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

Abstract

The invention relates to the technical field of cold station intelligent prediction, and provides a cold capacity determination method and device, electronic equipment and a readable storage medium. The method comprises the following steps: inputting at least one cooling capacity one-heat vector into a trained cooling capacity determination model, and performing linear transformation processing on the at least one cooling capacity one-heat vector to obtain each cooling capacity feature vector; all the cooling capacity feature vectors are processed through a gating residual network, and all regulation and control vectors are obtained; performing splicing processing on each cooling capacity feature vector to obtain a spliced vector; processing the splicing vector through a gated residual network to obtain a weight vector; performing exponential normalization processing on the weight vector to obtain each variable selection weight value; according to a preset mapping relation, performing splicing processing on each regulation and control vector and each variable selection weight value to obtain a target feature vector; and processing the target feature vector through a full connection layer to obtain target cooling capacity data.

Description

Technical Field

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

[0002] As the core facility for regulating indoor temperature and humidity, the operating efficiency and control accuracy of the cooling station system are directly related to indoor comfort and energy utilization. The traditional cooling station control system adopts experience-based fixed value control or feedback control, which is difficult to adapt to environmental changes, slow response, and high energy consumption. In order to improve control accuracy and energy-saving effects, modern cooling stations have begun to introduce computer-aided control and automation technology. Currently, environmental parameters and equipment operation data are usually collected through data collection, and machine learning models are trained through historical data. The real-time collected environmental parameters are input into the model to predict the cooling demand in the future. However, since traditional machine learning algorithms cannot understand the complex nonlinear relationship between multi-source heterogeneous data, the prediction error of cooling demand is large, and the model cannot automatically identify the key features that have the greatest impact on the prediction results, resulting in missing or redundant feature data.

[0003] It can be seen that the existing technology has problems such as large errors in determining the cooling capacity and data redundancy or omission due to insufficient understanding of the complex nonlinear relationship between multi-source heterogeneous data and the inability to identify the key features that affect the prediction results. 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, so as to solve the problems in the prior art of large cooling capacity determination errors and data redundancy or omissions due to insufficient understanding of the complex nonlinear relationship between multi-source heterogeneous data and the inability to identify key features that affect the prediction results.

[0005] According to a first aspect of an embodiment of the present disclosure, a method for determining cold quantity is provided, including: inputting at least one cold quantity unique hot vector into a trained cold quantity determination model, performing linear transformation processing on at least one cold quantity unique hot vector, and obtaining a cold quantity feature vector corresponding to each cold quantity unique hot vector; processing the cold quantity feature vector corresponding to each cold quantity unique hot vector through a gated residual network, and obtaining a control vector corresponding to each cold quantity feature vector; performing splicing processing on the cold quantity feature vector corresponding to each cold quantity unique hot vector, and obtaining a splicing vector corresponding to the cold quantity feature vector; processing the splicing vector corresponding to the cold quantity feature vector through a gated residual network, and obtaining a weight vector corresponding to the cold quantity feature vector; performing exponential normalization processing on the weight vector corresponding to the cold quantity feature vector, and obtaining a variable selection weight value corresponding to each cold quantity feature vector; splicing processing on the control vector corresponding to each cold quantity feature vector and the variable selection weight value corresponding to each cold quantity feature vector according to a preset mapping relationship, and obtaining a target feature vector corresponding to the cold quantity feature vector; processing the target feature vector corresponding to the cold quantity feature vector through a fully connected layer, and obtaining a target cold quantity data.

[0006] According to a second aspect of an embodiment of the present disclosure, a device for determining cooling capacity is provided, including: a first processing module, for inputting at least one cooling capacity unique hot vector into a trained cooling capacity determination model, performing linear transformation processing on at least one cooling capacity unique hot vector, and obtaining a cooling capacity feature vector corresponding to each cooling capacity unique hot vector; a second processing module, for processing the cooling capacity feature vector corresponding to each cooling capacity unique hot vector through a gated residual network, and obtaining a control vector corresponding to each cooling capacity feature vector; a third processing module, for performing splicing processing on the cooling capacity feature vectors corresponding to each cooling capacity unique hot vector, and obtaining a splicing vector corresponding to the cooling capacity feature vector; and a fourth processing module, for performing splicing processing on the cooling capacity feature vectors corresponding to the cooling capacity unique hot vector, and obtaining a splicing vector corresponding to the cooling capacity feature vector; and a fourth processing module, for performing splicing processing on the cooling capacity feature vectors corresponding to the cooling capacity unique hot vector, and obtaining a splicing vector corresponding to the cooling capacity feature vector. The concatenated vector corresponding to the cold quantity feature vector is processed through a gated residual network to obtain a weight vector corresponding to the cold quantity feature vector; the fifth processing module is used to perform exponential normalization on the weight vector corresponding to the cold quantity feature vector to obtain a variable selection weight value corresponding to each cold quantity feature vector; the determination module is used to perform concatenation processing on the control vector corresponding to each cold quantity feature vector and the variable selection weight value corresponding to each cold quantity feature vector according to a preset mapping relationship to obtain a target feature vector corresponding to the cold quantity feature vector; the sixth processing module is used to process the target feature vector corresponding to the cold quantity feature vector through a fully connected layer to obtain target cold quantity data.

[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 embodiments of the present disclosure have the following beneficial effects: through the trained cold quantity determination model, at least one cold quantity unique hot vector is linearly transformed and converted into a cold quantity feature vector, the cold quantity feature vector corresponding to the cold quantity unique hot vector can be processed by a gated residual network to obtain a control vector corresponding to the cold quantity feature vector, and then the cold quantity feature vectors corresponding to each cold quantity unique hot vector can be spliced ​​to obtain a spliced ​​vector corresponding to the cold quantity feature vector, and the spliced ​​vector corresponding to the cold quantity feature vector can be processed by a gated residual network to obtain a weight vector corresponding to the cold quantity feature vector, and the weight vector corresponding to the cold quantity feature vector contains weight values ​​corresponding to each control vector, and then the cold quantity feature vector can be normalized by a normalized exponential function. The weight vector corresponding to the vector is exponentially normalized to obtain the variable selection weight value corresponding to each cooling feature vector. According to the preset mapping relationship, the control vector and the variable selection weight value corresponding to the cooling feature vector are correspondingly multiplied and spliced ​​to obtain the target feature vector corresponding to the cooling feature vector. The target feature vector corresponding to the cooling feature vector can be processed by the fully connected layer to obtain the target cooling data. In this way, the control ability of the cooling feature vector and the accuracy of weight allocation of the cooling determination model are enhanced through the gated residual network. The accuracy and reliability of the target cooling data are improved through the processing of the fully connected layer, which provides strong technical support for the accurate prediction and control of cooling demand and improves the efficiency and accuracy of cooling determination. 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 flow chart of a method for determining cooling capacity provided by an embodiment of the present disclosure;

[0012] Figure 2 is a schematic diagram of the structure of a gated residual network provided by an embodiment of the present disclosure;

[0013] Figure 3 is a structural schematic diagram of a cooling capacity determination device provided by an embodiment of the present disclosure;

[0014] Figure 4It is a structural schematic diagram of an electronic device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0015] 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.

[0016] 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.

[0017] Figure 1 It is a flow chart of a method for determining cooling capacity provided in an embodiment of the present disclosure. Figure 1 The method for determining the cooling capacity can be executed by the server. Figure 1 As shown, the method for determining the cooling capacity includes:

[0018] Step 101: input at least one cooling unique hot vector into a trained cooling determination model, perform linear transformation on at least one cooling unique hot vector, and obtain cooling feature vectors corresponding to each cooling unique hot vector.

[0019] Specifically, at least one cold quantity unique hot vector can be input into a trained cold quantity determination model. The cold quantity determination model can be a convolutional neural network, a long short-term memory network, or a Transformer model, etc., which is not limited here. The cold quantity determination model can be used to perform linear transformation processing on the input cold quantity unique hot vector. In this embodiment, the cold quantity determination model can be used to extract feature information from the cold quantity unique hot vector to obtain the cold quantity feature vector corresponding to each cold quantity unique hot vector, thereby performing linear transformation processing on the cold quantity unique hot vector through the trained cold quantity determination model, thereby improving the accuracy of cold quantity determination, improving processing efficiency, and enhancing the scalability of the cold quantity determination model.

[0020] Among them, the cooling capacity unique hot vector can be a vector representation obtained by unique hot encoding the environmental monitoring data after each cleaning and the operation data of the cold station equipment after each cleaning, which can be used to characterize different cooling capacity types or states. In the cooling capacity unique hot vector, the value of one position is 1, and the values ​​of the other positions are 0. The position with a value of 1 corresponds to a specific cooling capacity type or state, and the position of 0 can be used to characterize the state or category that does not belong to the position represented; the cooling capacity determination model can be a machine learning model trained by training sample data, which can be used to determine the target cooling capacity data; the cooling capacity feature vector corresponding to the cooling capacity unique hot vector can be a feature vector obtained by linearly transforming the cooling capacity unique hot vector through the cooling capacity determination model. The cooling capacity feature vector corresponding to the cooling capacity unique hot vector can contain time series information, category information, semantic information, and / or context information, etc., which is not limited here.

[0021] For example, the cold quantity one-hot vector corresponding to the cold quantity data processed by one-hot encoding can be input into the cold quantity determination model trained based on the Transformer model, and the cold quantity one-hot vector can be linearly transformed by the cold quantity determination model to obtain the cold quantity feature vector corresponding to the cold quantity one-hot vector.

[0022] Step 102: Process the cooling feature vectors corresponding to each cooling unique hot vector through a gated residual network to obtain a control vector corresponding to each cooling feature vector.

[0023] Specifically, the residual connection structure of the gated residual network can be used to retain information from previous levels while learning new features, and the cold quantity feature vector corresponding to the one-hot vector can be screened and adjusted through the gating mechanism, so that the gated residual network can screen out features that have a greater impact on the determination process of the target cold quantity data. The processed cold quantity feature vector can be mapped to a control vector corresponding to the cold quantity feature vector, thereby improving the expression and learning capabilities of the cold quantity determination model. Through the combination of the residual connection and the gating mechanism in the gated residual network, the gradient vanishing problem in the deep network is eliminated, the training efficiency and stability of the cold quantity determination model are improved, and the accuracy and efficiency of the control are improved by obtaining the control vector corresponding to the cold quantity feature vector.

[0024] Among them, the gated residual network can be a deep neural network that combines residual connections and gating mechanisms, wherein the residual connections can skip levels through the network and directly transmit information to deeper layers. The gating mechanism can filter and adjust information through gating units, and can be used to support the network to focus on features that have a greater impact on the results. The gating mechanism can be composed of a gated linear unit and an activation function. The activation function can be a Sigmoid function, a Tanh function, or a ReLU function, etc., which is not limited here.

[0025] For example, the cold feature vectors corresponding to the cold unique hot vector can be screened through the gating mechanism in the gated residual network to screen out 16 cold feature vectors that have a large impact on generating target cold data, and then combined with the residual network, 8 cold feature vectors that have a large impact on generating target cold data can be selected from the 16 cold feature vectors.

[0026] Step 103, performing concatenation processing on the cold quantity feature vectors corresponding to each cold quantity unique heat vector to obtain a concatenated vector corresponding to the cold quantity feature vector.

[0027] Specifically, according to a preset order, the cold quantity feature vectors corresponding to each cold quantity unique hot vector can be spliced ​​in length in a same-dimensional manner to obtain a spliced ​​vector corresponding to the cold quantity feature vector, wherein the preset order includes but is not limited to the logical order of features or the importance ranking of features, etc. The spliced ​​vector corresponding to the cold quantity feature vector can be used to characterize feature data that affects the determination of target cold quantity data, including but not limited to environmental monitoring data and / or cold station equipment operation data, etc., thereby integrating the data of each cold quantity feature through the spliced ​​vector corresponding to the cold quantity feature vector, thereby improving the information integration effect of the cold quantity determination model, and improving the performance of the cold quantity determination model by integrating feature information, improving the processing efficiency of feature engineering, and improving the quality of feature information.

[0028] Among them, the splicing vector corresponding to the cold quantity feature vector can be a long vector spliced ​​by multiple cold quantity one-hot vectors, which can contain information of all cold quantity one-hot vectors. The length of the splicing vector corresponding to the cold quantity feature vector is equal to the sum of the lengths of all cold quantity one-hot vectors.

[0029] For example, the cold feature vectors corresponding to the 16 cold unique hot vectors may be sorted according to their importance, and the lengths of the cold feature vectors corresponding to the 16 cold unique hot vectors may be concatenated in the same dimension to obtain a concatenated vector corresponding to the cold feature vector.

[0030] Step 104, processing the concatenated vector corresponding to the cold feature vector through a gated residual network to obtain a weight vector corresponding to the cold feature vector.

[0031] Specifically, the concatenated vector corresponding to the cold quantity feature vector can be processed by the nonlinear activation function and residual structure in the gated residual network to obtain the weight vector corresponding to the cold quantity feature vector. The nonlinear activation function includes but is not limited to the Sigmoid function, the Tanh function, or the ReLU function, etc. The obtained weight vector contains the weight vectors corresponding to each cold quantity feature vector. The weight vector can be used to characterize the weight value of each cold quantity feature vector. The weight value can be a value between 0 and 1, which is used to characterize the importance of the corresponding cold quantity feature vector, thereby improving the accuracy of feature information selection, improving the performance of the cold quantity determination model, and enhancing the robustness of the cold quantity determination model.

[0032] Among them, the weight vector corresponding to the cold feature vector can be a vector with the same length as the concatenated vector of the cold feature vector, the elements in the weight vector corresponding to the cold feature vector can be used to characterize the importance of the corresponding cold feature, and the weight vector corresponding to the cold feature vector can be obtained by processing the concatenated vector through a gated residual network.

[0033] For example, the concatenated vector corresponding to the cold feature vector with a length of 16 dimensions can be processed by a gated residual network to obtain a weight vector corresponding to the cold feature vector, and the weight vector corresponding to the cold feature vector includes weight values ​​corresponding to 16 cold feature vectors.

[0034] Step 105, exponential normalization is performed on the weight vectors corresponding to the cooling characteristic vectors to obtain the variable selection weight values ​​corresponding to the respective cooling characteristic vectors.

[0035] Specifically, the weight vector corresponding to the cold quantity characteristic vector can be processed by a normalization function. The exponential normalization function includes but is not limited to a softmax function or a Sigmoid function. For example, the weight can be adjusted by multiplying by a constant or adding an offset to obtain the variable selection weight value corresponding to each cold quantity characteristic vector, thereby improving the rationality of the weight distribution, improving the stability of the cold quantity determination model, optimizing the processing flow, and improving the processing efficiency.

[0036] Among them, the variable selection weight value corresponding to the cooling characteristic vector can be the weight value corresponding to each cooling characteristic vector after exponential normalization processing, and the weight value corresponding to the cooling characteristic vector can be used to characterize the importance or influence of different cooling characteristics in the cooling determination model.

[0037] Step 106: According to a preset mapping relationship, the control vectors corresponding to each cooling characteristic vector and the variable selection weight values ​​corresponding to each cooling characteristic vector are concatenated to obtain a target characteristic vector corresponding to the cooling characteristic vector.

[0038] Specifically, the preset mapping relationship can correspond to a variable selection weight value for each cooling feature vector, and then the control vector corresponding to the cooling feature vector and the variable selection weight value corresponding to the cooling feature vector can be multiplied one by one, and multiple multiplication results can be spliced. The splicing method can be series connection, or it can be combined or converted according to the preset mapping relationship, which is not limited here. The target feature vector corresponding to the cooling feature vector is obtained, thereby improving the richness of the cooling feature vector, simplifying the processing flow, and improving the performance of the cooling determination model.

[0039] Among them, the preset mapping relationship can be a rule or function set in advance before the splicing process, which can be used to characterize the method of matching the control vector corresponding to the cooling feature vector and the variable selection weight value corresponding to the cooling feature vector, and performing splicing. The target feature vector corresponding to the cooling feature vector can be a feature vector obtained by splicing the control vector corresponding to the cooling feature vector and the variable selection weight value corresponding to the cooling feature vector. The target feature vector corresponding to the cooling feature vector can be used to determine the target cooling data.

[0040] For example, the preset mapping relationship can be a one-to-one correspondence between the control vector corresponding to the cooling feature vector and the variable selection weight value corresponding to the cooling feature vector, and then the control vector corresponding to the cooling feature vector and the variable selection weight value corresponding to the cooling feature vector are multiplied and concatenated respectively to obtain the target feature vector corresponding to the cooling feature vector.

[0041] Step 107, processing the target feature vector corresponding to the cooling feature vector through a fully connected layer to obtain target cooling data.

[0042] Specifically, the target feature vector corresponding to the cold feature vector can be multiplied by a weight matrix through a fully connected layer, for example, the dimension of the target feature vector can be n, for example, the dimension of the weight matrix can be m*n, where m can be the dimension of the output feature vector, and a bias vector can be added, for example, the size of the bias vector can be m, to obtain the linear transformation result of the target feature vector. The linear transformation result of the target feature vector can be obtained by a nonlinear activation function, including but not limited to a ReLU function, a Sigmoid function, or a Tanh function, etc., which can be used to introduce nonlinear properties and obtain target cold data. In this way, through the processing of the fully connected layer, the target feature vector corresponding to the cold feature vector can be mapped to a low-dimensional representation space, thereby improving the performance of the cold determination model. The fully connected layer can extract meaningful features from the cold feature vector by learning weights and biases, thereby enhancing the quality of feature information.

[0043] Among them, the fully connected layer can be a layer structure in a neural network, each neuron in the fully connected layer is connected to all neurons in the previous layer, the fully connected layer can be used to receive feature maps from the convolution layer or the pooling layer, and reduce the dimension, extract and combine the feature information, the target cooling data can be the final representation of the target feature vector after processing by the fully connected layer, and the target cooling data can be used to characterize the cooling demand in a specified time period.

[0044] According to the technical solution provided by the embodiment of the present disclosure, at least one cold quantity unique hot vector is linearly transformed into a cold quantity feature vector through a trained cold quantity determination model, and the cold quantity feature vector corresponding to the cold quantity unique hot vector can be processed through a gated residual network to obtain a control vector corresponding to the cold quantity feature vector, and then the cold quantity feature vectors corresponding to each cold quantity unique hot vector can be spliced ​​to obtain a spliced ​​vector corresponding to the cold quantity feature vector, and the spliced ​​vector corresponding to the cold quantity feature vector can be processed through a gated residual network to obtain a weight vector corresponding to the cold quantity feature vector, and the weight vector corresponding to the cold quantity feature vector contains weight values ​​corresponding to each control vector, and then the cold quantity feature vector can be normalized by a normalized exponential function. The corresponding weight vector is exponentially normalized to obtain the variable selection weight value corresponding to each cooling feature vector; according to the preset mapping relationship, the control vector corresponding to the cooling feature vector and the variable selection weight value are correspondingly multiplied and spliced ​​to obtain the target feature vector corresponding to the cooling feature vector; the target feature vector corresponding to the cooling feature vector can be processed by the fully connected layer to obtain the target cooling data; thereby, the control ability of the cooling feature vector and the accuracy of weight allocation of the cooling determination model are enhanced through the gated residual network; the accuracy and reliability of the target cooling data are improved through the processing of the fully connected layer, which provides strong technical support for the accurate prediction and control of cooling demand and improves the efficiency and accuracy of cooling determination.

[0045] In some embodiments, before inputting at least one cooling unique-hot vector into a trained cooling determination model and performing linear transformation processing on at least one cooling unique-hot vector to obtain cooling feature vectors corresponding to each cooling unique-hot vector, it also includes: acquiring at least one environmental monitoring data and at least one cooling station equipment operation data; performing data cleaning processing on each environmental monitoring data and each cooling station equipment operation data to obtain at least one cleaned environmental monitoring data and at least one cleaned cooling station equipment operation data; performing one-hot encoding processing on each cleaned environmental monitoring data and each cleaned cooling station equipment operation data to obtain at least one cooling unique-hot vector.

[0046] Specifically, at least one environmental monitoring data and at least one cold station equipment operation data can be obtained. The environmental monitoring data may include indoor and outdoor temperature, humidity, pedestrian flow, wind force and speed, and / or weather, etc., which are not limited here. The cold station equipment operation data may include chiller power, water pump frequency, chilled water supply and return water temperature, chilled water supply and return water pressure, chilled water supply and return water flow, and / or cooling tower operating status, etc., which are not limited here. The environmental monitoring data and the cold station equipment operation data can be cleaned. The data cleaning process can be used to remove noise, fill missing values, and / or correct erroneous data, etc., which are not limited here. Then, at least one cleaned environmental monitoring data and at least one cleaned cold station equipment operation data can be obtained. The cleaned data can be one-hot encoded, and the categorical variables can be converted into a format that can be processed by the machine learning model to obtain a cold quantity one-hot vector.

[0047] Among them, environmental monitoring data may refer to data used to monitor the environmental status inside and outside the building, including but not limited to indoor and outdoor temperature, humidity, pedestrian flow, wind force and speed, and / or weather, etc. Environmental monitoring data can be used to characterize the indoor and outdoor thermal and humid environment and comfort level. Cold station equipment operation data may be the operating status and performance data of various cooling equipment in the cold station, including but not limited to chiller power, water pump frequency, chilled water supply and return water temperature, chilled water supply and return water pressure, chilled water supply and return water flow, and / or the operating status of the cooling tower, etc. Cold station equipment operation data may be used to characterize the operating efficiency and performance of the cold station equipment. The cleaned environmental monitoring data may be the environmental monitoring data processed by data cleaning. Data cleaning may be used to remove noise, missing values, and erroneous data in the original data, which are not limited here. The cleaned cold station equipment operation data may be the cold station equipment operation data processed by data cleaning. The cleaned cold station equipment operation data may be used to characterize the actual operating status and performance of the cold station equipment.

[0048] For example, the indoor temperature data and the operating status data of the cooling device can be obtained, the indoor temperature data and the operating status data of the cooling device can be cleaned, the cleaned indoor temperature data and the operating status data of the cooling device can be obtained, and the cleaned indoor temperature data and the operating status data of the cooling device can be one-hot encoded to obtain a cooling quantity one-hot vector.

[0049] According to the technical solution provided by the embodiment of the present disclosure, by acquiring at least one environmental monitoring data and at least one cold station equipment operation data, data cleaning processing can be performed on the environmental monitoring data and the cold station equipment operation data, and then at least one cleaned environmental monitoring data and at least one cleaned cold station equipment operation data can be obtained. The cleaned data can be one-hot encoded to obtain a cold quantity one-hot vector. In this way, by acquiring, cleaning and one-hot encoding the environmental monitoring data and the cold station equipment operation data, the accuracy and reliability of the data are improved, the accuracy of the cold quantity determination model for data determination is improved, the operating efficiency of the cold station equipment is improved, energy waste is reduced, and operating costs are reduced.

[0050] In some embodiments, the gated residual network includes a first fully connected layer, a second fully connected layer, and a gated linear unit; the cold quantity feature vector corresponding to each cold quantity unique hot vector is processed by the gated residual network to obtain a control vector corresponding to each cold quantity feature vector, including: processing the cold quantity feature vector corresponding to the cold quantity unique hot vector by the first fully connected layer to obtain a first cold quantity fully connected vector corresponding to the cold quantity feature vector; performing nonlinear activation processing on the first cold quantity fully connected vector corresponding to the cold quantity feature vector to obtain an activation feature vector corresponding to the cold quantity feature vector; processing the activation feature vector corresponding to the cold quantity feature vector by the second fully connected layer to obtain a second cold quantity fully connected vector corresponding to the cold quantity feature vector; performing regularization processing on the second cold quantity fully connected vector corresponding to the cold quantity feature vector to obtain a regularized feature vector corresponding to the cold quantity feature vector; processing the regularized feature vector corresponding to the cold quantity feature vector by the gated linear unit to obtain a gated cold quantity feature vector corresponding to the cold quantity feature vector; performing residual connection processing on the gated cold quantity feature vector corresponding to the cold quantity feature vector and the cold quantity feature vector corresponding to the cold quantity unique hot vector to obtain a control vector corresponding to the cold quantity feature vector.

[0051] Specifically, the cold quantity unique hot vector can be processed by the first fully connected layer in the gated residual network. The first fully connected layer can be a densely connected neural network layer, which can be used to learn the linear relationship between input and output, and then the cold quantity first fully connected vector corresponding to the cold quantity feature vector can be obtained. The obtained cold quantity first fully connected vector can be subjected to nonlinear activation processing. The nonlinear activation function corresponding to the nonlinear activation processing includes but is not limited to the ReLU function, the Sigmoid function, or the Tanh function, etc., which are not limited here. The activation feature vector corresponding to the cold quantity feature vector can be obtained. The activation feature vector corresponding to the cold quantity feature vector can then be processed by the second fully connected layer to obtain the cold quantity second fully connected vector corresponding to the cold quantity feature vector. The cold quantity second fully connected vector can be regularized to obtain the regularized feature vector corresponding to the cold quantity feature vector. The regularized feature vector corresponding to the cold quantity feature vector can be processed by a gated linear unit to obtain the gated cold quantity feature vector corresponding to the cold quantity feature vector. The gated cold quantity feature vector and the cold quantity feature vector corresponding to the cold quantity unique hot vector can be residually connected to obtain the control vector corresponding to the cold quantity feature vector.

[0052] Among them, the first fully connected layer can be a layer in a neural network, each input node is connected to the output node, and can be used to learn the linear relationship between input and output. The second fully connected layer can be a fully connected layer in a gated residual network that is at a different position from the first fully connected layer. The gated linear unit can be a neural network unit that combines a gating mechanism and a linear transformation and can be used to dynamically control the passage of information flow. The first cold quantity fully connected vector corresponding to the cold quantity feature vector can be the output vector obtained after processing the cold quantity feature vector by the first fully connected layer. The activation feature vector corresponding to the cold quantity feature vector can be the vector obtained after performing nonlinear activation processing on the first cold quantity fully connected vector. The second cold quantity fully connected vector corresponding to the cold quantity feature vector can be the output vector obtained after processing the activation feature vector by the second fully connected layer. The regularized feature vector corresponding to the cold quantity feature vector can be the vector obtained after regularizing the second cold quantity fully connected vector. The gated cold quantity feature vector corresponding to the cold quantity feature vector can be the vector obtained after processing the regularized feature vector by the gated linear unit.

[0053] For example, the cold quantity unique hot vector can be processed by the first fully connected layer in the gated residual network to obtain the cold quantity first fully connected vector corresponding to the cold quantity feature vector. The obtained cold quantity first fully connected vector can be nonlinearly activated by the Sigmoid function to obtain the activation feature vector corresponding to the cold quantity feature vector. The activation feature vector corresponding to the cold quantity feature vector can be processed by the second fully connected layer to obtain the cold quantity second fully connected vector corresponding to the cold quantity feature vector. The cold quantity second fully connected vector can be weight attenuated regularized to obtain the regularized feature vector corresponding to the cold quantity feature vector. The regularized feature vector corresponding to the cold quantity feature vector can be processed by the gated linear unit to obtain the gated cold quantity feature vector corresponding to the cold quantity feature vector. The gated cold quantity feature vector and the cold quantity feature vector corresponding to the cold quantity unique hot vector can be residually connected to obtain the control vector corresponding to the cold quantity feature vector.

[0054] According to the technical solution provided by the embodiment of the present disclosure, the cold quantity unique hot vector is processed by the first fully connected layer in the gated residual network, and then the cold quantity first fully connected vector corresponding to the cold quantity feature vector can be obtained, and the obtained cold quantity first fully connected vector can be subjected to nonlinear activation processing to obtain the activation feature vector corresponding to the cold quantity feature vector, and the activation feature vector corresponding to the cold quantity feature vector can be subjected to processing by the second fully connected layer to obtain the cold quantity second fully connected vector corresponding to the cold quantity feature vector, and the cold quantity second fully connected vector can be subjected to regularization processing to obtain the regularized feature vector corresponding to the cold quantity feature vector, and the regularized feature vector corresponding to the cold quantity feature vector can be processed by a gated linear unit to obtain the gated cold quantity feature vector corresponding to the cold quantity feature vector, and the gated cold quantity feature vector and the cold quantity feature vector corresponding to the cold quantity unique hot vector can be subjected to residual connection to obtain the control vector corresponding to the cold quantity feature vector, thereby avoiding the loss phenomenon in the process of information transmission through the residual connection, improving the dynamic ability of the information flow through the gated linear unit, preventing the overfitting phenomenon through regularization processing, and improving the generalization ability of the cold quantity determination model.

[0055] In some embodiments, before inputting at least one cooling quantity unique-hot vector into a trained cooling quantity determination model and performing linear transformation processing on at least one cooling quantity unique-hot vector to obtain cooling quantity feature vectors corresponding to each cooling quantity unique-hot vector, it also includes: obtaining multiple training sample data sets and labels corresponding to each training sample data set, each training sample data set includes multiple environmental monitoring training data and multiple cooling station equipment operation training data, and the labels corresponding to the training sample data sets are used to characterize the actual cooling quantity of the training sample data sets; performing data cleaning processing on each training sample data set to obtain at least one cleaned environmental monitoring training data and at least one cleaned cooling station equipment operation training data; performing one-hot encoding processing on each cleaned environmental monitoring training data and each cleaned cooling station equipment operation training data to obtain at least one cooling quantity unique-hot training vector; performing linear transformation processing on at least one cooling quantity unique-hot training vector to obtain cooling quantity feature training vectors corresponding to each cooling quantity unique-hot training vector; and encoding the cooling quantity feature vectors corresponding to each cooling quantity unique-hot training vector through a gated residual network. The training vectors are processed to obtain the control vectors corresponding to each cold quantity feature training vector; the cold quantity feature training vectors corresponding to each cold quantity unique hot training vector are spliced ​​to obtain the splicing vectors corresponding to the cold quantity feature training vectors; the splicing vectors corresponding to the cold quantity feature training vectors are processed through the gated residual network to obtain the weight vectors corresponding to the cold quantity feature training vectors; the weight vectors corresponding to the cold quantity feature training vectors are exponentially normalized to obtain the variable selection weight values ​​corresponding to each cold quantity feature training vector; according to the preset mapping relationship, the control vectors corresponding to each cold quantity feature training vector and the variable selection weight values ​​corresponding to each cold quantity feature vector are spliced ​​to obtain the target feature vector corresponding to the cold quantity feature vector; the target feature vector corresponding to the cold quantity feature training vector is processed through the fully connected layer to obtain the training cold quantity data; the loss of the cold quantity determination model is determined according to the root mean square error loss function, the training cold quantity data, and the labels corresponding to the training sample data set; the parameters in the cold quantity determination model are updated according to the loss of the cold quantity determination model through a loop iteration.

[0056] Specifically, by obtaining multiple training sample data sets, each set contains multiple environmental monitoring data and multiple cold station equipment operation data, and is accompanied by a label representing the actual cold capacity, the training sample data sets can be cleaned to remove abnormal or erroneous data, and the cleaned environmental monitoring training data and the cleaned cold station equipment operation training data can be converted into a cold capacity one-hot training vector through one-hot encoding, and can continue to perform linear transformation to obtain a cold capacity feature training vector corresponding to the cold capacity one-hot training vector, and the cold capacity feature training vector can be processed through a gated residual network to obtain a control vector and a corresponding cold capacity feature training vector. The training vectors are spliced ​​to obtain the splicing vector corresponding to the cold quantity feature training vector, and then the splicing vector corresponding to the cold quantity feature training vector is processed through the gated residual network to obtain the weight vector corresponding to the cold quantity feature training vector, and then the weight vector corresponding to the cold quantity feature training vector can be exponentially normalized to obtain the variable selection weight value, and can be spliced ​​with the control vector into a target feature vector, and the target feature vector can be processed through a fully connected layer to obtain training cold quantity data, and the difference between the training cold quantity data and the real cold quantity label can be compared according to the root mean square error loss function to determine the loss of the cold quantity determination model, and the model parameters can be optimized through a cyclic iteration.

[0057] Among them, the training sample data set can be a data set containing multiple environmental monitoring training data and cold station equipment operation training data, which can be used in the training process of the cold capacity determination model. The label corresponding to the training sample data set can be the actual cold capacity value of each training sample data set, which can be used for model performance evaluation in the supervised learning process. The actual cold capacity corresponding to the training sample data set can be the actually measured or recorded cold capacity value, which can be used for the target value obtained by the cold capacity determination model. The root mean square error loss function can be a loss function for regression problems, which can be used to measure the average difference between the predicted value and the actual value of the cold capacity determination model. The loss of the cold capacity determination model can be a measure of the difference between the cold capacity value predicted by the cold capacity determination model during the training process and the actual cold capacity value, which can be used to guide the parameter update and optimization of the cold capacity determination model.

[0058] In addition, it should be noted that the parameters can be updated through adaptive optimization algorithms. For example, the Adam algorithm can set the initial learning rate to 0.001 for dynamic attenuation. Regularization can prevent overfitting through strategies such as random dropout, L1 norm regularization, and weight decay regularization, which are not limited here.

[0059] According to the technical solution provided by the embodiment of the present disclosure, by obtaining multiple training sample data sets, each set contains multiple environmental monitoring data and multiple cold station equipment operation data, and is accompanied by a label representing the actual cold capacity, the training sample data sets can be cleaned to remove abnormal or erroneous data, and the cleaned environmental monitoring training data and the cleaned cold station equipment operation training data can be converted into a cold capacity one-hot training vector through one-hot encoding, and can continue to be linearly transformed to obtain a cold capacity feature training vector corresponding to the cold capacity one-hot training vector, the cold capacity feature training vector can be processed by a gated residual network to obtain a control vector corresponding to the cold capacity feature training vector and at the same time, the cold capacity feature training vector can be spliced ​​to obtain a splicing vector corresponding to the cold capacity feature training vector, and then the cold capacity feature training vector corresponding to the gated residual network can be processed. The concatenated vector is processed to obtain the weight vector corresponding to the cold quantity feature training vector, and then the weight vector corresponding to the cold quantity feature training vector can be exponentially normalized to obtain the variable selection weight value, and can be spliced ​​with the control vector as the target feature vector, and the target feature vector can be processed by the fully connected layer to obtain the training cold quantity data, and the difference between the training cold quantity data and the real cold quantity label can be compared according to the root mean square error loss function to determine the loss of the cold quantity determination model, and the model parameters are optimized through a cyclic iteration method. In this way, the accuracy of cold quantity determination is improved by combining the environmental monitoring training data and the cold station equipment operation training data with the gated residual network, and the data quality is improved by data cleaning and unique hot encoding processing. The global information capture ability and feature selection mechanism of the gated residual network enhance the generalization ability and prediction accuracy of the cold quantity determination model.

[0060] In some embodiments, the method also includes: acquiring at least one real-time environmental monitoring data and at least one real-time cold station equipment operation data; performing data cleaning processing on at least one real-time environmental monitoring data and at least one real-time cold station equipment operation data to obtain at least one cleaned real-time environmental monitoring data and at least one cleaned real-time cold station equipment operation data; performing one-hot encoding processing on each cleaned real-time environmental monitoring data and each cleaned real-time cold station equipment operation data to obtain at least one real-time cold quantity one-hot vector; according to a preset sliding window and a preset cold quantity determination period, processing at least one real-time cold quantity one-hot vector through a cold quantity determination model to obtain real-time cold quantity data.

[0061] Specifically, at least one real-time environmental monitoring data and at least one real-time cold station equipment operation data can be obtained, and then the real-time environmental monitoring data and the real-time cold station equipment operation data can be cleaned to remove noise, outliers, and data in non-standard formats. The cleaned data can be one-hot encoded to obtain at least one real-time cold quantity one-hot vector. The real-time cold quantity one-hot vector can be processed by a trained cold quantity determination model according to a preset sliding window and cold quantity determination period to obtain real-time cold quantity data.

[0062] Among them, the real-time environmental monitoring data can be data about environmental conditions obtained in real time through the environmental monitoring system, including but not limited to real-time indoor and outdoor temperature, humidity, human flow, wind speed, weather, etc. The real-time cold station equipment operation data can be data generated in real time during the operation of the cold station equipment, including but not limited to real-time obtained chiller power, water pump frequency, chilled water supply and return water temperature, chilled water supply and return water pressure, chilled water supply and return water flow, cooling tower operation status, etc. The real-time cooling capacity unique hot vector can be a vector obtained by performing unique hot encoding processing on the cleaned real-time environmental monitoring data and the cleaned real-time cold station equipment operation data, which can be used to characterize the real-time cooling capacity state. The preset sliding window can be a time range, which can be used to select data within a time period from the real-time data for analysis. The size of the sliding window can be set according to actual needs, including but not limited to 1 hour, 2 hours, or 5 hours, etc. The preset cooling capacity determination period can be a specific time period for determining the cooling capacity, including but not limited to the next 4 hours, 8 hours, or 24 hours, etc. The real-time cooling capacity data can be the data obtained by processing the real-time cooling capacity unique hot vector through the cooling capacity determination model, which can be used to characterize the cooling capacity size at the current moment.

[0063] For example, the real-time environmental monitoring data and the real-time cold station equipment operation data can be cleaned, and the cleaned data can be one-hot encoded to obtain a real-time cooling capacity one-hot vector. The preset sliding window can have a step length of 1 hour, and the preset cooling capacity determination period can be the future 24-hour cooling capacity demand. The real-time cooling capacity data is obtained through the cooling capacity determination model.

[0064] According to the technical solution provided by the embodiment of the present disclosure, by acquiring at least one real-time environmental monitoring data and at least one real-time cold station equipment operation data, data cleaning processing can be performed on the real-time environmental monitoring data and the real-time cold station equipment operation data, and the cleaned data can be processed by one-hot encoding to obtain at least one real-time cold quantity one-hot vector, and the real-time cold quantity one-hot vector can be processed by a trained cold quantity determination model according to a preset sliding window and cold quantity determination period to obtain real-time cold quantity data. In this way, by acquiring environmental monitoring data and cold station equipment operation data in real time, combined with data cleaning and one-hot encoding processing, and determining cold quantity through a preset sliding window and a preset cold quantity determination period, dynamic, accurate and real-time monitoring and prediction of cold quantity are guaranteed, energy consumption is reduced, and operating efficiency is improved.

[0065] Based on the aforementioned embodiment, the method also includes: storing real-time cooling data in a preset database; based on the preset data accumulation time, screening and processing the real-time cooling data stored in the preset database to obtain real-time cooling data corresponding to the preset data accumulation time; and updating the parameters in the cooling determination model based on the real-time cooling data corresponding to the preset data accumulation time.

[0066] Specifically, the real-time cooling data can be stored in a preset database, which can be used to store and manage data related to cooling. The real-time cooling data stored in the preset database can be screened based on the preset data accumulation time. The screening process may include removing abnormal values, calculating statistics such as average values, maximum values, minimum values, and / or selecting real-time cooling data within the preset data accumulation time to obtain real-time cooling data within the preset data accumulation time. The cooling determination model can be trained based on the real-time cooling data corresponding to the preset data accumulation time, and the parameters in the cooling determination model can be updated.

[0067] Among them, the preset database can be a preset data storage structure, a database for storing and managing real-time cooling data, the preset data accumulation time can be a preset time range, which can be used to determine the screening and processing of real-time cooling data in a specific time period in the database, and the preset data accumulation time can be set according to actual needs, including but not limited to 7 days, 1 month, 1 year, etc. The real-time cooling data corresponding to the preset data accumulation time can be the real-time cooling data screened from the preset database within the preset data accumulation time. The real-time cooling data corresponding to the preset data accumulation time can be used to train the cooling determination model, and then the parameters of the cooling determination model can be updated.

[0068] For example, the preset data accumulation time can be 7 days. The real-time cooling data obtained in the last 7 days can be filtered out from the preset database, and the real-time cooling data obtained in the filtered 7 days can be used as a training sample data set to train the cooling determination model, and the parameters in the cooling determination model can be updated according to the obtained losses.

[0069] In addition, it should be noted that the validation set can be used to verify the performance of the cold capacity determination model trained based on the real-time cold capacity data corresponding to the preset data accumulation time. If the result of the model on the validation set is better than the current online model, the new model is updated and deployed through hot update.

[0070] According to the technical solution provided by the embodiment of the present disclosure, by storing real-time cold capacity data in a preset database, the real-time cold capacity data stored in the preset database can be screened and processed based on the preset data accumulation time to obtain the real-time cold capacity data within the preset data accumulation time, and the cold capacity determination model can be trained based on the real-time cold capacity data corresponding to the preset data accumulation time, and then the parameters in the cold capacity determination model can be updated. In this way, by storing the real-time cold capacity data in the preset database, centralized management of cold capacity data and preservation of historical data are realized, and the real-time cold capacity data is screened and processed based on the preset data accumulation time, which improves the effectiveness of updating the parameters of the cold capacity determination model, and regularly updates the parameters of the cold capacity determination model, thereby improving the accuracy and adaptability of the cold capacity determination model.

[0071] In some embodiments, the environmental monitoring data includes indoor temperature and humidity data; the cold station equipment operation data includes cooling capacity data; the cleaned environmental monitoring data includes indoor temperature and humidity analysis data and indoor wet-bulb temperature data; the cleaned cold station equipment operation data includes non-negative cooling capacity data; data cleaning processing is performed on each environmental monitoring data and each cold station equipment operation data to obtain at least one cooling capacity-related clean data, including: negative value truncation processing is performed on the environmental monitoring cooling capacity data to obtain non-negative cooling capacity data; exploratory data analysis processing is performed on the indoor temperature and humidity data to obtain indoor temperature and humidity analysis data; and the indoor wet-bulb temperature data is determined based on the indoor temperature and humidity data.

[0072] Specifically, data cleaning processing can be performed on the environmental monitoring data and the cold station equipment operation data. For the indoor temperature and humidity data in the environmental monitoring data, exploratory data analysis processing can be performed, including but not limited to data validity checking, outlier processing, and / or missing value filling and other steps, to obtain indoor temperature and humidity analysis data. The indoor wet-bulb temperature data can be determined based on the temperature and humidity information in the indoor temperature and humidity data through the wet-bulb temperature calculation formula or table lookup method. For the cooling data in the cold station equipment operation data, negative value truncation processing can be performed, and non-negative cooling data can be obtained by setting the negative cooling data to zero or a suitable non-negative value.

[0073] Among them, the indoor temperature and humidity data can be the temperature and humidity information in the indoor environment, which can be obtained through temperature and humidity sensors and other equipment. The cooling capacity data can be the cooling capacity data generated by the cold station equipment during operation, which can be used to characterize the cooling capacity and efficiency of the equipment. The indoor temperature and humidity analysis data can be the indoor temperature and humidity data obtained after exploratory data analysis and processing. The indoor wet-bulb temperature data can be the wet-bulb temperature information calculated based on the indoor temperature and humidity data, which can be used to evaluate the comfort of the indoor environment. The non-negative cooling capacity data can be the cooling capacity data after negative value truncation processing, which can be used to ensure that all cooling capacity values ​​are non-negative values, which conforms to physical laws.

[0074] For example, the acquired cooling data can be subjected to negative value truncation processing and negative values ​​can be truncated to zero, exploratory data analysis can be performed on the indoor temperature and humidity data, and statistical feature extraction can be performed, and the data can be filtered by filtering abnormal sensors. Based on the acquired indoor temperature and humidity data, the wet-bulb temperature can be calculated as additional feature data.

[0075] In addition, it should be noted that additional date and time feature data can be constructed based on the obtained time data, including but not limited to holiday time, adjusted holiday time, number of days until holiday, and / or business hours, etc.; the cooling data can also be accumulated to obtain the accumulated cooling per hour, which can be used to ensure the stability of the feature data; the tilted feature data can also be processed, and the processing methods include but are not limited to conversion, normalization, and / or truncation processing, etc.

[0076] According to the technical solution provided by the embodiment of the present disclosure, by performing data cleaning processing on the environmental monitoring data and the cold station equipment operation data, exploratory data analysis processing can be performed on the indoor temperature and humidity data in the environmental monitoring data, and indoor temperature and humidity analysis data can be obtained. The indoor wet-bulb temperature data can be determined based on the temperature and humidity information in the indoor temperature and humidity data, and negative value truncation processing can be performed on the cold capacity data in the cold station equipment operation data. Through data cleaning processing, the accuracy of the cold capacity cleaning data is improved, the operation efficiency of the cold station is improved, and the energy consumption is reduced.

[0077] Figure 2 is a schematic diagram of the structure of a gated residual network provided by an embodiment of the present disclosure, such as Figure 2 As shown, the structure of the gated residual network includes:

[0078] The first fully connected layer 201 may be a layer in a neural network, where each input node is connected to an output node, and is used to learn the linear relationship between input and output, thereby obtaining a first fully connected vector of coldness corresponding to a coldness feature vector;

[0079] The second fully connected layer 202 may be a fully connected layer in the gated residual network that is located at a different position from the first fully connected layer;

[0080] The gated linear unit 203 may be a neural network unit that combines a gate mechanism and a linear transformation and may be used to dynamically control the passage of information flow;

[0081] According to the technical solution provided by the embodiment of the present disclosure, the first fully connected vector corresponding to the cold quantity feature vector is obtained by inputting the cold quantity unique hot vector into the first fully connected layer 201 of the gated residual network for processing, and then the first fully connected vector corresponding to the cold quantity feature vector can be subjected to nonlinear activation processing, and the first fully connected vector corresponding to the input cold quantity feature vector can be multiplied by the result of Tanh(log(1+Sigmoid(x))), where x can be the input cold quantity feature vector, and then the activation feature vector corresponding to the cold quantity feature vector is obtained, and the activation feature vector corresponding to the cold quantity feature vector can be processed by the second fully connected layer 202 to obtain the cold quantity second fully connected vector corresponding to the cold quantity feature vector, and the weight attenuation regularization processing can be performed on the cold quantity second fully connected vector to obtain The regularized feature vector corresponding to the cold quantity feature vector can be processed by the gated linear unit 203 to obtain the gated cold quantity feature vector corresponding to the cold quantity feature vector, wherein the gated linear unit 203 can be composed of a linear transformation and a Sigmoid function product, and the gated cold quantity feature vector and the cold quantity feature vector corresponding to the cold quantity unique hot vector can be residually connected to obtain the control vector corresponding to the cold quantity feature vector, wherein the residual link can be processed by a gated residual network, and can be obtained by ReLU function and linear transformation, thereby avoiding the loss of information during transmission through residual connection, improving the dynamic ability of information flow through the gated linear unit, preventing overfitting through regularization processing, and improving the generalization ability of the cold quantity determination model.

[0082] 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.

[0083] 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.

[0084] Figure 3 Schematic diagram of a cooling capacity determination device provided by an embodiment of the present disclosure. Figure 3 As shown, the cooling capacity determining device comprises:

[0085] The first processing module 301 is used to input at least one cooling unique hot vector into the trained cooling determination model, perform linear transformation processing on at least one cooling unique hot vector, and obtain cooling feature vectors corresponding to each cooling unique hot vector;

[0086] The second processing module 302 is used to process the cooling feature vectors corresponding to each cooling unique heat vector through a gated residual network to obtain a control vector corresponding to each cooling feature vector;

[0087] The third processing module 303 is used to perform splicing processing on the cold quantity feature vectors corresponding to each cold quantity unique heat vector to obtain a splicing vector corresponding to the cold quantity feature vector;

[0088] The fourth processing module 304 is used to process the concatenated vector corresponding to the cold feature vector through a gated residual network to obtain a weight vector corresponding to the cold feature vector;

[0089] The fifth processing module 305 is used to perform exponential normalization processing on the weight vector corresponding to the cooling characteristic vector to obtain the variable selection weight value corresponding to each cooling characteristic vector;

[0090] The determination module 306 is used to perform splicing processing on the control vectors corresponding to each cooling characteristic vector and the variable selection weight values ​​corresponding to each cooling characteristic vector according to a preset mapping relationship to obtain a target characteristic vector corresponding to the cooling characteristic vector;

[0091] The sixth processing module 307 is used to process the target feature vector corresponding to the cooling feature vector through a fully connected layer to obtain target cooling data.

[0092] According to the technical solution provided by the embodiment of the present disclosure, at least one cold quantity unique hot vector is linearly transformed into a cold quantity feature vector through a trained cold quantity determination model, and the cold quantity feature vector corresponding to the cold quantity unique hot vector can be processed through a gated residual network to obtain a control vector corresponding to the cold quantity feature vector, and then the cold quantity feature vectors corresponding to each cold quantity unique hot vector can be spliced ​​to obtain a spliced ​​vector corresponding to the cold quantity feature vector, and the spliced ​​vector corresponding to the cold quantity feature vector can be processed through a gated residual network to obtain a weight vector corresponding to the cold quantity feature vector, and the weight vector corresponding to the cold quantity feature vector contains weight values ​​corresponding to each control vector, and then the cold quantity feature vector can be normalized by a normalized exponential function. The corresponding weight vector is exponentially normalized to obtain the variable selection weight value corresponding to each cooling feature vector; according to the preset mapping relationship, the control vector corresponding to the cooling feature vector and the variable selection weight value are correspondingly multiplied and spliced ​​to obtain the target feature vector corresponding to the cooling feature vector; the target feature vector corresponding to the cooling feature vector can be processed by the fully connected layer to obtain the target cooling data; thereby, the control ability of the cooling feature vector and the accuracy of weight allocation of the cooling determination model are enhanced through the gated residual network; the accuracy and reliability of the target cooling data are improved through the processing of the fully connected layer, which provides strong technical support for the accurate prediction and control of cooling demand and improves the efficiency and accuracy of cooling determination.

[0093] In some embodiments, the above-mentioned cooling capacity determination device is also used to obtain at least one environmental monitoring data and at least one cooling station equipment operation data; perform data cleaning processing on each environmental monitoring data and each cooling station equipment operation data to obtain at least one cleaned environmental monitoring data and at least one cleaned cooling station equipment operation data; perform unique hot encoding processing on each cleaned environmental monitoring data and each cleaned cooling station equipment operation data to obtain at least one cooling capacity unique hot vector.

[0094] In some embodiments, the second processing module 302 is specifically used to process the cold quantity feature vector corresponding to the cold quantity unique hot vector through the first fully connected layer to obtain the first cold quantity fully connected vector corresponding to the cold quantity feature vector; perform nonlinear activation processing on the first cold quantity fully connected vector corresponding to the cold quantity feature vector to obtain the activation feature vector corresponding to the cold quantity feature vector; process the activation feature vector corresponding to the cold quantity feature vector through the second fully connected layer to obtain the second cold quantity fully connected vector corresponding to the cold quantity feature vector; perform regularization processing on the second cold quantity fully connected vector corresponding to the cold quantity feature vector to obtain the regularized feature vector corresponding to the cold quantity feature vector; process the regularized feature vector corresponding to the cold quantity feature vector through a gated linear unit to obtain the gated cold quantity feature vector corresponding to the cold quantity feature vector; perform residual connection processing on the gated cold quantity feature vector corresponding to the cold quantity feature vector and the cold quantity feature vector corresponding to the cold quantity unique hot vector to obtain the control vector corresponding to the cold quantity feature vector.

[0095] In some embodiments, the above-mentioned cold capacity determination device is also used to obtain multiple training sample data sets and labels corresponding to each training sample data set, each training sample data set includes multiple environmental monitoring training data and multiple cold station equipment operation training data, and the labels corresponding to the training sample data sets are used to characterize the actual cold capacity corresponding to the training sample data sets; perform data cleaning processing on each training sample data set to obtain at least one cleaned environmental monitoring training data and at least one cleaned cold station equipment operation training data; perform one-hot encoding processing on each cleaned environmental monitoring training data and each cleaned cold station equipment operation training data to obtain at least one cold capacity one-hot training vector; perform linear transformation processing on at least one cold capacity one-hot training vector to obtain a cold capacity feature training vector corresponding to each cold capacity one-hot training vector; process the cold capacity feature training vector corresponding to each cold capacity one-hot training vector through a gated residual network to obtain a control vector corresponding to each cold capacity feature training vector; The cold quantity feature training vectors corresponding to each cold quantity unique hot training vector are concatenated to obtain the concatenated vector corresponding to the cold quantity feature training vector; the concatenated vector corresponding to the cold quantity feature training vector is processed through a gated residual network to obtain the weight vector corresponding to the cold quantity feature training vector; the weight vector corresponding to the cold quantity feature training vector is exponentially normalized to obtain the variable selection weight value corresponding to each cold quantity feature training vector; according to the preset mapping relationship, the control vector corresponding to each cold quantity feature training vector and the variable selection weight value corresponding to each cold quantity feature vector are concatenated to obtain the target feature vector corresponding to the cold quantity feature vector; the target feature vector corresponding to the cold quantity feature training vector is processed through a fully connected layer to obtain the training cold quantity data; the loss of the cold quantity determination model is determined according to the root mean square error loss function, the training cold quantity data, and the labels corresponding to the training sample data set; the parameters in the cold quantity determination model are updated according to the loss of the cold quantity determination model through a loop iteration.

[0096] In some embodiments, the above-mentioned cooling capacity determination device is also used to obtain at least one real-time environmental monitoring data and at least one real-time cold station equipment operation data; perform data cleaning processing on at least one real-time environmental monitoring data and at least one real-time cold station equipment operation data to obtain at least one cleaned real-time environmental monitoring data and at least one cleaned real-time cold station equipment operation data; perform one-hot encoding processing on each cleaned real-time environmental monitoring data and each cleaned real-time cold station equipment operation data to obtain at least one real-time cooling capacity one-hot vector; according to a preset sliding window and a preset cooling capacity determination period, at least one real-time cooling capacity one-hot vector is processed through a cooling capacity determination model to obtain real-time cooling capacity data.

[0097] In some embodiments, the above-mentioned cooling capacity determination device is also used to store real-time cooling capacity data in a preset database; based on the preset data accumulation time, filter and process the real-time cooling capacity data stored in the preset database to obtain the real-time cooling capacity data corresponding to the preset data accumulation time; based on the real-time cooling capacity data corresponding to the preset data accumulation time, update the parameters in the cooling capacity determination model.

[0098] In some embodiments, data cleaning processing is performed on each environmental monitoring data and each cold station equipment operation data to obtain at least one cooling capacity-related clean data, which is specifically used for: performing negative value truncation processing on the environmental monitoring cooling capacity data to obtain non-negative cooling capacity data; performing exploratory data analysis processing on the indoor temperature and humidity data to obtain indoor temperature and humidity analysis data; and determining the indoor wet-bulb temperature data based on the indoor temperature and humidity data.

[0099] 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.

[0100] 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.

[0101] 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.

[0102] 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.

[0103] 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.

[0104] 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.

[0105] 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, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc.

[0106] 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: Inputting at least one cooling unique hot vector into the trained cooling determination model, performing linear transformation processing on at least one cooling unique hot vector, and obtaining cooling feature vectors corresponding to each cooling unique hot vector; The cooling quantity feature vectors corresponding to each cooling quantity unique hot vector are processed by a gated residual network to obtain control vectors corresponding to each cooling quantity feature vector; Performing splicing processing on the cold quantity feature vectors corresponding to the respective cold quantity unique heat vectors to obtain a splicing vector corresponding to the cold quantity feature vectors; Processing the concatenated vector corresponding to the cold feature vector through the gated residual network to obtain a weight vector corresponding to the cold feature vector; Performing exponential normalization processing on the weight vectors corresponding to the cooling characteristic vectors to obtain variable selection weight values ​​corresponding to the cooling characteristic vectors; According to a preset mapping relationship, the control vectors corresponding to the cooling quantity characteristic vectors and the variable selection weight values ​​corresponding to the cooling quantity characteristic vectors are concatenated to obtain the target characteristic vector corresponding to the cooling quantity characteristic vector; The target feature vector corresponding to the cooling feature vector is processed through a fully connected layer to obtain target cooling data.

2. The method for determining cooling capacity according to claim 1, characterized in that: Before inputting at least one cooling capacity unique hot vector into the trained cooling capacity determination model and performing linear transformation processing on at least one cooling capacity unique hot vector to obtain cooling capacity feature vectors corresponding to each cooling capacity unique hot vector, the method further includes: Acquire at least one piece of environmental monitoring data and at least one piece of cold station equipment operation data; Performing data cleaning processing on each of the environmental monitoring data and each of the cold station equipment operation data to obtain at least one cleaned environmental monitoring data and at least one cleaned cold station equipment operation data; Perform one-hot encoding processing on each of the environmental monitoring data after cleaning and each of the cold station equipment operation data after cleaning to obtain at least one of the cold quantity one-hot vectors.

3. The method for determining cooling capacity according to claim 1, characterized in that: The gated residual network includes a first fully connected layer, a second fully connected layer, and a gated linear unit; The gated residual network is used to process the cooling feature vectors corresponding to each cooling unique hot vector to obtain the control vectors corresponding to each cooling feature vector, including: Processing the cold quantity feature vector corresponding to the cold quantity unique hot vector through the first fully connected layer to obtain a cold quantity first fully connected vector corresponding to the cold quantity feature vector; Performing nonlinear activation processing on the first fully connected vector of cooling corresponding to the cooling feature vector to obtain an activated feature vector corresponding to the cooling feature vector; Processing the activation feature vector corresponding to the cooling feature vector through the second fully connected layer to obtain a second cooling second fully connected vector corresponding to the cooling feature vector; Regularizing the second fully connected vector of cooling corresponding to the cooling feature vector to obtain a regularized feature vector corresponding to the cooling feature vector; Processing the regularized feature vector corresponding to the cooling feature vector by the gated linear unit to obtain a gated cooling feature vector corresponding to the cooling feature vector; Residual connection processing is performed on the gated cooling feature vector corresponding to the cooling feature vector and the cooling feature vector corresponding to the cooling unique hot vector to obtain a control vector corresponding to the cooling feature vector.

4. The method for determining cooling capacity according to claim 1, characterized in that: Before inputting at least one cooling capacity unique hot vector into the trained cooling capacity determination model and performing linear transformation processing on at least one cooling capacity unique hot vector to obtain cooling capacity feature vectors corresponding to each cooling capacity unique hot vector, the method further includes: Acquire multiple training sample data sets and labels corresponding to each of the training sample data sets, each of the training sample data sets comprising multiple environmental monitoring training data and multiple cold station equipment operation training data, and the labels corresponding to the training sample data sets are used to characterize the actual cooling capacity corresponding to the training sample data sets; Performing data cleaning processing on each of the training sample data sets to obtain at least one cleaned environmental monitoring training data and at least one cleaned cold station equipment operation training data; Performing one-hot encoding processing on each of the cleaned environmental monitoring training data and each of the cleaned cold station equipment operation training data to obtain at least one cold quantity one-hot training vector; Performing linear transformation processing on at least one of the cold quantity unique hot training vectors to obtain cold quantity feature training vectors corresponding to each of the cold quantity unique hot training vectors; The cold quantity feature training vectors corresponding to each of the cold quantity unique hot training vectors are processed by the gated residual network to obtain the control vectors corresponding to each of the cold quantity feature training vectors; Performing splicing processing on the cold quantity feature training vectors corresponding to each of the cold quantity unique hot training vectors to obtain a splicing vector corresponding to the cold quantity feature training vector; Processing the concatenated vector corresponding to the cold feature training vector through the gated residual network to obtain a weight vector corresponding to the cold feature training vector; Performing exponential normalization processing on the weight vectors corresponding to the cooling feature training vectors to obtain variable selection weight values ​​corresponding to the cooling feature training vectors; According to the preset mapping relationship, the control vectors corresponding to the respective cooling feature training vectors and the variable selection weight values ​​corresponding to the respective cooling feature vectors are concatenated to obtain the target feature vector corresponding to the cooling feature vector; Processing the target feature vector corresponding to the cold quantity feature training vector through the fully connected layer to obtain training cold quantity data; Determine the loss of the cooling determination model according to the root mean square error loss function, the training cooling data, and the labels corresponding to the training sample data set; The parameters in the cooling capacity determination model are updated according to the loss of the cooling capacity determination model in a loop iteration manner.

5. The method for determining cooling capacity according to claim 1, characterized in that: The method further comprises: Acquire at least one real-time environmental monitoring data and at least one real-time cold station equipment operation data; Performing data cleaning processing on at least one of the real-time environmental monitoring data and at least one of the real-time cold station equipment operation data to obtain at least one cleaned real-time environmental monitoring data and at least one cleaned real-time cold station equipment operation data; Performing one-hot encoding processing on each of the real-time environmental monitoring data after cleaning and each of the real-time cold station equipment operation data after cleaning to obtain at least one of the real-time cold capacity one-hot vectors; According to a preset sliding window and a preset cooling capacity determination period, at least one of the real-time cooling capacity unique heat vectors is processed by the cooling capacity determination model to obtain real-time cooling capacity data.

6. The method for determining cooling capacity according to claim 5, characterized in that: The method further comprises: The real-time cooling data is stored in a preset database; Based on the preset data accumulation time, the real-time cooling data stored in the preset database is screened and processed to obtain the real-time cooling data corresponding to the preset data accumulation time; The parameters in the cooling capacity determination model are updated based on the real-time cooling capacity data corresponding to the preset data accumulation time.

7. The method for determining cooling capacity according to claim 2, characterized in that: The environmental monitoring data includes indoor temperature and humidity data; the refrigeration station equipment operation data includes cooling capacity data; the cleaned environmental monitoring data includes indoor temperature and humidity analysis data and indoor wet-bulb temperature data; the cleaned refrigeration station equipment operation data includes non-negative cooling capacity data; The data cleaning process is performed on each of the environmental monitoring data and each of the cold station equipment operation data to obtain at least one cold capacity related clean data, including: Performing negative value truncation processing on the environmental monitoring cooling data to obtain the non-negative cooling data; Performing exploratory data analysis on the indoor temperature and humidity data to obtain the indoor temperature and humidity analysis data; The indoor wet-bulb temperature data is determined according to the indoor temperature and humidity data.

8. A device for determining cooling capacity, characterized in that: include: A first processing module is used to input at least one cooling capacity unique hot vector into a trained cooling capacity determination model, perform linear transformation processing on at least one cooling capacity unique hot vector, and obtain a cooling capacity feature vector corresponding to each cooling capacity unique hot vector; A second processing module is used to process the cooling feature vectors corresponding to each of the cooling unique hot vectors through a gated residual network to obtain a control vector corresponding to each of the cooling feature vectors; A third processing module is used to perform splicing processing on the cold quantity feature vectors corresponding to the cold quantity unique heat vectors to obtain a splicing vector corresponding to the cold quantity feature vector; A fourth processing module, used for processing the concatenated vector corresponding to the cold feature vector through the gated residual network to obtain a weight vector corresponding to the cold feature vector; A fifth processing module, configured to perform exponential normalization processing on the weight vectors corresponding to the cooling characteristic vectors to obtain variable selection weight values ​​corresponding to the cooling characteristic vectors; A determination module, configured to perform splicing processing on the control vectors corresponding to each of the cooling characteristic vectors and the variable selection weight values ​​corresponding to each of the cooling characteristic vectors according to a preset mapping relationship, so as to obtain a target characteristic vector corresponding to the cooling characteristic vector; The sixth processing module is used to process the target feature vector corresponding to the cooling feature vector through a fully connected layer to obtain target cooling data.

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