlling the cooling load distribution, only the minimum necessary cooling demand is guaranteed, and the overall energy consumption is reduced. The cooling capacity of low-priority areas is dynamically adjusted according to demand during the whole process to ensure that the cooling load of these areas is reduced as much as possible without affecting the overall comfort.

[0061] In this embodiment, specifically, the cooling capacity allocation strategy allocates priorities according to the temperature comfort requirements and occupant density of different areas in the building. Areas with higher priorities receive more cooling capacity allocation when the electricity price is lower, and prioritize minimum necessary cooling demand when the electricity price is higher.

[0062] In a building, the cooling load demand in different areas is usually different, which is affected by many factors such as population density, activity type and temperature comfort requirements. Therefore, the cooling capacity allocation not only requires overall energy consumption optimization, but also ensures that the cooling load demand of high-demand areas is met first in different time periods. This embodiment dynamically adjusts the cooling capacity according to the temperature comfort requirements and population density of different areas of the building. During periods when electricity prices are low, more cooling capacity is allocated to high-priority areas to ensure the temperature comfort of these areas. When electricity prices are at peak times, the cooling load allocation of these high-priority areas is reduced by controlling to ensure only the minimum necessary cooling demand and reduce overall energy consumption. The cooling capacity of low-priority areas is dynamically adjusted according to demand throughout the process to ensure that these areas reduce the cooling load as much as possible without affecting the overall comfort.

[0063] In this embodiment, specifically, during the establishment of the energy efficiency and function model of the central air-conditioning system, the historical operation data of the refrigeration unit, water pump and cooling tower are regularly collected to generate an equipment aging model, and the actual energy efficiency COP(t) of the equipment is dynamically adjusted in combination with the service life, routine maintenance frequency and historical failure rate of the equipment;

[0064] As the operating time of the refrigeration units, water pumps and cooling towers in the central air-conditioning system increases, the efficiency of the equipment will gradually decrease. Therefore, in order to ensure that the energy efficiency model can accurately reflect the current status of the equipment, it is necessary to establish dynamic adjustments based on the historical operating data, service life, maintenance frequency and failure rate parameters of the equipment. By regularly collecting the power consumption, operating time and failure frequency historical operating data of the equipment, the aging of the equipment is evaluated. The aging status of the equipment will affect the accuracy of its energy efficiency coefficient COP(t). Therefore, in the aging model, the actual energy efficiency parameters of the equipment will be dynamically adjusted with the changes in time and maintenance status. When the equipment has been running for a long time or is improperly maintained, the aging model will automatically reduce the energy efficiency coefficient of the equipment based on historical data, reflecting the actual performance decline of the equipment.

[0065] In this embodiment, specifically, the energy efficiency models of the refrigeration unit, water pump and cooling tower equipment are corrected by historical operation data and equipment performance curves provided by equipment manufacturers, and the load factor of the refrigeration unit and the parameters of the flow curve of the water pump in the energy efficiency model are corrected by real-time monitoring of the inlet and outlet water temperature difference of the equipment and the cooling air temperature environmental parameters, and comparing the real-time data with the equipment performance curve;

[0066] By real-time monitoring of the inlet and outlet water temperature difference of the refrigeration unit, the flow rate of the water pump and the air temperature parameters of the cooling tower, the current operating status of the equipment is obtained, and these data are compared with the performance curve provided by the equipment manufacturer. The manufacturer's performance curve usually provides the energy efficiency performance of the equipment under different working conditions. However, the actual operating environment is usually different from the ideal conditions in the performance curve. Therefore, according to the difference between the real-time monitored equipment status and the performance curve, the load factor of the equipment and the key parameters of the flow curve are dynamically adjusted to ensure that the energy efficiency model can accurately reflect the actual energy efficiency performance of the equipment.

[0067] In this embodiment, specifically, when the genetic algorithm is used to optimize the hourly operating parameters of the central air-conditioning system, an individual selection strategy based on the tournament selection method is adopted. In each generation of the population, several individuals are randomly selected from the current population for comparison, and individuals with higher fitness are subjected to crossover and mutation operations to generate new individuals. The crossover operation in the genetic algorithm uses multi-point crossover, and the crossover operation performs gene exchange at multiple gene points. During the crossover process, each new individual generated is subjected to a constraint condition check to ensure that the equipment operating power P equip (t) always remain within the safe operating range of the equipment;

[0068] Genetic algorithm is an optimization algorithm based on biological evolution, which is suitable for solving complex multi-objective optimization problems. In this embodiment, the genetic algorithm is used to optimize the hourly operating parameters of the central air-conditioning system to minimize energy consumption and ensure efficient operation of the system. The tournament selection method is a strategy for individual selection in the genetic algorithm. By randomly selecting several individuals for comparison, individuals with higher fitness are selected for subsequent crossover and mutation operations.

[0069] In each generation of the genetic algorithm, the system first randomly selects a part of individuals (i.e., different combinations of equipment operating parameters) for tournament comparison. The tournament comparison evaluates the fitness value of each individual through the fitness function. Individuals with higher fitness have a greater chance of being selected to participate in crossover and mutation in subsequent breeding operations, thereby generating new individuals. This can ensure that the genetic algorithm is always within the global search range and avoid the problem of local optimal solutions.

[0070] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0071] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A central air conditioning system optimization control method for building load prediction, characterized in that: The control steps include: Step 1: Collect historical electricity load data of the building , indoor and outdoor temperature data and , outdoor relative humidity , and equipment operating power of building equipment , which means using the above data to build a building load forecasting model based on the long short-term memory network LSTM, and the prediction formula is: , to predict the hourly load demand in future periods; in, Indicates time The future load forecast value of The data is obtained through LSTM model training, and the input includes historical power load, indoor and outdoor temperature, outdoor relative humidity and equipment operating power data. The back propagation algorithm is used to optimize the network weights and reduce the mean square error; Step 2: Use the predicted load output by the building load prediction model in step 1 , combined with time-of-use electricity prices Information, formulate hourly cooling capacity The allocation strategy can meet the load demand while reducing energy costs. Optimizing the objective function minimizes energy cost; in, Indicates time Time-of-use electricity prices, Indicates the air conditioning system at time Energy consumption, cooling capacity Based on the predicted load Dynamically adjust the system energy consumption pass Calculate, where is the energy efficiency coefficient of the refrigeration unit; Step 3: Evaluate the energy efficiency performance of the equipment under different working conditions by building energy efficiency and function models of key equipment such as refrigeration units, water pumps and cooling towers. The energy efficiency model deduces the energy efficiency coefficient of the equipment based on the equipment operation curve and environmental parameters. , and its calculation formula is: ; in, According to the outside temperature , Forecast Load Calculate the energy efficiency coefficient based on the equipment power Function of Step 4: Optimize the hourly operating parameters of the central air-conditioning system through genetic algorithms , including the load rate of the refrigeration unit, the flow rate of the water pump and the speed of the cooling tower fan. The optimization process is achieved through the following steps: The initial population of equipment operating parameters is randomly generated to represent the operating status of the equipment under different working conditions. The fitness value of each population is calculated. The fitness function is the negative value of the total energy consumption: in, Central air conditioning, pumps and fans in the time period The energy consumption is then determined. The roulette wheel selection algorithm is then used to select individuals with high fitness for crossover, and mutation operations are used to introduce new individuals to increase diversity. After multiple iterations of updating the population, the optimal equipment operating parameters that minimize energy consumption are finally obtained. Step 5: Based on load forecast and equipment operating power , for cooling capacity and equipment operating power Make dynamic adjustments to cope with fluctuations in building load demand, ensuring system operating efficiency while reducing peak energy costs.

2. The central air conditioning system optimization control method for building load prediction according to claim 1 is characterized in that: The historical electricity load data is collected through multiple data sources, including the building's electricity meter data, the historical energy consumption records of the central air-conditioning system, and the real-time load monitoring data of the building floors, and the system automatically updates the historical electricity load data every 15 minutes; the weighted average method is used to fuse the historical data from multiple data sources, and the K-means clustering algorithm is used to group similar historical load patterns.

3. The central air conditioning system optimization control method for building load prediction according to claim 1 is characterized in that: During the training process, the building load prediction model based on the long short-term memory network LSTM adjusts the learning rate according to the change rate of the loss function during the training process. When the loss function no longer decreases significantly in multiple training rounds, the model will dynamically reduce the learning rate to ensure that the LSTM model continues to converge when facing complex load changes. The error measurement in the training process uses the mean square error MSE indicator, and the optimization process of the loss function is completed through the Adam optimizer.

4. The central air conditioning system optimization control method for building load prediction according to claim 1 is characterized in that: The source of the time-of-use electricity price information is the real-time electricity price information of the electricity market. Through the network interface, it is connected to the electricity market platform in real time to obtain the time-of-use electricity price fluctuations of the day and future periods, compare the time-of-use electricity price information with the historical electricity price data of the building, and use the time series prediction ARIMA model to predict the electricity price fluctuations in future periods and adjust the cooling capacity in advance. allocation strategy.

5. The central air conditioning system optimization control method for building load prediction according to claim 4 is characterized in that: The cooling capacity The allocation strategy assigns priorities according to the temperature comfort requirements and occupancy density of different areas in the building. Higher-priority areas receive more cooling allocation when electricity prices are lower, while when electricity prices are higher, priority is given to ensuring the minimum necessary cooling demand.

6. The central air conditioning system optimization control method for building load prediction according to claim 5 is characterized in that: In the process of establishing the energy efficiency and function model of the central air-conditioning system, the historical operation data of the refrigeration unit, water pump and cooling tower are regularly collected to generate an equipment aging model, and the actual energy efficiency of the equipment is estimated based on the service life, daily maintenance frequency and historical failure rate of the equipment. Dynamic adjustment.

7. The central air conditioning system optimization control method for building load prediction according to claim 6 is characterized in that: The energy efficiency model of the refrigeration unit, water pump and cooling tower equipment is calibrated through historical operating data and equipment performance curves provided by equipment manufacturers. The load factor of the refrigeration unit and the parameters of the flow curve of the water pump in the energy efficiency model are calibrated by real-time monitoring of the inlet and outlet water temperature difference of the equipment and the cooling air temperature environmental parameters, and comparing the real-time data with the equipment performance curve.

8. The central air conditioning system optimization control method for building load prediction according to claim 1 is characterized in that: When the genetic algorithm is used to optimize the hourly operating parameters of the central air-conditioning system, an individual selection strategy based on the tournament selection method is adopted. In each generation of the population, several individuals are randomly selected from the current population for comparison, and individuals with higher fitness are subjected to crossover and mutation operations to generate new individuals.

9. The central air conditioning system optimization control method for building load prediction according to claim 1, characterized in that: The crossover operation in the genetic algorithm uses multi-point crossover, which exchanges genes at multiple gene points. During the crossover process, each new individual generated is checked for constraints to ensure that the equipment is running efficiently. Always stay within the safe operating range of the equipment.

Images