Hierarchical optimization method for energy consumption, computer device, and storage medium

Abstract

This application relates to a hierarchical optimization method for energy consumption, a computer device, and a storage medium, and relates to the field of energy. The method includes the steps of: acquiring target operating data related to the optimization of energy consumption; performing data hierarchical structuring on the target operating data, acquiring datasets corresponding to each optimization hierarchical level, and associating datasets of different optimization hierarchical levels with different optimization indicators; determining the optimization starting level based on the equipment operating mechanism and the current state indicated by the target operating data; and using the optimization starting level as the optimization starting point, performing hierarchical optimization calculations based on the optimization model of each optimization hierarchical level and the datasets corresponding to each optimization hierarchical level, acquiring a target energy consumption optimization policy, and optimizing the energy consumption of the heating equipment based on the target energy consumption optimization policy.

Description

[Technical Field]

[0001] (Cross-reference of related applications) This invention claims priority to a Chinese patent application filed with the China National Intellectual Property Administration on July 5, 2023, with application number 202310820243.9, and titled "Hierarchical Optimization Method for Energy Consumption, Computer Apparatus and Storage Medium," the entirety of which is incorporated into this invention by reference.

[0002] The embodiments of this application relate to the field of energy, and more particularly to a hierarchical optimization method for energy consumption, a computer device, and a storage medium. [Background technology]

[0003] Reducing carbon dioxide emissions is an important means of protecting the environment, and in scenarios where heating systems are in place (e.g., data centers), energy consumption can be reduced and carbon dioxide emissions reduced by optimizing the control policies for these heating systems.

[0004] In related technologies, methods for optimizing heating equipment control policies primarily involve directly constructing energy consumption prediction models between operational data related to energy consumption optimization and power usage efficiency using machine learning or deep learning. The operational data related to energy consumption optimization includes outdoor environment data, heating equipment, and keypoint data for IT equipment. When the predicted power usage efficiency is lower than the original power usage efficiency, a combination of input parameters for the model is considered an effective method for optimizing energy consumption.

[0005] However, the above optimization method is a purely data-driven model, and while it can theoretically achieve good energy consumption optimization effects, in practice, it may lead to problems such as frequent starting and stopping of equipment due to frequent adjustments of control parameters, or inconsistencies with actual operation and maintenance logic due to large differences in parameters, thus making the energy consumption optimization method impractical. [Overview of the project] [Problems that the invention aims to solve]

[0006] Embodiments of the present application provide a hierarchical energy consumption optimization method, a computer device, and a storage medium that can improve optimization efficiency and enhance the practicality of energy consumption optimization policies. The technical means are as follows: [Means for solving the problem]

[0007] According to one embodiment, a method for hierarchical optimization of energy consumption is provided, which includes the steps of: acquiring target operating data related to the optimization of energy consumption; performing data hierarchicalization on the target operating data to acquire a dataset corresponding to each optimization hierarchy and associating the datasets of different optimization hierarchies with different optimization indicators; determining the optimization start hierarchy based on the equipment operating mechanism and the current state indicated by the target operating data; and using the optimization start hierarchy as the optimization starting point, performing hierarchical optimization calculations based on the optimization model of each optimization hierarchy and the dataset corresponding to each optimization hierarchy, acquiring a target energy consumption optimization policy, and optimizing the energy consumption of the heating equipment based on the target energy consumption optimization policy.

[0008] In another embodiment, the present invention provides a computer device comprising a processor and memory storing at least one computer program, wherein the hierarchical optimization method for energy consumption is realized when the at least one computer program is loaded and executed by the processor.

[0009] In another embodiment, a readable storage medium is provided, which stores at least one computer program, and the hierarchical optimization method for energy consumption is realized when the computer program is loaded and executed by a processor.

[0010] In another embodiment, a computer program product is provided, which includes at least one computer program, which is loaded and executed by a processor to realize the hierarchical optimization method of energy consumption provided in the various selectable embodiments described above. [Brief explanation of the drawing]

[0011] Herein, the drawings are incorporated into the specification and constitute part of this specification, illustrating embodiments consistent with the present application and are used together with the specification to illustrate the principles of the present application. [Figure 1] A flowchart of a hierarchical optimization method for energy consumption according to one exemplary embodiment of the present invention is shown. [Figure 2] A flowchart of a hierarchical optimization method for energy consumption according to one exemplary embodiment of the present invention is shown. [Figure 3] A flowchart of a hierarchical optimization method for energy consumption according to one exemplary embodiment of the present invention is shown. [Figure 4] A schematic diagram of a hierarchical energy consumption optimization system according to one exemplary embodiment of the present invention is shown. [Figure 5] A structural block diagram of a computer device according to one exemplary embodiment of the present invention is shown. [Modes for carrying out the invention]

[0012] The following describes exemplary embodiments in detail, illustrated in the drawings. Where the following description relates to the drawings, unless otherwise specified, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the present application, which are described in detail in the appended claims.

[0013] In the scenario where heating equipment is installed, the purpose of energy conservation and reduction of energy consumption can be achieved by optimizing the energy consumption of the heating equipment. Exemplarily, when performing energy conservation and reduction of energy consumption, it can be achieved by optimizing the energy consumption of the heating equipment. At the current stage, there are the following problems in formulating the energy consumption optimization policy of the heating equipment.

[0014] 1) The relationship between the control parameters of the heating equipment is not considered, and all control parameters are treated equally (for example, the operating speed of the equipment and the internal control parameters of the equipment are adjusted equally each time optimization is performed). The changes in the output control parameters (especially the differences in the operating speed of the equipment) cause frequent startup and shutdown of the equipment, which does not conform to the actual operation and maintenance of the equipment.

[0015] 2) The combination of types of heating equipment (i.e., the operating mode) changes according to the climate, and the corresponding internal control parameters of the equipment are also different, which is not considered. Therefore, the difference between the internal control parameters given by the conventional model and the current actual operation and maintenance parameter setting logic of the conventional equipment is large, the reliability is low, and it is difficult to accept.

[0016] 3) In the modeling process, safety and effectiveness are not fully considered in accordance with the actual operation and maintenance. Therefore, the optimized control policy output by the model does not have high reliability during subsequent distribution.

[0017] 4) The large difference between the actually collected messy data and the simulated ideal data is not considered. There are many problems in the actually collected data, such as many data omissions, inconsistencies between equipment status indicators and equipment frequency feedback, and abnormal data collection. It is necessary to perform hierarchical processing and accuracy verification by combining the equipment operation mechanism. If effective data hierarchical processing is not performed, the accuracy of the model will decrease and subsequent applications will become invalid.

[0018] To address the above-mentioned problems, the embodiments of the present application provide a hierarchical energy consumption optimization method that solves all or part of the above-mentioned problems, optimizes energy consumption, and improves the practicality of energy consumption optimization policies. The hierarchical energy consumption optimization method of the present application will be described below with reference to several embodiments.

[0019] Figure 1 shows a flowchart of a hierarchical optimization method for energy consumption according to one exemplary embodiment of the present invention, the method which may be performed by a computer device, the computer device which may be implemented as a heating equipment management device, and as shown in Figure 1, the hierarchical optimization method for energy consumption may include the following steps 110 to 140.

[0020] In step 110, target operating data related to the optimization of energy consumption is obtained.

[0021] In the embodiments of this application, the target operating data related to the optimization of energy consumption may include temperature and humidity at each stage indoors and outdoors, energy consumption of various heating equipment, status information of key assemblies of heating equipment, and control parameters. The target operating data may be obtained by a computer device monitoring each heating equipment.

[0022] In step 120, the target driving data is hierarchically structured, a dataset corresponding to each optimization hierarchy is obtained, and datasets from different optimization hierarchies are associated with different optimization metrics.

[0023] When formulating an energy consumption optimization policy based on target operating data, the target operating data is first stratified based on optimization indicators corresponding to each optimization hierarchy to obtain a dataset corresponding to each optimization hierarchy. Different optimization hierarchies have different optimization indicators, which is reflected in the number and types of optimization indicators associated with different optimization hierarchies. Some of the optimization indicators associated with different optimization hierarchies may be the same, or they may be entirely different. For example, one optimization hierarchy may be associated with optimization indicator A, and another optimization hierarchy may be associated with optimization indicators A and B. The number of optimization hierarchies and the setting of optimization indicators associated with each optimization hierarchy may be set by the stakeholders according to their actual needs, and this application is not limited thereto.

[0024] In step 130, the optimization start level is determined based on the current state indicated by the equipment operating mechanism and target operating data.

[0025] In the embodiments of the present invention, there may be hierarchical relationships between each optimization level; that is, in the process of formulating an energy consumption optimization policy, the computer device can perform optimization calculations hierarchically based on the hierarchical relationships between each optimization level. Therefore, in order to avoid meaningless optimization calculation processes, the computer device can first determine the optimization starting level in the current state based on the current state indicated by the equipment operating mechanism and target operating data, and then determine the optimization starting point in each optimization level.

[0026] In step 140, the optimization starting layer is used as the optimization starting point, and hierarchical optimization calculations are performed based on the optimization model of each optimization layer and the dataset corresponding to each optimization layer to obtain a target energy consumption optimization policy, and the energy consumption of the heating equipment is optimized based on the target energy consumption optimization policy.

[0027] For example, if four optimization hierarchies with a hierarchical relationship are arranged, and the optimization starting hierarchy determined based on the current state indicated by the equipment operating mechanism and target operating data is the second optimization hierarchy, then when formulating an energy consumption optimization policy, the first optimization hierarchy can be skipped and hierarchical optimization calculations and verifications can be performed from the second optimization hierarchy until the target energy consumption optimization policy is obtained or all optimization hierarchies have been traversed. Hierarchical optimization enables the formulation of energy consumption optimization policies based on different optimization indicators, and by determining the optimization starting hierarchy, unnecessary optimization calculations and verification processes can be avoided, improving the efficiency of formulating energy consumption optimization policies and enhancing the applicability of the energy consumption optimization policy to the current state.

[0028] In each optimization tier, when performing optimization calculations, the optimization model of the optimization tier is used to perform optimization calculations on the dataset of the optimization tier to obtain the optimization result of the optimization tier, i.e., the energy consumption optimization policy of the optimization tier. In one possible embodiment, the computer device may directly determine the energy consumption optimization policy obtained in the optimization tier as the target energy consumption optimization policy, or in another possible embodiment, the computer device may determine the energy consumption optimization policy as the target energy consumption optimization policy if the energy consumption optimization policy satisfies specified conditions.

[0029] As described above, the hierarchical energy consumption optimization method according to the embodiment of the present invention acquires target operating data related to energy consumption optimization, performs data hierarchical structuring on the target operating data to obtain datasets for each optimization hierarchical level corresponding to different optimization indicators, determines the optimization starting hierarchical level based on the equipment operating mechanism and the current state, uses the optimization starting hierarchical level in each optimization hierarchical level as the optimization starting point, performs hierarchical optimization calculations based on the optimization model of each optimization hierarchical level and the dataset of each optimization hierarchical level to obtain a target energy consumption optimization policy that matches the current actual needs, and optimizes the energy consumption of the heating equipment based on the target energy consumption optimization policy. By the above method, an optimization starting hierarchical level that is adapted to the equipment operating mechanism and the current state is determined, and hierarchical optimization is performed using the optimization starting hierarchical level as the hierarchical optimization starting point, thereby formulating an energy consumption optimization policy, avoiding unnecessary optimization processes, improving optimization efficiency, and improving the practicality of the energy consumption optimization policy because the acquired target energy consumption optimization policy is adapted to the current state and the equipment operating mechanism.

[0030] In one possible embodiment, if the energy consumption optimization policy output in the optimization hierarchy passes safety and effectiveness verification, the energy consumption optimization policy is determined to be the target energy consumption optimization policy. In embodiments of the present application, a computer device can perform safety and effectiveness verification of the energy consumption optimization policy by a safety and effectiveness verification model. In this case, Figure 2 shows a flowchart of a hierarchical energy consumption optimization method according to one exemplary embodiment of the present application, the method which may be performed by a computer device which may be implemented as a heating equipment management device, and the hierarchical energy consumption optimization method which may include the following steps 201 to 208 as shown in Figure 2.

[0031] In step 201, target operating data related to the optimization of energy consumption is obtained.

[0032] The computer device can acquire actual operating data obtained from an actual data center as target operating data, and the actual data center may include an IT center and a power monitoring system (Supervision System, SS). In the embodiment of the present invention, the target operating data may include changes in indoor and outdoor temperature and humidity and information technology (IT) load, and the target operating data is used to indicate the current state.

[0033] In step 202, the target driving data is hierarchically structured, a dataset corresponding to each optimization hierarchy is obtained, and datasets from different optimization hierarchies are associated with different optimization metrics.

[0034] The computer device can implement a process by which a data hierarchical processing module performs data hierarchical processing on target operating data and obtains a dataset corresponding to each optimization hierarchical level. This process may include steps of performing data hierarchical cleaning, data hierarchical validation, indicator hierarchical fusion, indicator hierarchical extraction, and indicator hierarchical classification on the target operating data and obtaining a dataset to be applied to a control policy hierarchical optimization model. The control policy hierarchical optimization model includes optimization models for each optimization hierarchical level, and the dataset applied to the control policy hierarchical optimization model includes a subset of data applied to the optimization models for each optimization hierarchical level.

[0035] In the data tiering process, data tiering cleaning cleans up redundant data in target operating data, such as checking for data consistency and handling invalid and missing values.

[0036] Hierarchical data validation is used to verify the accuracy of each piece of data. Based on the service logic, it can verify whether there are any anomalies in the data collection of each piece of equipment. If an anomaly is determined, it can determine whether the anomaly is due to equipment failure or a data collection error, statistically analyze the size of the anomaly data, and then take corresponding data correction or data removal measures based on a pre-specified data accuracy extraction logic. For example, if the service logic on which hierarchical data validation is based is the service logic of a water cooling system, it is necessary to determine whether the cooling pump, primary refrigeration pump, water cooler, or plate heat exchanger in the chiller unit meet the conditions for simultaneous opening and closing, and then determine whether there are any anomalies in the data collection of each piece of equipment corresponding to the water cooling system. For example, if the equipment status bit for a piece of equipment indicates that the equipment is off, but there is a value in the equipment's frequency or current percentage, indicating that it is on, then an anomaly is determined.

[0037] Hierarchical fusion of indicators is used to perform data statistics on related indicators and generate new indicators. For example, a computer device may determine the actual operating status of equipment (i.e., the new indicator) using equipment status indicators, control indicators, and control feedback indicators. Hierarchical fusion of indicators may also be achieved by performing statistics on the number of operations for the same type of equipment, generating average temperature and humidity values, generating average current percentage values, generating average frequency feedback values, etc.

[0038] Hierarchical metric extraction is used to extract the necessary sets of metrics for different models and construct corresponding data subsets. For example, the data subset Data_E applied to the energy consumption effectiveness verification model includes metrics related to energy consumption, such as PUE (Power Usage Effectiveness) or CLF (Cooling Load Factor), where PUE is the ratio of all energy consumed to the energy used by the IT load, and CLF is the ratio of the power consumption of the cooling equipment to the power consumption of the IT equipment. The data subset Data_C applied to the cooling capacity safety verification model includes metrics related to cooling capacity. The data subset Data_S applied to the state safety verification includes metrics related to temperature and pressure states. The data subset for each optimization hierarchy required in the control policy hierarchical optimization model includes metrics related to the current optimization hierarchy.

[0039] Hierarchical classification of indicators is used to classify each indicator according to its type. For example, optimization indicators can be classified into different types based on classification criteria such as the type of indicator or the function of the indicator. For instance, if the classification criterion is the type of indicator, each indicator can be classified into environmental indicators, equipment operating status indicators, equipment control feedback indicators, equipment control parameter setting indicators, various energy consumption indicators, etc. If the classification criterion is the function of the indicator, each indicator can be classified into safety verification-related indicators, effectiveness verification-related indicators, policy hierarchical optimization-related indicators, etc. Hierarchical classification of indicators facilitates the iteration and modification of data hierarchical processing processes.

[0040] In one embodiment, before performing data tiering on target operating data and obtaining datasets corresponding to each optimization tier, the method further includes the step of performing an optimization pre-check on the target operating data using optimization pre-check indicators and obtaining optimization pre-check results. The optimization pre-check indicators include at least one of checking whether the indicators required for each model are missing, and checking whether service relationships exist between the indicators. If the optimization pre-check results indicate that the target operating data has passed the pre-check, the target operating data is tiered and datasets corresponding to each optimization tier are obtained.

[0041] In other words, when entering the online inference process for hierarchical optimization of energy consumption, first, a pre-optimization check is performed on the target operating data, and only after the target operating data passes the pre-optimization check is the subsequent processing process carried out. Checking whether the necessary indicators for each model are missing may involve checking whether there are large gaps in indoor and outdoor ambient temperatures, equipment operating conditions, equipment control feedback, equipment setting parameters, etc. Checking whether the service relationships between indicators are established may involve checking whether the state of each piece of equipment in the chiller unit satisfies simultaneous opening and closing, etc.

[0042] If the optimization pre-test results indicate that the target operating data failed the pre-test, it is determined that the heating system is not suitable for energy consumption optimization.

[0043] If the optimization pre-inspection results indicate that the target operating data failed the pre-inspection, it suggests a possible equipment malfunction or data collection error. In this case, it is determined that the heating equipment is not suitable for energy consumption optimization, and the system temporarily does not proceed to the subsequent process. In one embodiment, in this case, a prompt indicating "not suitable for optimization" may be issued via audio or visual information, for example, by emitting a prompt sound or displaying prompt information in the form of text or images on a display device.

[0044] In step 203, the optimization start level is determined based on the current state indicated by the equipment operating mechanism and target operating data.

[0045] As an example, consider a computer system with three hierarchical optimization levels: the first, second, and third optimization levels. When performing hierarchical optimization calculations, if the optimization starts at the first optimization level, the optimization is performed in order from the first to the third optimization level. The optimization metrics for the three optimization levels differ: the first optimization level refers to optimizing the internal control parameters of the heating system; the second optimization level refers to optimizing the number of operating units and internal control parameters of the heating system; and the third optimization level refers to optimizing the operating mode, number of operating units, and internal control parameters of the heating system. The operating mode reflects the combination method of the equipment; for example, in a water-cooling system, it can be divided into free mode, pre-cooling mode, cooling mode, etc. Based on this, an embodiment of the present invention provides a feasible method for determining an optimization start level based on climate change, the process of determining the optimization start level by a computer device can be realized by the steps of: determining a first optimization level as the optimization start level, which indicates that the internal control parameters of the heating equipment should be optimized when the outdoor temperature is within a preset outdoor temperature range, the indoor and outdoor temperature and humidity changes periodically, and the IT load changes periodically; determining a second optimization level as the optimization start level, which indicates that the number of operating units and internal control parameters of the heating equipment should be optimized when the outdoor temperature is within an outdoor temperature range, the indoor and outdoor temperature and humidity changes indicate a rapid change in indoor and outdoor temperature and humidity, or the IT load changes indicate a rapid change in IT load; and determining a third optimization level as the optimization start level, which indicates that the operating mode, number of operating units, and internal control parameters of the heating equipment should be optimized when the outdoor temperature is within a critical temperature range between different outdoor temperature ranges.

[0046] That is, the optimization index of the first optimization layer is the internal control parameter of the heating equipment, the optimization index of the second optimization layer is the operation number and internal control parameter of the heating equipment, and the optimization index of the third optimization layer is the operation mode, operation number and internal control parameter of the heating equipment.

[0047] When the outdoor temperature is within a preset outdoor temperature range, it indicates that there is no significant change in the climate, and the preset outdoor temperature range may be set according to seasons respectively. For example, the outdoor temperature range in winter may be T≤10, the outdoor temperature range in spring and autumn may be 10<T≤26, and the outdoor temperature range in summer may be T>26. It should be noted that the setting for the above outdoor temperature range is merely exemplary, and the outdoor temperature range may have different settings according to actual situations, and the present application is not limited thereto. When there is no significant change in the climate and the indoor and outdoor temperature and humidity and IT load change periodically, when performing energy consumption optimization, an optimization method of optimizing the internal control parameter of the heating equipment without adjusting the operation mode and operation number of the heating equipment is preferentially selected, that is, an optimization calculation process with the first optimization layer as the initial optimization layer is executed.

[0048] When the outdoor temperature is within the outdoor temperature range, i.e., there is no significant change in climate, and the changes in indoor and outdoor temperature and humidity indicate a rapid change in indoor and outdoor temperature and humidity, or the changes in IT load indicate a rapid change in IT load, when performing energy consumption optimization, the optimization method that optimizes the number of heating units and the internal control parameters of the equipment without adjusting the operating mode of the heating equipment is preferentially selected, i.e., an optimization calculation process is executed with the second optimization layer as the initial optimization layer. In one embodiment, the computer device can execute an optimization calculation process with the second optimization layer as the initial optimization layer if there is no significant change in climate, there is a rapid change in indoor and outdoor temperature and humidity, and this change exceeds a predetermined ratio threshold within a predetermined time interval, or there is a rapid change in IT load, and this change exceeds a predetermined ratio threshold within a certain time. For example, if the outdoor temperature is within the outdoor temperature range, there is a rapid change in IT load, and the change before and after within 4 hours exceeds 50%, it can be determined to execute an optimization calculation process with the second optimization layer as the initial optimization layer. The predetermined time interval and predetermined ratio threshold may both be set according to actual needs, and this application is not limited thereto.

[0049] When the outdoor temperature falls within the critical temperature range between different outdoor temperature ranges, it indicates that the climate is in a transitional period. For example, if the outdoor temperature T is approximately 10 or 26, corresponding to the outdoor temperature range set according to the season, it is determined that the climate is in a transitional period. In this case, an optimization method that optimizes the heating equipment's operating mode, the number of heating units in operation, and the internal control parameters of the equipment is preferentially selected, and an optimization calculation process is executed with the third optimization layer as the initial optimization layer.

[0050] Furthermore, the classification of the optimization hierarchies described above, and the setting of climate conditions when each optimization hierarchy is designated as the initial optimization hierarchy, are merely feasible classification and setting methods provided by the embodiments of this application. Parties concerned may set more or fewer optimization hierarchies and corresponding climate conditions according to their actual needs, and this application is not limited to this.

[0051] In step 204, if the system enters a target optimization hierarchy, which is one of the optimization hierarchies that uses the optimization start hierarchy as the starting point for optimization, the optimization model of the target optimization hierarchy is used to perform optimization calculations on the corresponding dataset and obtain a candidate energy consumption optimization policy.

[0052] In a hierarchical optimization process, the computer device starts the optimization process with the optimization start level as the optimization start point, and terminates the optimization process when the target energy consumption optimization policy has been obtained or each optimization level has been traversed, which is the optimization end point. In the process of performing a hierarchical optimization process with the determined optimization start level as the optimization start point, in order to further improve the practicality of the obtained energy consumption optimization policy, after obtaining the energy consumption optimization policy at each optimization level in the optimization process, a safety and effectiveness verification process is performed, and based on the verification results, it is decided whether or not to decide on it as the target energy consumption optimization policy.

[0053] In step 205, the safety and effectiveness of the candidate energy consumption optimization policy are verified using a safety and effectiveness verification model, and the verification results for the candidate energy consumption optimization policy are obtained.

[0054] In the embodiments of the present invention, the safety and effectiveness verification performed by the computer device on the energy consumption optimization policy may include energy consumption effectiveness verification, cooling capacity safety verification, and application state safety verification, and accordingly, the safety and effectiveness verification model may include an energy consumption effectiveness verification model, a cooling capacity safety verification model, and an application state safety verification model.

[0055] The steps of performing safety and effectiveness verification on an energy consumption optimization policy obtained using an optimization model by a safety and effectiveness verification model include: performing energy consumption effectiveness verification on the energy consumption optimization policy using an energy consumption effectiveness verification model and obtaining a first sub-verification result; performing cooling capacity safety verification on the energy consumption optimization policy using a cooling capacity safety verification model and obtaining a second sub-verification result; performing state safety verification on the energy consumption optimization policy using a state safety verification model and obtaining a third sub-verification result; and generating a verification result indicating whether the energy consumption optimization policy passed or failed verification based on the first sub-verification result, the second sub-verification result, and the third sub-verification result.

[0056] Each sub-verification result may be a predicted value of the corresponding model's verification metric, and whether the corresponding model has passed the verification is determined by determining whether the predicted value of the corresponding verification metric satisfies the verification conditions of the corresponding verification metric. For example, when an energy consumption efficiency verification model performs energy consumption efficiency verification on an energy consumption optimization policy, it means that data corresponding to the energy consumption efficiency verification in the energy consumption optimization policy is input into the energy consumption efficiency verification model, the first sub-verification result of the energy consumption efficiency verification model is obtained, i.e., a predicted value of the power usage efficiency in the energy consumption optimization policy, and if the first sub-verification result satisfies the power usage efficiency verification conditions, it is determined that the energy consumption efficiency verification has passed; if the first sub-verification result does not satisfy the power usage efficiency verification conditions, it is determined that the energy consumption efficiency verification has failed. For the verification process that indicates whether other sub-verification results have passed or failed, please refer to the explanation above, and the explanation will be omitted here.

[0057] In step 206, if the verification results of the candidate energy consumption optimization policy indicate that the candidate energy consumption optimization policy has passed the verification, the candidate energy consumption optimization policy is determined to be the target energy consumption optimization policy.

[0058] In one possible embodiment, if the first sub-verification result, the second sub-verification result, and the third sub-verification result all indicate that the verification has passed, the verification result is determined to indicate that the energy consumption optimization policy has passed verification. If any of the first sub-verification result, the second sub-verification result, and the third sub-verification result indicate that the verification has failed, the verification result is determined to indicate that the energy consumption optimization policy has failed verification. In one embodiment, the computer device may set a verification ratio threshold, and if the ratio of the number of sub-verification results indicating successful verification to the total number of sub-verification results is greater than the verification ratio threshold, the verification result is determined to indicate that the energy consumption optimization policy has passed verification; otherwise, the verification result is determined to indicate that the energy consumption optimization policy has failed verification. The value of the verification ratio threshold may be set according to actual needs, and the application is not limited thereto. As an example, if the results of the first sub-verification, the second sub-verification, and the third sub-verification all indicate that the verification has passed, then the verification result is determined to indicate that the energy consumption optimization policy has passed verification. In this case, the safety and effectiveness verification process for the above energy consumption optimization policy can be expressed as follows: OptimalControls=Model_x*(Data_new) (max(CurrentEnergy-Model_E(OptimalControls)) / CurrentEnergy)&(Model_C(OptimalControls)>=CurrentNeededCooling)&(MinSafeState<=Model_S(OptimalControls)<=MaxSafeState) OptimalControls represents the target energy consumption optimization policy, Model_x represents the xth optimization level, Data_new represents the target operating data, CurrentEnergy represents the current energy consumption, CurrentNeededCooling represents the current cooling capacity, MinSafeState represents the minimum state safety, and MaxSafeState represents the maximum safe state. Note that the setting of verification conditions for each of the above verification indicators is merely illustrative, and other verification conditions can be set according to actual needs, and this application is not limited thereto.

[0059] In the hierarchical optimization process, the energy consumption optimization policy of the current optimization hierarchical may fail safety and effectiveness verification. In this case, In step 207, if the target optimization hierarchy is not the highest-level optimization hierarchy among the various optimization hierarchies, and the validation results for the candidate energy consumption optimization policy indicate that the candidate energy consumption optimization policy has failed to validate, the optimization calculations and safety and effectiveness validations for the lower-level optimization hierarchies of the target optimization hierarchy are performed.

[0060] In step 208, if the target optimization hierarchy is the highest optimization hierarchy among the various optimization hierarchies, and the validation result of the candidate energy consumption optimization policy indicates that the candidate energy consumption optimization policy has failed to validate, then it is determined that the heating equipment is not suitable for energy consumption optimization.

[0061] In other words, if a higher optimization hierarchy exists above the target optimization hierarchy, and a target energy consumption optimization policy cannot be obtained based on the target optimization hierarchy, the optimization process of the lower optimization hierarchy is performed. For example, after performing optimization calculations and safety and effectiveness verification using the optimization model of the Nth optimization hierarchy, if the obtained energy consumption optimization policy passes the verification, it is determined and output as the target energy consumption optimization policy, and the hierarchical optimization process is terminated. If the obtained energy consumption optimization policy fails the verification, the optimization calculations and safety and effectiveness verification are continued using the optimization model of the N+1th optimization hierarchy until a target energy consumption optimization policy is obtained or all optimization hierarchys are traversed, where N is a positive integer. If a target energy consumption optimization policy is not obtained even after all optimization hierarchys have been traversed, it is determined that the current heating equipment is not suitable for energy consumption optimization, and a corresponding prompt is output.

[0062] Performing optimization in an optimization hierarchy means adjusting the parameters corresponding to each optimization metric in the dataset corresponding to that optimization hierarchy to achieve the objective of reducing energy consumption. Therefore, the same optimization hierarchy may output different energy consumption optimization policies in different iterations of optimization. For example, the optimization metrics of the first optimization hierarchy are the internal control parameters of the heating equipment, which may include parameters such as frequency and current percentage. Different energy consumption optimization policies can be generated based on different combinations of parameter settings. To ensure the comprehensiveness of the optimization process, the computer device can perform multiple iterations of optimization in a single optimization hierarchy. Based on this, if the verification result of the candidate energy consumption optimization policy output from the target optimization hierarchy is determined to have failed verification, the method further includes the steps of performing iterative optimization calculations and safety and effectiveness verification on the corresponding dataset based on the optimization model of the target optimization hierarchy; and if the number of iterative optimization calculations reaches the limit number of iterations but the target energy consumption optimization policy is not obtained, the method further includes the steps of performing optimization calculations and safety and effectiveness verification on the lower optimization hierarchy of the target optimization hierarchy based on the hierarchy status of the target optimization hierarchy, or determining that the heating equipment is not suitable for energy consumption optimization.

[0063] In the above process, if a target energy consumption optimization policy is obtained before the number of iterative optimization calculations reaches the limit number of iterations, the target energy consumption optimization policy is output and iterative optimization and hierarchical optimization are stopped. If a target energy consumption optimization policy is not obtained even after the number of iterative optimization calculations reaches the limit number of iterations, it is indicated that the current optimization hierarchical level (i.e., the target optimization hierarchical level) cannot perform good energy consumption optimization. If the current optimization hierarchical level is not the highest optimization hierarchical level, the optimization calculation and safety and effectiveness verification of the lower optimization hierarchical level of the current optimization hierarchical level are performed. If the current optimization hierarchical level is the highest optimization hierarchical level, it is determined that the heating equipment is not suitable for energy consumption optimization. The numerical value of the limit number of iterations may be set according to the actual needs, and the setting of the limit number of iterations corresponding to different optimization hierarchical levels may be the same or different, and this application is not limited thereto.

[0064] As described above, the hierarchical energy consumption optimization method according to the embodiment of the present invention acquires target operating data related to energy consumption optimization, performs data hierarchical structuring on the target operating data to obtain datasets for each optimization hierarchical level corresponding to different optimization indicators, determines the optimization starting hierarchical level based on the equipment operating mechanism and the current state, uses the optimization starting hierarchical level in each optimization hierarchical level as the optimization starting point, performs hierarchical optimization calculations based on the optimization model of each optimization hierarchical level and the dataset of each optimization hierarchical level to obtain a target energy consumption optimization policy that matches the current actual needs, and optimizes the energy consumption of the heating equipment based on the target energy consumption optimization policy. By the above method, an optimization starting hierarchical level that is adapted to the equipment operating mechanism and the current state is determined, and hierarchical optimization is performed using the optimization starting hierarchical level as the hierarchical optimization starting point, thereby formulating an energy consumption optimization policy, avoiding unnecessary optimization processes, improving optimization efficiency, and improving the practicality of the energy consumption optimization policy because the acquired target energy consumption optimization policy is adapted to the current state and the equipment operating mechanism.

[0065] In the process of hierarchical optimization of energy consumption, changes in the application scenario or application needs may cause the model applied to the formulation of the energy consumption optimization policy to no longer fit the current application scenario or application needs. In this case, the computer device can improve the practicality of the energy consumption optimization policy by first retraining the model applied to the formulation of the energy consumption optimization policy, and then formulating the energy consumption optimization policy using the model obtained through retraining. Therefore, the method further includes the following steps 209 to 212.

[0066] In step 209, if it is determined that the model retraining conditions are met, a sample set is constructed by combining historical driving data and target driving data.

[0067] The hierarchical optimization method for energy consumption according to the present invention, with respect to the application of the optimization model and safety and effectiveness verification model for each optimization hierarchical level, requires that the optimization model and safety and effectiveness verification model for each optimization hierarchical level be retrained when the model retraining conditions are met, in order to improve the accuracy of the hierarchical optimization of energy consumption. In one possible embodiment, a computer device may determine whether or not the model retraining conditions are met after acquiring target operating data related to the optimization of energy consumption, or in another possible embodiment, the computer device may determine whether or not model retraining is necessary based on a preset determination cycle, and the present invention is not limited thereto.

[0068] For example, after a computer device acquires target operating data related to energy consumption optimization, it determines whether the model retraining conditions are met. In this case, after acquiring the target operating data, the computer device must first determine whether the models involved in the energy consumption optimization process need to be retrained before formulating an energy consumption optimization policy based on the target operating data. If it is determined that the model retraining conditions are met, the model retraining process is performed first, and then the energy consumption optimization policy based on the target operating data is formulated based on the retrained model. If it is determined that the model retraining conditions are not met, the energy consumption optimization policy based on the target operating data can be formulated based on the current model.

[0069] The model retraining conditions may include at least one of the following: the current time reaching a predetermined training time, and the number of policy application negative feedbacks obtained reaching a threshold.

[0070] In one embodiment, the computer device can acquire target operating data and policy application feedback, which may be positive or negative policy application feedback. The computer device can statistically count the number of negative policy application feedbacks, and if it determines that the number of negative policy application feedbacks has reached a threshold, it indicates that the energy consumption optimization effect of the energy consumption optimization policy formulated by the current model is poor and cannot meet actual needs, and it determines that the model needs to be retrained. The number of negative policy application feedbacks obtained may be the cumulative number of negative policy application feedbacks obtained within a certain period of time, or it may be the number of consecutive negative policy application feedbacks obtained.

[0071] Because the policies formulated by the current model may no longer be applicable to actual needs, and a model trained using only historical driving data may be applicable to previous conditions but not to current needs, when retraining the model, a sample set is constructed by combining target driving data and historical driving data, and the model is trained using this combined sample set.

[0072] In step 210, data hierarchies are performed on the sample data in the sample set to obtain a sample dataset applied to the safety and effectiveness verification model and a sample dataset applied to the control policy hierarchical optimization model. The sample dataset applied to the safety and effectiveness verification model includes a subset of sample data applied to the energy consumption effectiveness verification model, a subset of sample data applied to the cooling capacity safety verification model, and a subset of sample data applied to the state safety verification model. The sample dataset applied to the control policy hierarchical optimization model includes a subset of sample data applied to the optimization model of each optimization hierarchy.

[0073] In the embodiments of the present invention, each model is trained, and different models correspond to different sample data subsets. A sample data subset corresponding to each model is obtained by performing data hierarchical structuring on the sample data in the sample set. This process can be achieved by performing data hierarchical cleaning, data hierarchical validation, metric hierarchical fusion, metric hierarchical extraction, and metric hierarchical classification on the sample data in the sample set to obtain a sample dataset applied to the safety and effectiveness validation model and a sample dataset applied to the control policy hierarchical optimization model.

[0074] The sample datasets applied to safety and effectiveness verification include a subset of sample data applied to the energy consumption effectiveness verification model, a subset of sample data applied to the cooling capacity safety verification model, and a subset of sample data applied to state safety verification. The datasets applied to the control policy hierarchical optimization model include a subset of sample data applied to the optimization model of each optimization hierarchy. Exemplarily, the sample datasets applied to safety and effectiveness verification may be represented as {Data_E,Data_C,Data_S}, where Data_E represents the subset of sample data applied to the energy consumption effectiveness verification model, Data_C represents the subset of sample data applied to the cooling capacity safety verification model, and Data_S represents the subset of sample data applied to the state safety verification model. The sample datasets applied to the control policy hierarchical optimization model may be represented as {Data_01,Data_02,Data_03}, where Data_01 represents the subset of sample data applied to the first optimization hierarchy, Data_02 represents the subset of sample data applied to the second optimization hierarchy, and Data_03 represents the subset of sample data applied to the third optimization hierarchy.

[0075] For details regarding the relationships between each processing method in data tiering, please refer to the correspondence in Step 202; the explanation is omitted here.

[0076] In step 211, the safety and effectiveness validation model is trained based on a sample dataset applied to the safety and effectiveness validation model, and the control policy hierarchical optimization model is trained based on a sample dataset applied to the control policy hierarchical optimization model.

[0077] The process of training safety and effectiveness validation models based on sample datasets applied to safety and effectiveness validation models can be achieved by the following steps: performing deep learning training on the energy consumption effectiveness validation model based on a subset of sample data applied to the energy consumption effectiveness validation model and obtaining a retrained energy consumption effectiveness validation model; performing deep learning training on the cooling capacity safety validation model based on a subset of sample data applied to the cooling capacity safety validation model and obtaining a retrained cooling capacity safety validation model; and performing deep learning training on the state safety validation model based on a subset of sample data applied to the state safety validation model and obtaining a retrained state safety validation model.

[0078] In one possible embodiment, a computer device can construct corresponding models using a neural network, such as a DNN (Deep Neural Network) or an LSTM (Long Short-Term Memory) network. For example, the computer device can construct an energy consumption effectiveness verification model and a cooling capacity safety verification model using a DNN, and a state safety verification model using an LSTM network. However, the construction of the above models is merely illustrative, and this application does not limit the type of neural network that forms the basis for model construction.

[0079] As an example, we construct an energy consumption effectiveness verification model and a cooling capacity safety verification model using a DNN, and a state safety verification model using an LSTM network. The loss function used in the training process for these three types of models can be expressed as follows: Model_E~{min(y pue -f DNN (x Data_E )) 2} Model_C ~ {min(y cooling -f DNN (x Data_C )) 2} Model_S ~ {min(y state -f LSTM (x Data_S )) 2} Model_E represents an energy consumption effectiveness verification model, Model_C represents a cooling capacity safety verification model, Model_S represents a state safety verification model, x Data_E , x Data_C and x Data_S represent the input data of the corresponding models respectively, and y pue , y cooling and y state represent the power usage efficiency, cooling capacity and state safety parameters of the corresponding input data respectively. Exemplarily, the state safety parameters may be intake / exhaust or water intake / water discharge temperature, intake / exhaust or water intake / water discharge pressure, etc. The above input data and y values are stored correspondingly in the dataset. Thus, during model training, model training is realized and the accuracy of the model is improved by updating the model parameters according to the difference between the y value and the predicted y value obtained by the model based on the input data. When the safety and effectiveness verification models are applied, by inputting the parameters corresponding to the candidate optimization policies determined by the control policy hierarchical optimization model into each of the safety and effectiveness verification models, the corresponding sub-verification results output from each model are obtained.

[0080] The process of training a control policy hierarchical optimization model based on a sample dataset applied to the control policy hierarchical optimization model can be achieved by the following steps: training the optimization model for the first optimization layer by combinatorial traverse based on a subset of sample data for the first optimization layer and the index parameter constraint range for the first optimization layer, testing the optimization model for the first optimization layer with a retrained safety and effectiveness validation model, and obtaining the retrained optimization model for the first optimization layer; training the optimization model for the second optimization layer by Bayesian optimization based on a subset of sample data for the second optimization layer and the index parameter constraint range for the second optimization layer, testing the optimization model for the second optimization layer with a retrained safety and effectiveness validation model, and obtaining the retrained optimization model for the second optimization layer; and training the optimization model for the third optimization layer by reinforcement learning based on a subset of sample data for the third optimization layer and the index parameter constraint range for the third optimization layer, testing the optimization model for the third optimization layer with a retrained safety and effectiveness validation model, and obtaining the retrained optimization model for the third optimization layer.

[0081] In the embodiments of the present invention, in order to improve the retraining effect of the optimization model, the computer device may first retrain a safety and effectiveness verification model, test the optimization model with the retrained safety and effectiveness verification model, and obtain the retrained optimization model. Testing the optimization model with the retrained safety and effectiveness verification model for any optimization model may mean that the retrained safety and effectiveness verification model performs safety and effectiveness verification on the parameters in the energy consumption optimization policy obtained by the optimization model based on the corresponding sample data subset, and if the verification result indicates that the verification has been passed, it is determined that the optimization model has passed the test and the retrained optimization model is obtained.

[0082] In the training process of the optimization model described above, for any optimization hierarchy, a set of parameter data combinations from the sample data subset corresponding to that hierarchy is input to the optimization model, and the optimization model obtains a predicted parameter data combination (i.e., the parameters in the energy consumption optimization policy predicted by the optimization model). If each parameter data in the predicted parameter data combination is within the index parameter constraint range of the optimization hierarchy, the retrained safety and effectiveness validation model performs safety and effectiveness validation on the predicted parameter data combination. If the validation result indicates a pass, the training completion condition is met, and the above process is repeated for the next set of parameter data combinations in the sample data subset until a retrained optimization model is obtained. The training completion condition may be that all parameter data combinations in the sample data subset have been iteratively trained, or that the number of times the trained optimization model has passed the tests of the retrained safety and effectiveness validation model is greater than the count threshold.

[0083] If any parameter data in a combination of predicted parameter data falls outside the metric parameter constraint range of the optimization hierarchy, or if the validation result indicates failure, the process of optimizing the model parameters of the optimization model and obtaining combinations of predicted parameter data is repeated, and subsequent processes are carried out until a retrained optimization model is obtained.

[0084] In the embodiments of this application, the constraint range of the index parameters for each optimization hierarchy indicates the optimization space of the corresponding optimization index. For example, if the frequency constraint range is 50Hz to 100Hz, the predicted frequency in the combination of predicted parameter data obtained by the optimization model must be within the range of 50Hz to 100Hz. If it exceeds this range, it indicates that the optimization result of the optimization model is inappropriate and that the model training must be performed again.

[0085] In step 212, hierarchical optimization calculations are performed based on the retrained control policy hierarchical optimization model, and safety and effectiveness validation is performed based on the retrained safety and effectiveness validation model.

[0086] After completing retraining for each model, the process of formulating a hierarchical optimization policy for energy consumption is carried out based on the retrained models.

[0087] By updating each model used in the hierarchical energy consumption optimization process in real time, the energy consumption optimization policy obtained through model processing can be more closely aligned with current actual needs, thereby improving the adaptability of the energy consumption optimization policy.

[0088] Based on the embodiment shown in Figure 2, the computer device acquires target operating data related to energy consumption optimization, and then, as an example, determines whether the model retraining conditions are met. Figure 3 shows a flowchart of a hierarchical energy consumption optimization method according to one exemplary embodiment of the present invention. This method may be performed by a computer device, which may be implemented as a heating equipment management device. As shown in Figure 3, the hierarchical energy consumption optimization method may include the following steps S301 to S310.

[0089] In step S301, target driving data is acquired.

[0090] In step S302, it is determined whether the model retraining conditions are met. If they are not met, step S303 is executed; otherwise, step S307 is executed.

[0091] The model retraining conditions may include at least one of the following: the current time reaching a preset training time, and the number of policy application negative feedbacks obtained reaching a threshold. If the model retraining conditions include a determination of the number of negative feedbacks, the computer device can obtain policy application feedback when acquiring target driving data.

[0092] In step S303, it is determined whether the target operating data passed the optimization pre-inspection. If it passed, S304 is executed; otherwise, it is determined that the heating equipment is not suitable for energy consumption optimization.

[0093] In step S304, data hierarchies are created for the target driving data.

[0094] The process can be understood by referring to the relevant details in the embodiment shown in Figure 2, and will not be explained here.

[0095] In step S305, the optimization starting hierarchy is determined based on the current state indicated by the equipment operating mechanism and target operating data, and a hierarchical optimization calculation is performed.

[0096] The process can be understood by referring to the relevant details in the embodiment shown in Figure 2, and will not be explained here.

[0097] In step S306, the target energy consumption optimization policy is output.

[0098] In step S307, historical driving data and target driving data are combined to form a sample set.

[0099] In step S308, data hierarchicalization is performed on the sample data in the sample set to obtain a sample dataset to be applied to the safety and effectiveness validation model and a sample dataset to be applied to the control policy hierarchical optimization model.

[0100] In step S309, the safety and efficacy validation model is trained based on the sample dataset applied to the safety and efficacy validation model.

[0101] In step S310, the control policy hierarchical optimization model is trained based on a sample dataset applied to the control policy hierarchical optimization model.

[0102] After obtaining the retrained safety and effectiveness verification model and the control policy hierarchical optimization model, step S305 is performed to obtain the target energy consumption optimization policy for the target operating data.

[0103] For embodiments of the above process, please refer to the relevant details in the embodiment shown in Figure 2; therefore, a detailed explanation is omitted here.

[0104] Figure 4 shows a schematic diagram of a hierarchical energy consumption optimization system according to one exemplary embodiment of the present invention, which is used to obtain a target energy consumption optimization policy by performing all or some of the steps of the embodiment shown in Figure 1, Figure 2, or Figure 3, and as shown in Figure 4, the hierarchical energy consumption optimization system may include a data hierarchical service module 410, a data hierarchical processing module 420, a safety and effectiveness verification module 430, and a control policy hierarchical optimization module 440, which constitute an energy consumption hierarchical intelligent optimization assembly.

[0105] The data tiering service module 410 is used to acquire target operating data and can act as an intermediary between the actual data center and an energy consumption hierarchical intelligent optimization assembly that performs hierarchical optimization of energy consumption. The data tiering service module can collect various indoor and outdoor environmental data, equipment operating status data, equipment control setting parameters, control feedback data, equipment combination operating modes, number of equipment in operation, IT center load and energy consumption data, cooling system and power supply energy consumption data, etc. from the actual data center. Considering that hierarchical processing is required thereafter, the data tiering service module can also perform hierarchical storage, hierarchical transfer, and constraint hierarchical arrangement on the acquired target operating data after acquiring it. On the other hand, the data tiering service module is also responsible for distributing the target energy consumption optimization policy acquired by the energy consumption hierarchical intelligent optimization assembly to the actual data center.

[0106] The data hierarchical processing module 420 performs data hierarchical processing on the input operating data, including data hierarchical cleaning, data hierarchical validation, metric hierarchical fusion, metric hierarchical extraction, and metric hierarchical classification, to obtain datasets to be applied to the safety and effectiveness validation model and datasets to be applied to the control policy hierarchical optimization model, and transmits the corresponding datasets to the safety and effectiveness validation module 430 and the control policy hierarchical optimization module 440, respectively.

[0107] The safety and effectiveness verification module 430 can be used to train a safety and effectiveness verification model and to perform safety and effectiveness verification on an energy consumption optimization policy using the safety and effectiveness verification model.

[0108] The control policy hierarchical optimization module 440 can be used to train a control policy hierarchical optimization model, determine the optimization starting hierarchical level, and perform hierarchical optimization calculations based on the optimization model at each optimization hierarchical level to obtain a target energy consumption optimization policy.

[0109] Figure 5 shows a structural block diagram of a computer device 500 according to one exemplary embodiment of the present application. The computer device may be implemented as a heating equipment management device in the above-mentioned technical means of the present application. The computer device 500 includes a central processing unit (CPU) 501, a system memory 504 including random access memory (RAM) 502 and read-only memory (ROM) 503, and a system bus 505 connecting the system memory 504 and the central processing unit 501. The computer device 500 further includes a mass storage device 506 for storing an operating system 509, application programs 510 and other program modules 511.

[0110] Without loss of generality, the computer-readable media described above may include computer storage media and communication media. Computer storage media include volatile and non-volatile, removable and non-removable media implemented by any method or technique for storing information such as computer-readable instructions, data structures, program modules, or other data. Computer storage media include RAM, ROM, Erasable Programmable Read Only Memory (EPROM), Electrically-Erasable Programmable Read-Only Memory (EEPROM), flash memory or other solid-state storage technologies, CD-ROM, Digital Versatile Disc (DVD) or other optical storage, magnetic tape cartridges, magnetic tapes, disk storage or other magnetic storage devices. Naturally, as will be apparent to those skilled in the art, the computer storage media described above are not limited to the above types. The system memory 504 and the mass storage device 506 may be collectively referred to as memory.

[0111] According to various embodiments of the present invention, the computer device 500 may further be connected to and operated by a remote computer on a network, such as the Internet. That is, the computer device 500 may be connected to a network 508 by a network interface unit 507 connected to the system bus 505, or it may be connected to other types of networks or remote computer systems (not shown) using the network interface unit 507.

[0112] The memory further includes at least one instruction, at least one program, code set, or instruction set, the at least one instruction, at least one program, code set, or instruction set being stored in the memory, and the central processing unit 501 executes the at least one instruction, at least one program, code set, or instruction set to realize all or part of the steps of the hierarchical energy consumption optimization method shown in each of the above embodiments.

[0113] In one exemplary embodiment, a computer-readable storage medium storing at least one computer program is further provided, and all or part of the steps of the hierarchical optimization method for energy consumption are realized when the computer program is loaded and executed by a processor. For example, the computer-readable storage medium may be read-only memory (ROM), random access memory (RAM), compact disc read-only memory (CD-ROM), magnetic tape, floppy disk, and optical data storage device.

[0114] In one exemplary embodiment, a computer program product is further provided, which includes at least one computer program, and which, when loaded and executed by a processor, realizes all or some steps of the hierarchical energy consumption optimization method shown in any of the embodiments of Figure 1, Figure 2, or Figure 3.

[0115] A person skilled in the art will readily conceive of other embodiments of the Application after reviewing the specification and practicing the invention disclosed herein. The Application is intended to include any variations, uses, or adaptive changes of the Application, which include common or conventional technical means in the Art not disclosed herein, in accordance with the general principles of the Application. The Specification and Examples are illustrative only, and the true scope and spirit of the Application are indicated by the Claims.

[0116] It should be understood that this application is not limited to the exact structure described above and shown in the drawings, and that various modifications and changes may be made without departing from its scope. The scope of this application is limited only to the attached claims.

Claims

1. The steps include obtaining target operating data related to the optimization of energy consumption, The steps include: performing data hierarchical structure on the aforementioned target operating data, obtaining datasets corresponding to each optimization hierarchy, and assigning datasets from different optimization hierarchies to different optimization indicators; A step of determining the optimization start level based on the equipment operating mechanism and the current state indicated by the target operating data, A method for hierarchical optimization of energy consumption, comprising the steps of: using the aforementioned optimization starting hierarchy as the optimization starting point, performing hierarchical optimization calculations based on the optimization model of each optimization hierarchy and the dataset corresponding to each optimization hierarchy, obtaining a target energy consumption optimization policy, and optimizing the energy consumption of heating equipment based on the aforementioned target energy consumption optimization policy.

2. The step of obtaining a target energy consumption optimization policy involves using the aforementioned optimization starting hierarchy as the optimization starting point, performing hierarchical optimization calculations based on the optimization model of each optimization hierarchy and the dataset corresponding to each optimization hierarchy, and obtaining the target energy consumption optimization policy. When the optimization starts from the aforementioned optimization starting layer, and the optimization is performed in one of the target optimization layers, the optimization calculation is performed on the corresponding dataset using the optimization model of the target optimization layer to obtain a candidate energy consumption optimization policy. The steps include: performing safety and effectiveness verification on the candidate energy consumption optimization policy using a safety and effectiveness verification model, and obtaining the verification results for the candidate energy consumption optimization policy; The method according to claim 1, comprising the step of determining the candidate energy consumption optimization policy as the target energy consumption optimization policy if the verification result of the candidate energy consumption optimization policy indicates that the candidate energy consumption optimization policy has passed the verification.

3. The aforementioned method, If the target optimization hierarchy is not the highest-level optimization hierarchy among the various optimization hierarchies, and the verification result of the candidate energy consumption optimization policy indicates that the candidate energy consumption optimization policy has failed to be verified, the steps include performing optimization calculations and safety and effectiveness verification for the lower-level optimization hierarchies of the target optimization hierarchy. The method according to claim 2, further comprising the step of determining that the heating equipment is not suitable for energy consumption optimization if, when the target optimization hierarchy is the highest optimization hierarchy among the various optimization hierarchies, the verification result of the candidate energy consumption optimization policy indicates that the candidate energy consumption optimization policy has failed to be verified.

4. If the verification result of the candidate energy consumption optimization policy indicates that the candidate energy consumption optimization policy has failed to be verified, the method The steps include performing iterative optimization calculations and safety and effectiveness verification on the corresponding dataset based on the optimization model of the aforementioned target optimization hierarchy, The method according to claim 3, further comprising the step of, if the number of iterative optimization calculations reaches the limit number of iterations but the target energy consumption optimization policy is not obtained, performing optimization calculations and safety and effectiveness verification of the optimization hierarchy below the target optimization hierarchy based on the hierarchy status of the target optimization hierarchy, or determining that the heating equipment is not suitable for energy consumption optimization.

5. The aforementioned target operating data includes changes in indoor and outdoor temperature and humidity and information technology (IT) load. The step of determining the optimization start level based on the equipment operating mechanism and the current state indicated by the target operating data is: The steps include determining a first optimization level as the optimization start level, which indicates that the internal control parameters of the heating equipment should be optimized when the outdoor temperature is within a preset outdoor temperature range, and the indoor and outdoor temperature and humidity changes periodically, as well as the IT load changes periodically. The steps include determining a second optimization level as the optimization start level, which indicates optimizing the number of heating equipment units and internal control parameters, if the outdoor temperature is within the outdoor temperature range and the changes in indoor and outdoor temperature and humidity indicate a rapid change in indoor and outdoor temperature and humidity or the changes in IT load indicate a rapid change in IT load, The method according to claim 1, comprising the step of determining a third optimization level as the optimization start level, which indicates that when the outdoor temperature is within a critical temperature range between different outdoor temperature ranges, the operating mode, number of units, and internal control parameters of the heating equipment are optimized.

6. The step of performing data hierarchical structure on the aforementioned target driving data and obtaining a dataset corresponding to each optimization hierarchy is: The steps include performing data hierarchical cleaning, data hierarchical validation, metric hierarchical fusion, metric hierarchical extraction, and metric hierarchical classification on the target operating data to obtain a dataset to be applied to the safety and effectiveness validation model, the control policy hierarchical optimization model includes the optimization model for each optimization hierarchy, and the dataset applied to the control policy hierarchical optimization model includes a subset of data applied to the optimization model for each optimization hierarchy. The method according to claim 1, wherein the data hierarchical validation is used to verify the accuracy of each sample data, the indicator hierarchical fusion is used to perform data statistics on related indicators and generate new indicators, the indicator hierarchical extraction is used to extract the sets of indicators required for different models and construct corresponding data subsets, and the indicator hierarchical classification is used to classify each indicator according to its type.

7. The safety and effectiveness verification model includes an energy consumption effectiveness verification model, a cooling capacity safety verification model, and a state safety verification model, and the step of performing safety and effectiveness verification on the energy consumption optimization policy obtained using the optimization model by the safety and effectiveness verification model is: The steps include: performing an energy consumption effectiveness verification on the energy consumption optimization policy using the aforementioned energy consumption effectiveness verification model and obtaining the first sub-verification result; The steps include: performing a cooling capacity safety verification against the energy consumption optimization policy using a cooling capacity safety verification model and obtaining the results of the second sub-verification; The process involves performing a state safety verification on the energy consumption optimization policy using a state safety verification model and obtaining the results of the third sub-verification, The method according to claim 2, comprising the step of generating a verification result indicating whether the energy consumption optimization policy passed or failed verification based on the first sub-verification result, the second sub-verification result, and the third sub-verification result.

8. The aforementioned method, If it is determined that the model retraining conditions are met, the steps include: constructing a sample set by combining historical operating data and the target operating data; A step of performing data hierarchicalization on the sample data in the sample set and obtaining a sample dataset to be applied to a safety and effectiveness verification model and a sample dataset to be applied to a control policy hierarchical optimization model, wherein the sample dataset to be applied to the safety and effectiveness verification model includes a subset of sample data to be applied to an energy consumption effectiveness verification model, a subset of sample data to be applied to a cooling capacity safety verification model, and a subset of sample data to be applied to a state safety verification model, and the sample dataset to be applied to the control policy hierarchical optimization model includes a subset of sample data to be applied to the optimization model of each optimization hierarchy, The steps include: training the safety and effectiveness validation model based on a sample dataset applied to the safety and effectiveness validation model, and training the control policy hierarchical optimization model based on a sample dataset applied to the control policy hierarchical optimization model; The method according to claim 1, further comprising the steps of performing hierarchical optimization calculations based on a retrained control policy hierarchical optimization model and performing safety and effectiveness verification based on a retrained safety and effectiveness verification model.

9. The step of training the safety and efficacy validation model based on the sample dataset to which the safety and efficacy validation model is applied is: The steps include: performing deep learning training on the energy consumption effectiveness validation model based on a subset of sample data applied to the energy consumption effectiveness validation model, and obtaining a retrained energy consumption effectiveness validation model; The steps include: performing deep learning training on the cooling capacity safety verification model based on a subset of sample data applied to the cooling capacity safety verification model, and obtaining a retrained cooling capacity safety verification model; The method according to claim 8, comprising the steps of performing deep learning training on a state safety validation model based on a subset of sample data applied to the state safety validation model, and obtaining a retrained state safety validation model.

10. The step of training a control policy hierarchical optimization model based on a sample dataset applied to the control policy hierarchical optimization model is: The steps include: training an optimization model for the first optimization tier by combinatorial traversal based on a sample data subset of the first optimization tier and the constraint range of the index parameters of the first optimization tier; testing the optimization model for the first optimization tier with a retrained safety and effectiveness validation model; and obtaining a retrained optimization model for the first optimization tier. Steps include: training an optimization model for the second optimization layer by Bayesian optimization based on a sample data subset of the second optimization layer and the constraint range of the index parameters of the second optimization layer; testing the optimization model for the second optimization layer with a retrained safety and effectiveness validation model; and obtaining the retrained optimization model for the second optimization layer. The method according to claim 8, comprising the steps of: training an optimization model for the third optimization layer by reinforcement learning based on a sample data subset of the third optimization layer and the index parameter constraint range of the third optimization layer; testing the optimization model for the third optimization layer with a retrained safety and effectiveness validation model; and obtaining a retrained optimization model for the third optimization layer.

11. The method according to claim 8, wherein the model retraining condition includes at least one of the following: the current time reaching a preset training time, and the number of policy application negative feedbacks obtained reaching a count threshold.

12. Before performing data hierarchical structuring on the aforementioned target operating data and obtaining a dataset corresponding to each optimization hierarchy, the method, The process further includes the step of performing an optimization pre-test on the target operating data using an optimization pre-test index and obtaining the optimization pre-test results, wherein the optimization pre-test index includes at least one of checking whether the necessary index for each model is missing, and checking whether the service relationships between the indexes are established. The step of performing data hierarchical structure on the aforementioned target driving data and obtaining a dataset corresponding to each optimization hierarchy is: The method according to claim 1, further comprising the step of performing data hierarchical structuring on the target operating data and obtaining a dataset corresponding to each optimization hierarchy if the optimization pre-inspection result indicates that the target operating data has passed the pre-inspection.

13. The aforementioned method, The method according to claim 12, further comprising the step of determining that the heating equipment is not suitable for energy consumption optimization if the optimization pre-inspection results indicate that the target operating data failed the pre-inspection.

14. A computer device comprising a processor and memory storing at least one computer program, wherein the hierarchical optimization method for energy consumption described in any one of claims 1 to 13 is realized by loading and executing the at least one computer program by the processor.

15. A computer-readable storage medium in which at least one computer program is stored, and the hierarchical optimization method for energy consumption described in any one of claims 1 to 13 is realized when the computer program is loaded and executed by a processor.

Images