Cold capacity determination method and device, electronic equipment and readable storage medium

Abstract

The invention relates to the technical field of cold quantity prediction, and provides a cold quantity determination method and device, electronic equipment and a readable storage medium. The method comprises the steps that attention mechanism processing is conducted on at least one cooling capacity feature tensor corresponding to a to-be-processed object through a trained cooling capacity determination model, and all cooling capacity feature weight vectors corresponding to the to-be-processed object are obtained; performing normalization processing on each cooling capacity feature weight vector corresponding to the to-be-processed object to obtain each normalized feature weight value corresponding to the to-be-processed object; performing weighting processing on each cooling capacity feature tensor corresponding to the to-be-processed object and each normalized feature weight value corresponding to the to-be-processed object to obtain each weighted feature vector corresponding to the to-be-processed object; and linear transformation is carried out on each weighted feature vector corresponding to the to-be-processed object to obtain the target cooling capacity data corresponding to the to-be-processed object, and the target cooling capacity data is used for cooling capacity release control, so that the accuracy of cooling capacity determination is improved.

Description

Technical Field

[0001] The present disclosure relates to the technical field of cooling capacity prediction, and in particular to a cooling capacity determination method, device, electronic device and readable storage medium. Background Art

[0002] In large shopping malls, the cooling capacity is usually determined and generated through a complete refrigeration system. The cooling capacity determination process is usually through manual judgment and input, which increases the workload and reduces the cooling capacity release efficiency. Therefore, it is necessary to predict the cooling capacity in the future. Conventional cooling capacity prediction methods are mainly based on machine learning and statistical models. The above methods require a large amount of historical data for training models, such as support vector machines, decision trees, or neural networks. The core of these models is to predict future cooling demand by learning patterns in historical data. In shopping malls with sufficient data, data based on characteristics such as ambient temperature, humidity, and traffic flow can more accurately predict cooling demand. However, this model relies on manual or statistical methods for feature selection and cannot dynamically adjust feature weights according to the importance of the target task, resulting in the model being overly dependent on unimportant features.

[0003] It can be seen that the existing technology has the problem that the weight cannot be adjusted during feature selection, resulting in low accuracy and weak correlation in the determination results of the cooling capacity. Summary of the invention

[0004] In view of this, the embodiments of the present disclosure provide a method, device, electronic device and readable storage medium for determining cooling capacity to solve the problem in the prior art that the weight cannot be adjusted due to feature selection, resulting in low accuracy and weak correlation in the cooling capacity determination results.

[0005] According to a first aspect of an embodiment of the present disclosure, a method for determining cold capacity is provided, comprising: performing attention mechanism processing on at least one cold capacity feature tensor corresponding to a processed object through a trained cold capacity determination model to obtain each cold capacity feature weight vector corresponding to the processed object; performing normalization processing on each cold capacity feature weight vector corresponding to the processed object to obtain each normalized feature weight value corresponding to the processed object; performing weighted processing on each cold capacity feature tensor corresponding to the processed object and each normalized feature weight value corresponding to the processed object to obtain each weighted feature vector corresponding to the processed object; performing linear transformation on each weighted feature vector corresponding to the processed object to obtain target cold capacity data corresponding to the processed object, and sending the target cold capacity data to a terminal device for cold capacity release control.

[0006] According to a second aspect of an embodiment of the present disclosure, a device for determining cold capacity is provided, including: a first processing module, which is used to perform attention mechanism processing on at least one cold capacity feature tensor corresponding to a processed object through a trained cold capacity determination model, and obtain each cold capacity feature weight vector corresponding to the processed object; a second processing module, which is used to perform normalization processing on each cold capacity feature weight vector corresponding to the processed object, and obtain each normalized feature weight value corresponding to the processed object; a third processing module, which is used to perform weighted processing on each cold capacity feature tensor corresponding to the processed object and each normalized feature weight value corresponding to the processed object, and obtain each weighted feature vector corresponding to the processed object; a fourth processing module, which is used to perform linear transformation on each weighted feature vector corresponding to the processed object, and obtain target cold capacity data corresponding to the processed object, and send the target cold capacity data to a terminal device for cold capacity release control.

[0007] According to a third aspect of an embodiment of the present disclosure, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0008] According to a fourth aspect of an embodiment of the present disclosure, a readable storage medium is provided, which stores a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0009] Compared with the prior art, the disclosed embodiment has the following beneficial effects: by inputting at least one cold quantity feature tensor corresponding to the object to be processed into the trained cold quantity determination model, the cold quantity feature tensor can be processed by the cold quantity determination model through an attention mechanism and converted into a cold quantity feature weight vector corresponding to the object to be processed, and then each cold quantity feature weight vector corresponding to the object to be processed can be normalized by a normalized exponential function to obtain each normalized feature weight value corresponding to the object to be processed, each cold quantity feature tensor corresponding to the object to be processed and each normalized feature weight value corresponding to the object to be processed can be weighted, and the weight value of the cold quantity feature tensor can be adjusted to obtain a redistributed weight. The weighted feature vectors corresponding to the object to be processed can then be linearly transformed through the activation function to obtain the target cooling data corresponding to the object to be processed. The target cooling data can be used to guide the corresponding terminal device to release cooling in a specified time period. The cooling determination model completed through training can perform refined analysis and processing on the specific cooling features of the object to be processed. Through the attention mechanism, the cooling determination model can identify and weight key cooling features, thereby improving the accuracy and reliability of cooling determination and improving processing efficiency. The flexibility of the cooling determination method is enhanced by combining and weighting different cooling features. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0011] Figure 1 is a schematic diagram of an application scenario of an embodiment of the present disclosure; Figure 2 is a flow chart of a method for determining cooling capacity provided by an embodiment of the present disclosure; Figure 3 is a structural schematic diagram of a device for determining cooling capacity provided by an embodiment of the present disclosure; Figure 4 It is a structural schematic diagram of an electronic device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0012] In the following description, specific details such as specific system structures and technologies are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present disclosure. However, it should be clear to those skilled in the art that the present disclosure may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obstructing the description of the present disclosure with unnecessary details.

[0013] It should be noted that the user information (including but not limited to terminal device information, user personal information, etc.) and data (including but not limited to data for display, data for analysis, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties.

[0014] A method and device for determining cooling capacity according to an embodiment of the present disclosure will be described in detail below with reference to the accompanying drawings.

[0015] Figure 1 Schematic diagram of an application scenario of an embodiment of the present disclosure. The application scenario may include terminal devices 1, 2 and 3, a server 4 and a network 5.

[0016] Terminal devices 1, 2 and 3 can be hardware or software. When terminal devices 1, 2 and 3 are hardware, they can be various electronic devices with display screens and supporting communication with server 4, including but not limited to wall-mounted air conditioners, floor-standing air conditioners, central air conditioners and refrigeration control panels, etc.; when terminal devices 1, 2 and 3 are software, they can be installed in the above electronic devices. Terminal devices 1, 2 and 3 can be implemented as multiple software or software modules, or as a single software or software module, and the embodiments of the present disclosure are not limited to this. Furthermore, various applications can be installed on terminal devices 1, 2 and 3, such as data processing applications, instant messaging tools, social platform software, search applications, shopping applications, etc.

[0017] The server 4 may be a server that provides various services, for example, a background server that receives a request sent by a terminal device that establishes a communication connection with the server, and the background server may receive and analyze the request sent by the terminal device, and generate a processing result. The server 4 may be a single server, or a server cluster composed of several servers, or a cloud computing service center, which is not limited in the embodiments of the present disclosure.

[0018] It should be noted that the server 4 can be hardware or software. When the server 4 is hardware, it can be various electronic devices that provide various services for the terminal devices 1, 2, and 3. When the server 4 is software, it can be multiple software or software modules that provide various services for the terminal devices 1, 2, and 3, or it can be a single software or software module that provides various services for the terminal devices 1, 2, and 3, and the embodiments of the present disclosure are not limited to this.

[0019] The network 5 may be a wired network connected by coaxial cable, twisted pair, and optical fiber, or a wireless network that can interconnect various communication devices without wiring, such as Bluetooth, Near Field Communication (NFC), infrared, etc., which is not limited in the embodiments of the present disclosure.

[0020] The user can establish a communication connection with the server 4 via the network 5 through the terminal devices 1, 2 and 3 to receive or send information, etc. Specifically, the server 4 can obtain at least one cold feature tensor corresponding to the object to be processed via the terminal devices 1, 2 and 3, and by inputting at least one cold feature tensor corresponding to the object to be processed into the trained cold determination model, the cold feature tensor can be processed by the cold determination model for attention mechanism, and converted into a cold feature weight vector corresponding to the object to be processed, and then each cold feature weight vector corresponding to the object to be processed can be normalized by a normalized exponential function to obtain each normalized feature weight value corresponding to the object to be processed, each cold feature tensor corresponding to the object to be processed and each normalized feature weight value corresponding to the object to be processed can be weighted, and the weight value of the cold feature tensor can be adjusted to obtain each weighted feature vector corresponding to the object to be processed with a redistributed weight, and then each weighted feature vector corresponding to the object to be processed can be linearly transformed by an activation function to obtain the target cold data corresponding to the object to be processed, and the target cold data can be used to guide the corresponding device terminal to release cold in a specified time period.

[0021] It should be noted that the specific types, quantities and combinations of the terminal devices 1, 2 and 3, the server 4 and the network 5 can be adjusted according to the actual needs of the application scenario, and the embodiments of the present disclosure are not limited to this.

[0022] Figure 2 It is a flow chart of a method for determining cooling capacity provided in an embodiment of the present disclosure. Figure 2 The cooling capacity can be determined by Figure 1 The server executes. Figure 2 As shown, the method for determining the cooling capacity includes: Step 201, using the trained cold quantity determination model, performs attention mechanism processing on at least one cold quantity feature tensor corresponding to the object to be processed, and obtains each cold quantity feature weight vector corresponding to the object to be processed.

[0023] Specifically, a cold capacity determination model that has been pre-trained can be a model based on deep learning training. The cold capacity determination model can learn and identify patterns related to cold capacity, and perform attention mechanism processing on at least one cold capacity feature tensor corresponding to a given object to be processed, wherein the object to be processed can be a building or refrigeration equipment, etc., which is not limited here. The cold capacity feature tensor can be a tensor structure containing multidimensional cold capacity data such as temperature, humidity, and wind speed. Through the attention mechanism processing, weights can be configured for each cold capacity feature tensor, and then features that are more relevant to the cold capacity determination can be identified, and each cold capacity feature weight vector corresponding to the object to be processed can be obtained. The specific content of the cold capacity determination includes but is not limited to energy consumption prediction or performance optimization, etc., thereby enhancing the cold capacity determination model's ability to capture cold capacity characteristics, improving the accuracy of predicting or analyzing cold capacity in the future, and improving efficiency and performance in application scenarios such as energy management and equipment maintenance.

[0024] Among them, the cooling capacity determination model can be a deep learning model obtained by training a large amount of historical cooling data. The cooling capacity determination model can identify features related to cooling capacity from environmental parameters, and predict cooling capacity for a period of time in the future based on the identified features, which can be used for subsequent analysis, prediction or optimization tasks; the object to be processed can refer to the target entity that needs to be determined or analyzed for cooling capacity, including but not limited to buildings, refrigeration equipment in industrial production lines, or data center cooling systems.

[0025] The cooling characteristic tensor corresponding to the object to be processed can be a multidimensional data set describing the cooling state of the object to be processed. The cooling characteristic tensor corresponding to the object to be processed can include measurement values ​​of multiple dimensions such as indoor temperature, humidity, air flow velocity, and / or cooling water temperature, and can be stored in the form of a tensor.

[0026] The cold feature weight vector corresponding to the object to be processed can be the result of attention mechanism processing. The cold feature weight vector corresponding to the object to be processed can be used to characterize the correlation or influence of each cold feature in the cold determination. The cold feature weight vector corresponding to the object to be processed can be a vector containing weight values.

[0027] For example, in an intelligent building energy management system, if it is necessary to predict the cooling capacity of the building's cooling system in the next hour, the real-time cooling characteristic data of the building, such as indoor temperature, humidity, and / or external temperature, can be collected through sensors to form a cooling feature tensor. The cooling feature tensor can then be processed through the trained cooling determination model and attention mechanism to obtain various cooling feature weight vectors. The cooling feature weight vectors can be used to characterize environmental factors that have a more significant impact on the cooling system's cooling determination.

[0028] Step 202 , normalizing each cooling characteristic weight vector corresponding to the object to be processed to obtain each normalized characteristic weight value corresponding to the object to be processed.

[0029] Specifically, normalization processing can be achieved by dividing each weight of each cold feature weight vector corresponding to the object to be processed by the sum of all components in the cold feature weight vector, which can ensure that the sum of the weight values ​​of all features is 1, thereby obtaining each normalized feature weight value corresponding to the object to be processed. In this way, the dimensional difference between different feature weights is eliminated through normalization processing, the stability and interpretability of the cold determination model are enhanced, and the dependence of the cold determination model on each feature is made more intuitive and easy to understand; since the normalized feature weight vector conforms to the actual physical meaning, the prediction accuracy of the model is improved, and the data processing efficiency is improved.

[0030] Among them, the normalized feature weight value corresponding to the object to be processed can be obtained by normalizing the cold feature weight vectors corresponding to the object to be processed into a vector whose sum of all components is 1. Each component of the vector can be the normalized feature weight value of the corresponding feature. The normalized feature weight value can be calculated by dividing each component in the cold feature weight vector by the sum of the cold feature weight vector. The normalized feature weight value corresponding to the object to be processed can be used to characterize the relative importance of each feature in describing the cold characteristics of the object to be processed.

[0031] For example, in the energy consumption management of intelligent buildings, the cooling capacity of the air-conditioning system in different rooms can be determined by taking the room temperature, humidity, light intensity, and / or population density as cooling characteristics. Then, by weighting and normalizing the cooling characteristics, the normalized feature weight value corresponding to each room can be obtained. The normalized feature weight value can be used to judge the actual cooling demand of each room, thereby optimizing the operation strategy of the air-conditioning system, such as adjusting the cooling capacity, air supply speed, etc.

[0032] Step 203 , weighted processing is performed on each cold quantity feature tensor corresponding to the object to be processed and each normalized feature weight value corresponding to the object to be processed to obtain each weighted feature vector corresponding to the object to be processed.

[0033] Specifically, through weighted processing, each element in each cold feature tensor corresponding to the object to be processed can be multiplied by its corresponding normalized feature weight value, and the above multiplication step can be performed on each feature in the cold feature tensor to ensure that the importance of each feature can be reflected according to its normalized weight value, and then the weighted feature vectors corresponding to the object to be processed can be obtained, thereby enhancing the sensitivity of the cold determination model to the features. Since the weighted processing reduces the interference of non-relevant features on the results, the model focuses more on the information of relevant features, thereby improving the prediction accuracy and robustness of the cold determination model. The weighted feature vector can be flexibly adjusted according to different task requirements, providing a more accurate and efficient data basis for subsequent model training, feature selection or decision making.

[0034] Among them, the weighted feature vector corresponding to the object to be processed can be obtained by multiplying the elements in each cold feature tensor of the object to be processed with its corresponding normalized feature weight value. The weighted feature vector contains the information of the cold feature tensor and can represent the relative importance of each feature in determining the cold through the weight value.

[0035] For example, in an intelligent temperature control system, in order to determine the cooling demand in a building, multiple factors such as indoor temperature, external temperature, and solar radiation can be used as cooling characteristics. After obtaining the normalized feature weight value corresponding to the building, weighted processing is performed to obtain the weighted feature vector corresponding to the building.

[0036] Step 204 , linearly transform each weighted feature vector corresponding to the object to be processed to obtain target cooling data corresponding to the object to be processed, and send the target cooling data to the device terminal for cooling release control.

[0037] Specifically, according to each weighted feature vector corresponding to the object to be processed, these weighted feature vectors can be processed by an activation function, which is not limited here, and then the target cooling data corresponding to the object to be processed can be obtained, wherein the weighted feature vector may include factors such as ambient temperature, humidity, energy consumption rate, and / or occupancy rate, which are not limited here, thereby enhancing the sensitivity of the cooling determination model to environmental changes, improving the accuracy of cooling determination, improving energy utilization efficiency, and reducing energy waste.

[0038] Among them, the target cooling data corresponding to the object to be processed can be the cooling supply amount obtained by calculation based on the current state and environmental conditions of the object to be processed; the terminal device can be a refrigeration device that can receive and execute control instructions, including but not limited to the control unit of the air-conditioning system, the water pump and valve system of the cooling tower, etc.

[0039] For example, in the management of large data centers, the target cooling data required for the data center to maintain the optimal operating temperature can be calculated through linear transformation processing based on weighted feature vectors such as the current server load, ambient temperature, and humidity. The target cooling data is sent to the air-conditioning system inside the data center, and the air-conditioning system can adjust parameters such as the supply air temperature and wind speed based on the target cooling data.

[0040] According to the technical solution provided by the embodiment of the present disclosure, by inputting at least one cold quantity feature tensor corresponding to the object to be processed into the trained cold quantity determination model, the cold quantity feature tensor can be processed by the cold quantity determination model through the attention mechanism and converted into a cold quantity feature weight vector corresponding to the object to be processed, and then each cold quantity feature weight vector corresponding to the object to be processed can be normalized by a normalized exponential function to obtain each normalized feature weight value corresponding to the object to be processed, each cold quantity feature tensor corresponding to the object to be processed and each normalized feature weight value corresponding to the object to be processed can be weighted, and the weight value of the cold quantity feature tensor can be adjusted to obtain the object to be processed with the reallocated weight. The weighted feature vectors corresponding to the object can be linearly transformed through the activation function to obtain the target cooling data corresponding to the object to be processed. The target cooling data can be used to guide the corresponding terminal device to release cooling in a specified time period. The cooling determination model completed through training can perform refined analysis and processing on the specific cooling features of the object to be processed. Through the attention mechanism, the cooling determination model can identify and weight key cooling features, thereby improving the accuracy and reliability of cooling determination and improving processing efficiency. The flexibility of the cooling determination method is enhanced by combining and weighting different cooling features.

[0041] In some embodiments, before the cold quantity determination model completed through training performs attention mechanism processing on at least one cold quantity feature tensor corresponding to the object to be processed to obtain each cold quantity feature weight vector corresponding to the object to be processed, it also includes: obtaining at least one historical cold quantity feature data corresponding to the object to be processed; performing data type check processing on each historical cold quantity feature data corresponding to the object to be processed to obtain date type data corresponding to the object to be processed, numerical feature data corresponding to the object to be processed, and discrete feature data corresponding to the object to be processed; performing feature construction processing on the date type data corresponding to the object to be processed to obtain time information derived data corresponding to the object to be processed; performing standardization processing on the numerical feature data corresponding to the object to be processed to obtain dimensional data corresponding to the object to be processed; performing one-hot encoding processing on the discrete feature data corresponding to the object to be processed to obtain one-hot data corresponding to the object to be processed; formatting the time information derived data corresponding to the object to be processed, the dimensional data corresponding to the object to be processed, and the one-hot data corresponding to the object to be processed to obtain at least one cold quantity feature tensor corresponding to the object to be processed.

[0042] Specifically, the historical cooling characteristic data corresponding to the object to be processed can be obtained from a database or an external data source. The object to be processed can be a building or a data center, etc., which is not limited here. The historical cooling characteristic data corresponding to the object to be processed can include environmental conditions and cooling usage at different time points, and then the data type of the historical cooling characteristic data can be checked. For date data, including but not limited to date or time, time information derivative data can be obtained through feature construction, wherein the feature construction includes but is not limited to time series analysis or time window division and other processing to obtain time-related laws and trends. For numerical feature data, including but not limited to quantifiable data such as temperature, humidity, and / or energy consumption, the numerical feature data can be converted into a unified dimension through standardization processing to eliminate the scale differences between different features. For discrete feature data, including but not limited to weather conditions or equipment status, etc., one-hot encoding processing can be performed to convert the discrete feature data into a form recognizable by the cooling determination model, and then the processed time information derivative data, dimensional data, and one-hot data can be formatted to obtain the cooling feature tensor corresponding to the object to be processed.

[0043] The historical cooling characteristic data corresponding to the object to be processed may be various historical record data related to the cooling demand and usage of the object to be processed in the past period of time, wherein the object to be processed includes but is not limited to buildings or data centers.

[0044] The date data corresponding to the object to be processed may be information related to time, including but not limited to a specific date or a timestamp, etc. The date data corresponding to the object to be processed may be used to analyze the change pattern of cooling demand over time.

[0045] The numerical characteristic data corresponding to the object to be processed may be characteristics that can be expressed by numerical quantification, including but not limited to temperature, humidity, and / or energy consumption, etc. The numerical characteristic data corresponding to the object to be processed may be used to characterize the physical quantity of cooling demand.

[0046] The discrete feature data corresponding to the object to be processed may be features that cannot be directly quantified by numerical values, including but not limited to weather conditions or equipment status, etc. The weather condition may be sunny or rainy, and the equipment status may be on or off, which are not limited here. The discrete feature data corresponding to the object to be processed may be converted into a form recognizable by the cooling capacity determination model through encoding.

[0047] The time information derived data corresponding to the object to be processed can be time-related information obtained by processing date data, including but not limited to time windows or seasonal changes. The time information derived data corresponding to the object to be processed can be used to capture the impact of time on cooling demand.

[0048] The dimensional data corresponding to the object to be processed may be numerical feature data after standardization, and the dimensional data corresponding to the object to be processed may be used to eliminate scale differences between different features.

[0049] The one-hot data corresponding to the object to be processed may be a result of one-hot encoding discrete feature data, wherein one-hot encoding may be a processing method for converting discrete features into a form recognizable by a model.

[0050] For example, in intelligent building management, we can collect numerical feature data such as the ambient temperature, humidity, and energy consumption of the building at different time points and the corresponding date data, as well as discrete feature data such as weather conditions and equipment operating status, and perform corresponding data type checks on the date data, numerical feature data, and discrete feature data: feature construction, standardization, and one-hot encoding processing, and then we can obtain the corresponding cooling feature tensor of the building.

[0051] According to the technical solution provided by the embodiment of the present disclosure, by obtaining the historical cold quantity characteristic data corresponding to the object to be processed from the database or external data source, the data type of the historical cold quantity characteristic data can be checked, and the time information derivative data can be obtained through the feature construction method to obtain time-related laws and trends. For numerical feature data, the numerical feature data can be converted into a unified dimension through standardization processing, and for discrete feature data, one-hot encoding processing can be performed to convert the discrete feature data into a form recognizable by the cold quantity determination model, and then the processed time information derivative data, dimensional data, and one-hot data can be formatted to obtain the cold quantity characteristic tensor corresponding to the object to be processed, thereby enhancing the flexibility and accuracy of data processing, ensuring the effective fusion of different types of data, improving the ability of the cold quantity determination model to capture time factors and complex features, improving the accuracy of cold quantity determination, improving the degree of automation of data processing, reducing manual intervention, and improving processing efficiency.

[0052] In some embodiments, in the cold quantity determination model completed through training, before performing attention mechanism processing on at least one cold quantity feature tensor corresponding to the object to be processed to obtain each cold quantity feature weight vector corresponding to the object to be processed, it also includes: obtaining multiple cold quantity feature tensors corresponding to the training object and the real cold quantity corresponding to each cold quantity feature tensor; inputting multiple cold quantity feature tensors corresponding to the training object into the cold quantity determination model for attention mechanism processing to obtain each cold quantity feature weight vector corresponding to the training object; normalizing each cold quantity feature weight vector corresponding to the training object to obtain each normalized feature weight value corresponding to the training object; weighting processing each cold quantity feature tensor corresponding to the training object and each normalized feature weight value corresponding to the training object to obtain each weighted feature vector corresponding to the training object; linearly transforming each weighted feature vector corresponding to the training object to obtain training cold quantity data corresponding to the training object; determining the mean square error loss corresponding to the cold quantity determination model according to the training cold quantity data corresponding to the training object and the real cold quantity corresponding to each cold quantity feature tensor; updating the parameters in the cold quantity determination model according to the mean square error loss corresponding to the cold quantity determination model through a cyclic iteration.

[0053] Specifically, multiple groups of cold feature tensors corresponding to the training object and the actual cold values ​​corresponding to each group of cold feature tensors can be obtained from the database as training data, the cold feature tensors can be input into the cold determination model, and the cold feature weight vectors corresponding to each feature tensor can be obtained through the attention mechanism. The cold feature weight vectors can be normalized to ensure that their sum is 1 to obtain the normalized feature weight values ​​corresponding to the training object. The cold feature tensors can be weighted by the normalized feature weight values ​​to obtain the weighted feature vectors corresponding to the training object, and then the weighted feature vectors can be linearly transformed by the activation function to obtain the cold data of the training object, that is, the training cold data corresponding to the training object, and then the mean square error loss between the predicted training cold data and the actual cold can be calculated, and the parameters in the cold determination model can be adjusted in a cyclic iterative manner based on the mean square error loss value until the performance of the cold determination model reaches the predetermined standard.

[0054] Among them, the cooling feature tensor corresponding to the training object can be a multidimensional array or matrix that describes the cooling-related properties of the training object, including but not limited to a set of environmental parameters such as temperature, humidity, and / or wind speed.

[0055] The actual cooling capacity corresponding to the cooling capacity feature tensor may be an actual cooling capacity value of the training object under specific conditions obtained through actual measurement or known standards.

[0056] The cold feature weight vector corresponding to the training object can be a vector obtained by processing the cold feature tensor through the attention mechanism. The cold feature weight vector corresponding to the training object can be used to characterize the importance or contribution of each feature in determining the cold.

[0057] The normalized feature weight value corresponding to the training object may be a value obtained by normalizing the cold feature weight vector, in order to ensure that the sum of the weights of all features is 1, which facilitates subsequent weighted processing.

[0058] The weighted feature vector corresponding to the training object can be a vector obtained by multiplying the cooling feature tensor and its corresponding normalized feature weight value. The weighted feature vector corresponding to the training object can be used to characterize the comprehensive influence of each feature in determining the cooling amount.

[0059] The mean square error loss corresponding to the cooling capacity determination model can be obtained by calculating the average of the squares of the differences between the training cooling capacity data obtained by the cooling capacity determination model and the actual cooling capacity. The mean square error loss corresponding to the cooling capacity determination model can be used to evaluate the performance of the cooling capacity determination model in cooling capacity determination, and the cooling capacity determination model parameters can be optimized by minimizing the loss.

[0060] For example, in an intelligent building air-conditioning system, the required cooling capacity can be obtained based on the cooling characteristic tensors such as the temperature, humidity, and / or occupancy density of the indoor environment through the trained cooling determination model, so that the operating parameters of the air-conditioning system can be automatically adjusted to achieve energy saving and consumption reduction.

[0061] According to the technical solution provided by the embodiment of the present disclosure, multiple groups of cold feature tensors corresponding to the training object and the actual cold values ​​corresponding to each group of cold feature tensors are obtained as training data through a database, the cold feature tensors can be input into the cold determination model, and the cold feature weight vectors corresponding to each feature tensor can be obtained through attention mechanism processing, the cold feature weight vectors can be processed by an exponential normalization function to obtain each normalized feature weight value corresponding to the training object, the cold feature tensors can be weighted by the normalized feature weight value to obtain each weighted feature vector corresponding to the training object, and then each weighted feature vector can be linearly transformed by an activation function to obtain the training cold data corresponding to the training object, and then the mean square error loss between the predicted training cold data and the actual cold can be calculated, and the parameters in the cold determination model can be adjusted in a cyclic iteration manner based on the mean square error loss value until the performance of the cold determination model reaches a predetermined standard, thereby enhancing the recognition ability of the cold determination model for key cold features, improving the accuracy of the cold determination result, and improving the generalization ability and stability of the cold determination model in practical applications.

[0062] In some embodiments, feature construction processing is performed on the date data corresponding to the object to be processed to obtain time information derivative data corresponding to the object to be processed, including: performing feature construction processing on the date data corresponding to the object to be processed to obtain construction feature values ​​corresponding to the date data; performing identification generation processing on the date data corresponding to the object to be processed to obtain identification data corresponding to the date data; performing feature integration processing on the construction feature values ​​corresponding to the date data and the identification data corresponding to the date data to obtain time information derivative data corresponding to the object to be processed.

[0063] Specifically, feature engineering methods can be used to construct features for date data in the object to be processed, and constructed feature values ​​such as "whether it is a weekend" or "quarter number" can be obtained. There is no limitation here. Hash functions, timestamp conversion, or custom encoding rules and other identification processing methods can be used to generate unique or meaningful identification data for the date data corresponding to the object to be processed, such as "specific holiday identification code" or "timestamp sequence". The obtained constructed feature values ​​and identification data can then be integrated, that is, the constructed feature values ​​and identification data are merged into a new data set as the time information derived data corresponding to the object to be processed.

[0064] Among them, the constructed feature values ​​corresponding to the date data can be new feature values ​​extracted or calculated from the original date data through algorithms or rules. The constructed feature values ​​corresponding to the date data can be used to characterize time characteristics, such as "month" or "whether it is a holiday", etc., which are not limited here.

[0065] The identification data corresponding to the date data can be a unique identifier or a code with a specified meaning generated for the date data through a hash function or timestamp conversion, etc., which is not limited here. The date data can be in the form of a digital sequence or a character string containing specified information, etc., which is not limited here. The identification data corresponding to the date data can be used to identify and distinguish different time points or time periods during data processing and analysis.

[0066] For example, in the user behavior analysis of e-commerce platforms, for the user's purchase records, the purchase date can be processed as date data, and through feature construction processing, constructed feature values ​​such as "purchase month" and "whether it is a promotion period" can be extracted. Then, through timestamp conversion or custom encoding rules, a unique "purchase time identification code" can be generated for each purchase record. The constructed feature value and identification data can be integrated and processed as derived data of the time information of the user's purchase behavior.

[0067] According to the technical solution provided by the embodiment of the present disclosure, feature construction processing is performed on the date data in the object to be processed through a feature engineering method to obtain a constructed feature value, and identification processing methods such as hash functions, timestamp conversion, or custom encoding rules can be used to generate unique or meaningful identification data for the date data corresponding to the object to be processed. Then, the constructed feature value and the identification data can be merged into a new data set as time information derived data corresponding to the object to be processed, thereby enhancing the richness and expressiveness of the data, improving the accuracy of data analysis and prediction, and enhancing the processing capability of the cooling capacity determination model for time-sensitive problems.

[0068] In some embodiments, discrete feature data corresponding to the object to be processed are one-hot encoded to obtain the one-hot data corresponding to the object to be processed, including: performing index allocation processing on the discrete feature data corresponding to the object to be processed to obtain a mapping relationship corresponding to the discrete feature data; performing one-hot encoding processing on the mapping relationship corresponding to the discrete feature data to obtain the one-hot data corresponding to the object to be processed.

[0069] Specifically, an index number can be assigned to each discrete value in the discrete feature data corresponding to the object to be processed, wherein the discrete feature data corresponding to the object to be processed includes but is not limited to gender, country, and / or product category, etc., to obtain a mapping relationship between discrete feature data and index numbers, and then each discrete feature value can be one-hot encoded based on the mapping relationship between discrete feature data and index numbers, that is, a vector containing only 0 and 1 is created, and the length of the vector is equal to the total number of discrete feature values. The position corresponding to the discrete feature value index number in the vector is 1, and the remaining positions are 0, so that the one-hot data corresponding to the object to be processed can be obtained.

[0070] Among them, the mapping relationship corresponding to the discrete feature data can be a corresponding relationship that maps discrete feature values ​​to unique index numbers. The mapping relationship corresponding to the discrete feature data can be obtained through a hash table, an enumeration type, or a key-value pair structure. The mapping relationship corresponding to the discrete feature data can be used to find the index number corresponding to each discrete feature value, and then perform feature conversion of one-hot encoding.

[0071] For example, in the construction of user portraits in the recommendation system, the user's occupation can be a discrete feature, and all appearing occupational categories can be assigned indexes, such as "teacher" is assigned index 1, "doctor" is assigned index 2, and so on. For each user's occupational information, one-hot encoding can be performed according to the index assignment result. For example, if the total number of occupational categories is N, then the occupational information of each user can be converted into a one-hot vector of length N, where the position corresponding to the occupational index number is 1, and the remaining positions are one-hot data 0.

[0072] According to the technical solution provided by the embodiment of the present disclosure, by assigning an index number to each discrete value in the discrete feature data corresponding to the object to be processed, a mapping relationship between the discrete feature data and the index number is obtained, and then each discrete feature value can be one-hot encoded based on the mapping relationship between the discrete feature data and the index number, so that the one-hot data corresponding to the object to be processed can be obtained, thereby enhancing the readability and interpretability of the data, improving the processing ability of the cold quantity determination model for discrete features, avoiding the deviation introduced by direct numerical comparison between discrete feature values, and improving the generalization ability of the cold quantity determination model.

[0073] In some embodiments, the numerical feature data corresponding to the object to be processed is standardized to obtain the dimensional data corresponding to the object to be processed, including: calculating the feature mean and standard deviation of the numerical feature data corresponding to the object to be processed to obtain the mean value corresponding to the numerical feature data and the standard deviation value corresponding to the numerical feature data; converting the mean value corresponding to the numerical feature data and the standard deviation value corresponding to the numerical feature data through a preset standardized formula to obtain the dimensional data corresponding to the object to be processed.

[0074] Specifically, the global mean and standard deviation can be calculated for the numerical feature data corresponding to the object to be processed, wherein the numerical feature data corresponding to the object to be processed includes but is not limited to age, income, and / or score, etc. The mean can be obtained by traversing the data set, accumulating each numerical feature value and dividing it by the total, and the standard deviation can be calculated by the variance formula, and then a preset standardization formula, including but not limited to the standard score (Z-score) standardization formula, etc., can be used to subtract the mean from each numerical feature value and divide it by the standard deviation, so as to obtain dimensionless standardized data, that is, the dimensional data corresponding to the object to be processed.

[0075] Among them, the mean corresponding to the numerical feature data can be the average value of the numerical feature in all samples, and the mean corresponding to the numerical feature data can be used to characterize the central trend of the numerical feature; the standard deviation value corresponding to the numerical feature data can refer to a measure of the degree of deviation of the numerical feature value from its mean, and the larger the standard deviation, the more dispersed the data distribution; the preset normalization formula can be a mathematical formula accepted and applied in data preprocessing, such as the Z-score normalization formula, etc. The preset normalization formula can be used to eliminate the influence of different dimensions and magnitudes.

[0076] For example, in the field of financial risk assessment, numerical features can be the borrower's age and income. Since the dimensions of age and income are time units and monetary units respectively, these two features can be standardized, and the mean and standard deviation of the age and income of all borrowers can be calculated. Then, the age and income of each borrower can be converted into standardized values ​​through the Z-score standardization formula.

[0077] According to the technical solution provided by the embodiment of the present disclosure, by calculating the global mean and standard deviation of the numerical feature data corresponding to the object to be processed, the mean can be obtained by traversing the data set, accumulating each numerical feature value and dividing it by the total number, and the standard deviation can be calculated by the variance formula, and then a preset normalization formula can be used to subtract the mean from each numerical feature value and divide it by the standard deviation, so as to obtain the dimensional data corresponding to the object to be processed, thereby enhancing the comparability of the data, improving the convergence speed and stability of the cold quantity determination model, improving the prediction accuracy and generalization ability of the cold quantity determination model, and reducing the deviation caused by data scale differences.

[0078] In some embodiments, the method for determining the cold quantity also includes: obtaining a training data set, the training data set including multiple cold quantity feature tensors corresponding to the training object and the actual cold quantity corresponding to each cold quantity feature tensor; dividing the training data set to obtain multiple training data subsets; performing multi-fold cross-validation processing on the cold quantity determination model according to the training data subsets to obtain multiple cold quantity determination candidate models and the mean square error value corresponding to each cold quantity determination candidate model; determining the cold quantity determination target model according to the multiple cold quantity determination candidate models and the mean square error value corresponding to each cold quantity determination candidate model.

[0079] Specifically, a training data set can be obtained through a database or a data center. The training data set can include multiple cooling feature tensors corresponding to the training object and the actual cooling corresponding to each cooling feature tensor. The training data set can be divided into multiple non-overlapping training data subsets through a partitioning strategy, wherein the partitioning strategy includes but is not limited to random partitioning or hierarchical partitioning, etc. A multi-fold cross-validation method can be used to train and verify the cooling determination model by using different training data subsets as training sets each time and the remaining training data subsets as verification sets, so that multiple cooling determination candidate models based on different training-verification combinations can be generated. For each cooling determination candidate model, the mean square error value between its target cooling on the verification set and the actual cooling is calculated, and then the cooling determination candidate model with the smallest error can be selected as the cooling determination target model based on the obtained mean square error value.

[0080] Among them, the training data set can be a data set containing a large number of training object cooling feature tensors and their corresponding real cooling values, and the training data set can be used to train and verify the cooling determination model; the training data subset can be a subset obtained by dividing the training data set, and the training data subset can be used as a training set and a verification set in turn during the multi-fold cross-validation process; the cooling determination candidate model can be a plurality of cooling prediction models obtained by training based on different training data subsets during the multi-fold cross-validation process; the mean square error value corresponding to the cooling determination candidate model can be obtained by calculating the difference between the target cooling and the real cooling of the cooling determination candidate model on the verification set, and the mean square error value can be used to evaluate the prediction performance of the cooling determination candidate model.

[0081] For example, in the field of air conditioning system design, a large number of cooling feature tensors of buildings under different environmental conditions and their corresponding real cooling data can be obtained as training data sets. Then, the training data sets are divided into multiple subsets through partitioning, where the cooling feature tensors include but are not limited to outdoor temperature, humidity, building type, and / or number of users, etc., and multiple cooling determination candidate models can be trained and verified through a multi-fold cross-validation method, and the cooling determination candidate model with the smallest mean square error is selected as the target model for predicting the cooling load demand of new buildings or under new conditions.

[0082] According to the technical solution provided by the embodiment of the present disclosure, a training data set is obtained through a database or a data center. The training data set may include multiple cold feature tensors corresponding to the training object and the actual cold corresponding to each cold feature tensor. Then, the training data set may be divided into multiple non-overlapping training data subsets through a partitioning strategy. A multi-fold cross-validation method may be used to train and verify the cold determination model by using different training data subsets as training sets each time and the remaining training data subsets as verification sets, so that multiple cold determination candidate models based on different training-verification combinations may be generated. For each cold determination candidate model, the mean square error value between its target cold on the verification set and the actual cold is calculated. Then, according to the obtained mean square error value, the cold determination candidate model with the smallest error may be selected as the cold determination target model, thereby enhancing the generalization ability of the cold determination model. Through multi-fold cross-validation, the cold determination model is trained and verified on multiple different data subsets, reducing the risk of overfitting, improving the stability of the cold determination model, and improving the accuracy of cold prediction. By selecting the model with the smallest mean square error, the closeness of the processing result to the actual cold is ensured.

[0083] All the above optional technical solutions can be arbitrarily combined to form optional embodiments of the present disclosure, and will not be described in detail here.

[0084] The following are embodiments of the device disclosed herein, which can be used to execute the method embodiments disclosed herein. For details not disclosed in the device embodiments disclosed herein, please refer to the method embodiments disclosed herein.

[0085] Figure 3 Schematic diagram of a cooling capacity determination device provided by an embodiment of the present disclosure. Figure 3 As shown, the device for determining the cooling capacity includes: The first processing module 301 is used to perform attention mechanism processing on at least one cold feature tensor corresponding to the object to be processed through the trained cold determination model to obtain each cold feature weight vector corresponding to the object to be processed; The second processing module 302 is used to normalize each cooling feature weight vector corresponding to the object to be processed to obtain each normalized feature weight value corresponding to the object to be processed; The third processing module 303 is used to perform weighted processing on each cold quantity feature tensor corresponding to the object to be processed and each normalized feature weight value corresponding to the object to be processed to obtain each weighted feature vector corresponding to the object to be processed; The fourth processing module 304 is used to perform linear transformation on each weighted feature vector corresponding to the object to be processed, obtain target cooling data corresponding to the object to be processed, and send the target cooling data to the terminal device for cooling release control.

[0086] According to the technical solution provided by the embodiment of the present disclosure, by inputting at least one cold quantity feature tensor corresponding to the object to be processed into the trained cold quantity determination model, the cold quantity feature tensor can be processed by the cold quantity determination model through the attention mechanism and converted into a cold quantity feature weight vector corresponding to the object to be processed, and then each cold quantity feature weight vector corresponding to the object to be processed can be normalized by a normalized exponential function to obtain each normalized feature weight value corresponding to the object to be processed, each cold quantity feature tensor corresponding to the object to be processed and each normalized feature weight value corresponding to the object to be processed can be weighted, and the weight value of the cold quantity feature tensor can be adjusted to obtain the object to be processed with the reallocated weight. The weighted feature vectors corresponding to the object can be linearly transformed through the activation function to obtain the target cooling data corresponding to the object to be processed. The target cooling data can be used to guide the corresponding terminal device to release cooling in a specified time period. The cooling determination model completed through training can perform refined analysis and processing on the specific cooling features of the object to be processed. Through the attention mechanism, the cooling determination model can identify and weight key cooling features, thereby improving the accuracy and reliability of cooling determination and improving processing efficiency. The flexibility of the cooling determination method is enhanced by combining and weighting different cooling features.

[0087] In some embodiments, the above-mentioned cold capacity determination device is also used to obtain at least one historical cold capacity characteristic data corresponding to the object to be processed; perform data type check processing on each historical cold capacity characteristic data corresponding to the object to be processed to obtain date type data corresponding to the object to be processed, numerical characteristic data corresponding to the object to be processed, and discrete characteristic data corresponding to the object to be processed; perform feature construction processing on the date type data corresponding to the object to be processed to obtain time information derived data corresponding to the object to be processed; perform standardization processing on the numerical characteristic data corresponding to the object to be processed to obtain dimensional data corresponding to the object to be processed; perform one-hot encoding processing on the discrete characteristic data corresponding to the object to be processed to obtain one-hot data corresponding to the object to be processed; format the time information derived data corresponding to the object to be processed, the dimensional data corresponding to the object to be processed, and the one-hot data corresponding to the object to be processed to obtain at least one cold capacity characteristic tensor corresponding to the object to be processed.

[0088] In some embodiments, the above-mentioned cold quantity determination device is also used to obtain multiple cold quantity feature tensors corresponding to the training object and the actual cold quantity corresponding to each cold quantity feature tensor; input the multiple cold quantity feature tensors corresponding to the training object into the cold quantity determination model for attention mechanism processing to obtain each cold quantity feature weight vector corresponding to the training object; normalize each cold quantity feature weight vector corresponding to the training object to obtain each normalized feature weight value corresponding to the training object; weightedly process each cold quantity feature tensor corresponding to the training object and each normalized feature weight value corresponding to the training object to obtain each weighted feature vector corresponding to the training object; linearly transform each weighted feature vector corresponding to the training object to obtain training cold quantity data corresponding to the training object; determine the mean square error loss corresponding to the cold quantity determination model according to the training cold quantity data corresponding to the training object and the actual cold quantity corresponding to each cold quantity feature tensor; update the parameters in the cold quantity determination model according to the mean square error loss corresponding to the cold quantity determination model through a cyclic iteration.

[0089] In some embodiments, feature construction processing is performed on the date data corresponding to the object to be processed to obtain time information derivative data corresponding to the object to be processed. Specifically, feature construction processing is performed on the date data corresponding to the object to be processed to obtain construction feature values ​​corresponding to the date data; identification generation processing is performed on the date data corresponding to the object to be processed to obtain identification data corresponding to the date data; feature integration processing is performed on the construction feature values ​​corresponding to the date data and the identification data corresponding to the date data to obtain time information derivative data corresponding to the object to be processed.

[0090] In some embodiments, one-hot encoding is performed on discrete feature data corresponding to the object to be processed to obtain the one-hot data corresponding to the object to be processed, which is specifically used to perform index allocation processing on the discrete feature data corresponding to the object to be processed to obtain a mapping relationship corresponding to the discrete feature data; one-hot encoding is performed on the mapping relationship corresponding to the discrete feature data to obtain the one-hot data corresponding to the object to be processed.

[0091] In some embodiments, the numerical feature data corresponding to the object to be processed are standardized to obtain the dimensional data corresponding to the object to be processed, which is specifically used to calculate the feature mean and standard deviation of the numerical feature data corresponding to the object to be processed to obtain the mean value corresponding to the numerical feature data and the standard deviation value corresponding to the numerical feature data; through a preset standardized formula, the mean value corresponding to the numerical feature data and the standard deviation value corresponding to the numerical feature data are converted to obtain the dimensional data corresponding to the object to be processed.

[0092] In some embodiments, the above-mentioned cold capacity determination device is also used to obtain a training data set, the training data set includes multiple cold capacity feature tensors corresponding to the training object and the actual cold capacity corresponding to each cold capacity feature tensor; divide the training data set to obtain multiple training data subsets; perform multi-fold cross-validation processing on the cold capacity determination model according to the training data subsets to obtain multiple cold capacity determination candidate models and the mean square error value corresponding to each cold capacity determination candidate model; determine the cold capacity determination target model according to the multiple cold capacity determination candidate models and the mean square error value corresponding to each cold capacity determination candidate model.

[0093] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present disclosure.

[0094] Figure 4 Schematic diagram of an electronic device 4 provided in an embodiment of the present disclosure. Figure 4 As shown, the electronic device 4 of this embodiment includes: a processor 401, a memory 402, and a computer program 403 stored in the memory 402 and executable on the processor 401. When the processor 401 executes the computer program 403, the steps in the above-mentioned method embodiments are implemented. Alternatively, when the processor 401 executes the computer program 403, the functions of the modules / units in the above-mentioned device embodiments are implemented.

[0095] The electronic device 4 may be a desktop computer, a notebook, a PDA, a cloud server, or other electronic device. The electronic device 4 may include, but is not limited to, a processor 401 and a memory 402. Those skilled in the art will appreciate that Figure 4 The electronic device 4 is merely an example and does not limit the electronic device 4 , and may include more or less components than those shown in the figure, or different components.

[0096] The processor 401 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0097] The memory 402 may be an internal storage unit of the electronic device 4, for example, a hard disk or memory of the electronic device 4. The memory 402 may also be an external storage device of the electronic device 4, for example, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device 4. The memory 402 may also include both an internal storage unit and an external storage device of the electronic device 4. The memory 402 is used to store computer programs and other programs and data required by the electronic device.

[0098] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units.

[0099] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium (e.g., a computer-readable storage medium). Based on this understanding, the present disclosure implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. The computer program may include computer program code, and the computer program code may be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable storage medium may include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc.

[0100] The above embodiments are only used to illustrate the technical solutions of the present disclosure, rather than to limit them. Although the present disclosure has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present disclosure, and should all be included in the protection scope of the present disclosure.

Claims

1. A method for determining cooling capacity, characterized in that: include: Attention mechanism processing is performed on at least one cold quantity feature tensor corresponding to the object to be processed by the trained cold quantity determination model to obtain each cold quantity feature weight vector corresponding to the object to be processed; Normalizing each cooling feature weight vector corresponding to the object to be processed to obtain each normalized feature weight value corresponding to the object to be processed; Performing weighted processing on each cold quantity feature tensor corresponding to the object to be processed and each normalized feature weight value corresponding to the object to be processed to obtain each weighted feature vector corresponding to the object to be processed; Each weighted feature vector corresponding to the object to be processed is linearly transformed to obtain target cooling data corresponding to the object to be processed, and the target cooling data is sent to the terminal device for cooling release control.

2. The method for determining cooling capacity according to claim 1, characterized in that: Before the cold quantity determination model completed through training performs attention mechanism processing on at least one cold quantity feature tensor corresponding to the object to be processed to obtain each cold quantity feature weight vector corresponding to the object to be processed, the method further includes: Acquire at least one historical cooling characteristic data corresponding to the object to be processed; Performing data type checking processing on each historical cooling characteristic data corresponding to the object to be processed, obtaining date data corresponding to the object to be processed, numerical characteristic data corresponding to the object to be processed, and discrete characteristic data corresponding to the object to be processed; Performing feature construction processing on the date data corresponding to the object to be processed to obtain time information derived data corresponding to the object to be processed; Standardizing the numerical feature data corresponding to the object to be processed to obtain the dimensional data corresponding to the object to be processed; Performing one-hot encoding processing on the discrete feature data corresponding to the object to be processed to obtain one-hot data corresponding to the object to be processed; The time information derived data corresponding to the object to be processed, the dimensional data corresponding to the object to be processed, and the one-hot data corresponding to the object to be processed are formatted to obtain at least one cold feature tensor corresponding to the object to be processed.

3. The method for determining cooling capacity according to claim 1, characterized in that: Before the cold quantity determination model completed through training performs attention mechanism processing on at least one cold quantity feature tensor corresponding to the object to be processed to obtain each cold quantity feature weight vector corresponding to the object to be processed, the method further includes: Acquire multiple cooling feature tensors corresponding to the training object and the actual cooling corresponding to each of the cooling feature tensors; Inputting a plurality of cold feature tensors corresponding to the training object into the cold determination model for attention mechanism processing to obtain each cold feature weight vector corresponding to the training object; Normalizing each cooling feature weight vector corresponding to the training object to obtain each normalized feature weight value corresponding to the training object; Performing weighted processing on each cold feature tensor corresponding to the training object and each normalized feature weight value corresponding to the training object to obtain each weighted feature vector corresponding to the training object; Performing linear transformation on each weighted feature vector corresponding to the training object to obtain training cooling data corresponding to the training object; Determine the mean square error loss corresponding to the cooling capacity determination model according to the training cooling capacity data corresponding to the training object and the real cooling capacity corresponding to each cooling capacity feature tensor; The parameters in the cooling capacity determination model are updated according to the mean square error loss corresponding to the cooling capacity determination model in a loop iteration manner.

4. The method for determining cooling capacity according to claim 2, characterized in that: The performing feature construction processing on the date data corresponding to the object to be processed to obtain time information derived data corresponding to the object to be processed includes: Performing feature construction processing on the date data corresponding to the object to be processed to obtain a construction feature value corresponding to the date data; Performing identification generation processing on the date data corresponding to the object to be processed to obtain identification data corresponding to the date data; Feature integration processing is performed on the construction feature value corresponding to the date data and the identification data corresponding to the date data to obtain time information derivative data corresponding to the object to be processed.

5. The method for determining cooling capacity according to claim 2, characterized in that: The performing one-hot encoding processing on the discrete feature data corresponding to the object to be processed to obtain the one-hot data corresponding to the object to be processed includes: Performing index allocation processing on the discrete feature data corresponding to the object to be processed to obtain a mapping relationship corresponding to the discrete feature data; The mapping relationship corresponding to the discrete feature data is subjected to one-hot encoding processing to obtain one-hot data corresponding to the object to be processed.

6. The method for determining cooling capacity according to claim 2, characterized in that: The step of performing standardization processing on the numerical feature data corresponding to the object to be processed to obtain the dimensional data corresponding to the object to be processed includes: Calculating the characteristic mean and standard deviation of the numerical characteristic data corresponding to the object to be processed to obtain the mean value corresponding to the numerical characteristic data and the standard deviation value corresponding to the numerical characteristic data; By using a preset standardization formula, the mean value corresponding to the numerical feature data and the standard deviation value corresponding to the numerical feature data are converted to obtain the dimensional data corresponding to the object to be processed.

7. The method for determining cooling capacity according to claim 3, characterized in that: The method further comprises: Acquire a training data set, wherein the training data set includes a plurality of cooling feature tensors corresponding to the training object and real cooling corresponding to each of the cooling feature tensors; Dividing the training data set to obtain multiple training data subsets; Performing a multi-fold cross validation process on the cooling capacity determination model according to the training data subset to obtain a plurality of cooling capacity determination candidate models and mean square error values ​​corresponding to each of the cooling capacity determination candidate models; A cooling capacity determination target model is determined according to the plurality of cooling capacity determination candidate models and the mean square error values ​​corresponding to the respective cooling capacity determination candidate models.

8. A device for determining cooling capacity, characterized in that: include: A first processing module is used to perform attention mechanism processing on at least one cold feature tensor corresponding to the object to be processed through the trained cold determination model to obtain each cold feature weight vector corresponding to the object to be processed; A second processing module is used to normalize each cooling feature weight vector corresponding to the object to be processed to obtain each normalized feature weight value corresponding to the object to be processed; A third processing module is used to perform weighted processing on each cold feature tensor corresponding to the object to be processed and each normalized feature weight value corresponding to the object to be processed to obtain each weighted feature vector corresponding to the object to be processed; The fourth processing module is used to perform linear transformation on each weighted feature vector corresponding to the object to be processed to obtain target cooling data corresponding to the object to be processed, and send the target cooling data to the terminal device for cooling release control.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images