Method, device and equipment for flexibly adjusting load of refrigerating system of central air conditioner

Abstract

The invention relates to the technical field of power regulation, and discloses a central air-conditioning refrigeration system load flexible regulation method which comprises the following steps: performing multi-modal feature extraction and short-term load analysis on a historical sensing data sequence to obtain short-term load data; and performing multi-scale expansion convolution, bidirectional time sequence feature extraction and long-term load analysis on the historical sensing data sequence to obtain long-term load data, and obtaining real-time sensing data according to the pre-obtained real-time sensing data. The short-term load data and the long-term load data are used for carrying out fitness calculation and global optimal iteration on the adjustment strategy set to obtain a primary adjustment strategy; and state space construction and state feedback updating are conducted on the primary adjusting strategy according to the real-time sensing data and the analysis load data, a standard adjusting strategy is obtained, and load adjusting is conducted on the central air conditioner through the standard adjusting strategy. The invention further provides a device and equipment for flexibly adjusting the load of the central air-conditioning refrigerating system. The sales volume of air conditioner load adjustment can be increased.

Description

Technical Field

[0001] The present invention relates to the technical field of power regulation, and in particular to a method, device and equipment for flexible regulation of load of a central air-conditioning refrigeration system. Background Art

[0002] The central air conditioning refrigeration system is a large-scale air handling equipment used to control the temperature, humidity and air quality in a building. The central air conditioner circulates the refrigerant to bring the heat inside the building to the outside to achieve a cooling effect. In order to save the energy consumption of the central air conditioner and extend the service life of the equipment, the central air conditioning refrigeration system needs to be load regulated.

[0003] Existing load regulation methods are often based on fixed rules or simple proportional integral derivative control (PID). In practical applications, load regulation methods based on fixed rules or simple PID control often lag behind changes in load demand and cannot quickly respond to sudden load changes caused by gathering of people or sudden rise in external temperature. It is also difficult to globally optimize energy consumption and comfort, resulting in high energy consumption or difficulty in ensuring comfort, which may lead to low efficiency in load regulation. Summary of the invention

[0004] The present invention provides a method, device and equipment for flexible load regulation of a central air-conditioning refrigeration system, the main purpose of which is to solve the problem of low efficiency during load regulation.

[0005] To achieve the above object, the present invention provides a method for flexible load adjustment of a central air-conditioning refrigeration system, comprising:

[0006] Perform multimodal feature extraction and short-term load analysis on the pre-acquired historical sensor data sequence to obtain short-term load data, wherein the multimodal feature extraction and short-term load analysis on the pre-acquired historical sensor data sequence to obtain short-term load data includes: performing format unification and data normalization operations on the pre-acquired historical sensor data sequence to obtain a standard sensor data sequence; performing multimodal feature extraction on the standard sensor data sequence to obtain a historical load feature sequence group; and calculating the load correlation coefficient between each historical load feature sequence in the historical load feature sequence group using the following load correlation algorithm:

[0007]

[0008] Among them, p(C i ,C j ) refers to the historical load characteristic sequence C in the historical load characteristic sequence group i and historical load characteristic sequence Cj The load correlation coefficient, i, j are the serial numbers of the historical load characteristic sequence in the historical load characteristic sequence group, K i is the historical load characteristic sequence C i The total number of features, K j is the historical load characteristic sequence C j The total number of features, k and l are the feature numbers, cov() is the covariance operator, σ() is the standard deviation operator, x ik Refers to the historical load characteristic sequence C i The kth feature in w ik is the historical load characteristic sequence C i The prior weight of the kth feature in , x jl Refers to the historical load characteristic sequence C j The lth feature in w jl is the historical load characteristic sequence C j a priori weight of the lth feature in; performing correlation dimension reduction on the historical load feature sequence group according to the load correlation coefficient to obtain a reduced dimension load feature sequence group; performing principal component dimension reduction and short-term load analysis on the reduced dimension load feature sequence group to obtain short-term load data;

[0009] Extracting a significant lag coefficient group from the historical sensor data sequence, and performing multi-scale dilated convolution and bidirectional time series feature extraction on the historical sensor data sequence according to the significant lag coefficient group to obtain a bidirectional sensor time series feature sequence;

[0010] Performing a long-term load analysis on the historical sensing data sequence according to the bidirectional sensing time series feature sequence to obtain long-term load data, and generating analytical load data according to the short-term load data and the long-term load data;

[0011] Initializing a regulation strategy set, performing fitness calculation and global optimal iteration on the regulation strategy set according to the pre-acquired real-time sensor data and the analyzed load data, and obtaining a primary regulation strategy;

[0012] The state space of the primary regulation strategy is constructed and the state feedback is updated according to the real-time sensing data and the analyzed load data to obtain a standard regulation strategy, and the standard regulation strategy is used to regulate the load of the central air conditioner.

[0013] Optionally, performing principal component dimensionality reduction and short-term load analysis on the dimensionality-reduced load feature sequence group to obtain short-term load data includes:

[0014] Performing a matrix operation on the reduced-dimensionality load feature sequence group to obtain a reduced-dimensionality load feature matrix;

[0015] Performing covariance matrix calculation on the reduced-dimensional load characteristic matrix to obtain a load characteristic relationship matrix;

[0016] Performing eigenvalue decomposition on the load characteristic relationship matrix to obtain a load characteristic vector group and a load characteristic value group;

[0017] Performing principal component screening on the load characteristic vector group according to the load characteristic value group to obtain a principal component load characteristic group;

[0018] Performing radial basis calculation on the principal component load feature group to obtain a radial basis load data group;

[0019] Performing activation response on the principal component load feature group to obtain an activation response weight group;

[0020] The radial basis load data set is weightedly fused according to the activation response weight set to obtain short-term load data.

[0021] Optionally, extracting a significant lag coefficient group from the historical sensor data sequence includes:

[0022] Performing self-attention encoding on the historical sensor data sequence to obtain attention sensor weights;

[0023] Performing hysteresis coefficient analysis on the historical sensor data sequence according to the attention sensor weight to obtain a sensor hysteresis coefficient group;

[0024] The following sensor autocorrelation algorithm is used to perform autocorrelation analysis on the historical sensor data sequence according to the sensor lag coefficient group to obtain a sensor autocorrelation array:

[0025]

[0026] Among them, ρ r is the rth sensor autocorrelation coefficient in the sensor hysteresis coefficient group, r is the coefficient number, U r is the rth sensor hysteresis coefficient in the sensor hysteresis coefficient group, t is the data sequence number, T is the length of the historical sensor data sequence, x t is the tth historical sensor data in the historical sensor data sequence, x t+r is the t+rth historical sensor data in the historical sensor data sequence;

[0027] A significant lag coefficient group is screened out from the lag coefficient group according to the lag coefficient array.

[0028] Optionally, performing multi-scale dilated convolution and bidirectional time series feature extraction on the historical sensing data sequence according to the significant lag coefficient group to obtain a bidirectional sensing time series feature sequence includes:

[0029] Perform expansion rate configuration according to the significant hysteresis coefficient group to obtain an expansion rate group;

[0030] Initializing a multi-scale dilated convolution kernel group according to the dilation rate group;

[0031] Performing dilated convolution on the historical sensor data sequence using the multi-scale dilated convolution kernel group to obtain a dilated sensor feature sequence group;

[0032] Performing feature fusion on the expanded sensing feature sequence group to obtain a multi-scale sensing feature sequence;

[0033] Performing forward time series feature extraction on the multi-scale sensing feature sequence to obtain a forward sensing time series feature sequence;

[0034] Performing backward time series feature extraction on the multi-scale sensing feature sequence to obtain a backward sensing time series feature sequence;

[0035] The forward time series feature sequence and the backward time series feature sequence are feature concatenated to obtain a bidirectional sensing time series feature sequence.

[0036] Optionally, performing a long-term load analysis on the historical sensing data sequence according to the bidirectional sensing time series feature sequence to obtain long-term load data includes:

[0037] Performing self-attention encoding on the bidirectional sensing temporal feature sequence to obtain a bidirectional self-attention weight;

[0038] Using the bidirectional self-attention weight, a weighted load analysis is performed on the bidirectional sensing time series feature sequence to obtain attention long-term load data;

[0039] Performing autoregressive analysis on the bidirectional sensing time series feature sequence to obtain an autoregressive sensing data sequence;

[0040] Performing residual calculation on the historical sensor data sequence and the autoregressive sensor data sequence to obtain a residual sensor data sequence;

[0041] Perform multi-step decoding on the bidirectional sensing time series feature sequence according to the residual sensing data sequence to obtain decoded long-term load data;

[0042] The attention long-term load data and the decoding long-term load data are weightedly fused to obtain long-term load data.

[0043] Optionally, performing fitness calculation and global optimal iteration on the adjustment strategy set according to the pre-acquired real-time sensor data and the analyzed load data to obtain a primary adjustment strategy includes:

[0044] Selecting the regulation strategies in the regulation strategy set one by one as the target regulation strategy, performing load analysis on the real-time sensor data according to the target regulation strategy, and obtaining regulation load data;

[0045] Extracting real-time temperature data from the real-time sensor data, performing thermodynamic analysis on the target adjustment strategy according to the real-time temperature data, and obtaining target cooling capacity data;

[0046] The fitness corresponding to the target adjustment strategy is calculated according to the adjustment load data, the target cooling capacity data and the analysis load data using the following fitness algorithm:

[0047]

[0048] Wherein, F refers to the fitness, w1 is the preset load weight coefficient, N is the total time length of the load data analysis, n is the time index, is the load value at the nth moment in the analyzed load data, M is the total number of devices corresponding to the regulated load data, m is the device index, It refers to the power consumption value of the mth device in the load regulation data at the nth time, v m refers to the efficiency coefficient of the mth device in the load regulation data, || is the absolute value symbol, w2 is the preset energy efficiency weight coefficient, Q m It refers to the cooling capacity value corresponding to the mth device in the target cooling capacity data;

[0049] The fitness of all target adjustment strategies in the adjustment strategy set is aggregated into a fitness set, and the adjustment strategy set is globally optimally iterated according to the fitness set to obtain a primary adjustment strategy.

[0050] Optionally, constructing the state space and updating the state feedback of the primary regulation strategy according to the real-time sensing data and the analyzed load data to obtain a standard regulation strategy includes:

[0051] Constructing a state space according to the real-time sensor data and the primary adjustment strategy to obtain an air conditioning load state space;

[0052] Calculating a cost matrix of the air conditioning load state space according to the analyzed load data;

[0053] Solving the cost matrix for gain to obtain a feedback gain matrix;

[0054] The primary regulation strategy is feedback updated according to the gain feedback matrix to obtain a standard regulation strategy.

[0055] In order to solve the above problems, the present invention also provides a central air-conditioning refrigeration system load flexible adjustment device, the device comprising:

[0056] The short-term analysis module is used to perform multimodal feature extraction and short-term load analysis on the pre-acquired historical sensor data sequence to obtain short-term load data, wherein the multimodal feature extraction and short-term load analysis on the pre-acquired historical sensor data sequence to obtain short-term load data includes: performing format unification and data normalization operations on the pre-acquired historical sensor data sequence to obtain a standard sensor data sequence; performing multimodal feature extraction on the standard sensor data sequence to obtain a historical load feature sequence group; and calculating the load correlation coefficient between each historical load feature sequence in the historical load feature sequence group using the following load correlation algorithm:

[0057]

[0058] Among them, p(C i ,C j ) refers to the historical load characteristic sequence C in the historical load characteristic sequence group i and historical load characteristic sequence C j The load correlation coefficient, i, j are the serial numbers of the historical load characteristic sequence in the historical load characteristic sequence group, K i is the historical load characteristic sequence C i The total number of features, K j is the historical load characteristic sequence C j The total number of features, k and l are the feature numbers, cov() is the covariance operator, σ() is the standard deviation operator, x ik Refers to the historical load characteristic sequence C i The kth feature in w ik is the historical load characteristic sequence C i The prior weight of the kth feature in , x jl Refers to the historical load characteristic sequence C j The lth feature in w jl is the historical load characteristic sequence C j a priori weight of the lth feature in; performing correlation dimension reduction on the historical load feature sequence group according to the load correlation coefficient to obtain a reduced dimension load feature sequence group; performing principal component dimension reduction and short-term load analysis on the reduced dimension load feature sequence group to obtain short-term load data;

[0059] A bidirectional analysis module, used to extract a significant lag coefficient group from the historical sensor data sequence, and perform multi-scale dilated convolution and bidirectional time series feature extraction on the historical sensor data sequence according to the significant lag coefficient group to obtain a bidirectional sensor time series feature sequence;

[0060] A long-term analysis module, configured to perform long-term load analysis on the historical sensing data sequence according to the bidirectional sensing time series feature sequence to obtain long-term load data, and generate analysis load data according to the short-term load data and the long-term load data;

[0061] A global optimization module is used to initialize a set of adjustment strategies, perform fitness calculation and global optimal iteration on the set of adjustment strategies according to the real-time sensor data acquired in advance and the analyzed load data, and obtain a primary adjustment strategy;

[0062] The load regulation module is used to construct the state space and update the state feedback of the primary regulation strategy according to the real-time sensor data and the analyzed load data, obtain the standard regulation strategy, and use the standard regulation strategy to regulate the load of the central air conditioner.

[0063] In order to solve the above problem, the present invention further provides an electronic device, the electronic device comprising:

[0064] at least one processor; and,

[0065] a memory communicatively connected to the at least one processor; wherein,

[0066] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the above-mentioned central air-conditioning refrigeration system load flexible adjustment method.

[0067] The present invention can realize nonlinear load analysis by combining multimodal sensor feature data through multimodal feature extraction and short-term load analysis, thereby performing short-term analysis on the load data of the central air conditioner and improving the accuracy of the load analysis. By extracting a significant lag coefficient group, the correlation of the sequence itself can be measured according to the mutual dependence between the data time steps, thereby discovering the periodicity and trend correlation of the data. By performing multi-scale dilated convolution and bidirectional time series feature extraction, bidirectional time series feature extraction can be realized under the premise of expanding the field of view, thereby improving the representation of the time series features and the accuracy of the load analysis.

[0068] By performing long-term load analysis, it is possible to combine autoregressive linear analysis and self-attention models to achieve more accurate long-term load data analysis, thereby improving the accuracy of load data analysis and the accuracy of subsequent load regulation. By performing fitness calculation and global optimal iteration, a primary regulation strategy is obtained, and the regulation strategy can be gradually adjusted to a strategy method that meets load requirements and reduces power consumption, thereby achieving flexible load regulation and improving the efficiency of load regulation. By performing state space construction and state feedback updates, the control state of the central air conditioner can continuously converge to the optimal state, achieving flexible regulation of the load of the central air conditioning system, and improving the efficiency of load regulation. Therefore, the central air conditioning refrigeration system load flexible regulation method, device and equipment proposed in the present invention can solve the problem of low efficiency when performing load regulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 A schematic diagram of a flow chart of a method for flexible load adjustment of a central air-conditioning refrigeration system provided by an embodiment of the present invention;

[0070] Figure 2 A schematic flow chart of a method for extracting short-term load data provided by an embodiment of the present invention;

[0071] Figure 3 A schematic flow chart of a method for extracting a significant lag coefficient group provided by an embodiment of the present invention;

[0072] Figure 4 A functional module diagram of a central air-conditioning refrigeration system load flexible adjustment device provided by an embodiment of the present invention;

[0073] Figure 5 A schematic diagram of the structure of an electronic device of a central air-conditioning refrigeration system load flexible adjustment device provided by an embodiment of the present invention;

[0074] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0075] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0076] The embodiment of the present application provides a method for flexible load adjustment of a central air-conditioning refrigeration system. The execution subject of the method for flexible load adjustment of a central air-conditioning refrigeration system includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided in the embodiment of the present application. In other words, the method for flexible load adjustment of a central air-conditioning refrigeration system can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc. The server can be an independent server, or it can be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN), and big data and artificial intelligence platforms.

[0077] Reference Figure 1 FIG. 1 is a flow chart of a method for flexible load adjustment of a central air-conditioning refrigeration system provided by an embodiment of the present invention. In this embodiment, the method for flexible load adjustment of a central air-conditioning refrigeration system includes:

[0078] S1. Perform multimodal feature extraction and short-term load analysis on the pre-acquired historical sensor data sequence to obtain short-term load data.

[0079] In detail, each historical sensor data in the historical sensor data sequence refers to a data sequence obtained by sensor monitoring and recording of data such as the load data of the central air conditioner, environmental sensor data, and operating status data of the air conditioner within a past time interval, and each historical sensor data in the historical sensor data sequence is sorted in chronological order.

[0080] Specifically, the central air conditioner is an air conditioning system composed of one or more cold and heat source systems and multiple air conditioning systems. The central air conditioner adopts the principle of liquid vaporization refrigeration to provide the required cooling capacity for the air conditioning system to offset the heat load of the indoor environment, and adopts a heating system to provide the required heat for the air conditioning system to offset the cold and heating loads of the indoor environment.

[0081] In detail, the load data includes load data of components such as the compressor and fan of the central air conditioner, the environmental sensor data refers to data such as the temperature and humidity of the space where the central air conditioner is located, and the operating status data includes data such as the operating power consumption and operating mode of components such as the compressor and fan of the central air conditioner.

[0082] In the embodiment of the present invention, refer to Figure 2As shown, the multimodal feature extraction and short-term load analysis are performed on the pre-acquired historical sensor data sequence to obtain short-term load data, including:

[0083] S21, performing format unification and data normalization operations on the pre-acquired historical sensor data sequence to obtain a standard sensor data sequence;

[0084] S22, performing multimodal feature extraction on the standard sensor data sequence to obtain a historical load feature sequence group;

[0085] S23, calculating the load correlation coefficient between each historical load characteristic sequence in the historical load characteristic sequence group;

[0086] S24, performing correlation dimension reduction on the historical load feature sequence group according to the load correlation coefficient to obtain a reduced dimension load feature sequence group;

[0087] S25, performing principal component dimensionality reduction and short-term load analysis on the reduced-dimensionality load feature sequence group to obtain short-term load data.

[0088] In detail, the format unification refers to unifying the data of the same data type in the historical sensor data sequence into the same data unit and data format, for example, unifying the power unit into watts, and the data normalization operation can be performed using a softmax or sigmoid function.

[0089] Specifically, the multimodal feature extraction refers to extracting data features of different modes in the historical sensor data sequence, and the historical load feature sequence group includes a historical load feature sequence composed of load data, a historical sensor feature sequence composed of environmental sensor data, and a historical operation feature sequence composed of air conditioning operation status data.

[0090] In detail, the load correlation coefficient is calculated using the following load correlation algorithm:

[0091]

[0092] Among them, p(C i ,C j ) refers to the historical load characteristic sequence C in the historical load characteristic sequence group i and historical load characteristic sequence C j The load correlation coefficient, i, j are the serial numbers of the historical load characteristic sequence in the historical load characteristic sequence group, K i is the historical load characteristic sequence C i The total number of features, K j is the historical load characteristic sequence C jThe total number of features, k and l are the feature numbers, cov() is the covariance operator, σ() is the standard deviation operator, x ik Refers to the historical load characteristic sequence C i The kth feature in w ik is the historical load characteristic sequence C i The prior weight of the kth feature in , x jl Refers to the historical load characteristic sequence C j The lth feature in w jl is the historical load characteristic sequence C j The prior weight of the lth feature in .

[0093] Specifically, the load correlation algorithm ensures the independence of the correlation calculation logic, and the weights are derived from the prior information of the data itself, so that the weighted average result can effectively characterize the correlation between features in multimodal fusion.

[0094] In detail, the priori weight is based on the feature priority corresponding to each historical load feature sequence in the historical load feature sequence group. If some features or modes are more important to the target, higher priori weights can be assigned before fusion.

[0095] Specifically, the correlation dimension reduction refers to reducing the dimension of the historical load feature sequences in the historical load feature sequence group according to the load correlation coefficient. For example, when the load correlation coefficients of two historical load feature sequences are close to 1 or -1, one of them can be removed to reduce redundancy.

[0096] In detail, the performing principal component dimensionality reduction and short-term load analysis on the dimensionality reduction load feature sequence group to obtain short-term load data includes:

[0097] Performing a matrix operation on the reduced-dimensionality load feature sequence group to obtain a reduced-dimensionality load feature matrix;

[0098] Performing covariance matrix calculation on the reduced-dimensional load characteristic matrix to obtain a load characteristic relationship matrix;

[0099] Performing eigenvalue decomposition on the load characteristic relationship matrix to obtain a load characteristic vector group and a load characteristic value group;

[0100] Performing principal component screening on the load characteristic vector group according to the load characteristic value group to obtain a principal component load characteristic group;

[0101] Performing radial basis calculation on the principal component load feature group to obtain a radial basis load data group;

[0102] Performing activation response on the principal component load feature group to obtain an activation response weight group;

[0103] The radial basis load data set is weightedly fused according to the activation response weight set to obtain short-term load data.

[0104] Specifically, the matrix operation refers to splicing each reduced-dimensionality load feature sequence in the reduced-dimensionality load feature sequence group as a matrix row vector into a reduced-dimensionality load feature matrix, and the eigenvalue decomposition (Eigendecomposition), also known as spectral decomposition, is an algorithm for decomposing a matrix into a set of products of eigenvalues ​​and eigenvectors.

[0105] In detail, the principal component screening refers to selecting several load characteristic values ​​with larger values ​​in the load characteristic value group as the principal component load characteristic value group, and aggregating the load characteristic vectors corresponding to the principal component load characteristic value group in the load characteristic vector group into the principal component load characteristic group.

[0106] Specifically, the radial basis activation refers to performing radial basis activation using the hidden layer of a radial basis function neural network (RBF) pre-trained using a large amount of load data labeled for future short-term time periods and the corresponding principal component load feature group as an activation function, and the radial basis function can be a Gaussian kernel function.

[0107] In detail, the activation response refers to the node response value of the radial basis function neural network, and the response value represents the degree of matching between the input and the node center of the node of the radial basis function neural network. The closer the distance, the higher the response, and the farther the distance, the closer the response value is to zero. The weighted fusion refers to taking the corresponding weights of each activation in the activation response weight group as the weights of each radial basis load data in the radial basis load data group for weighted summation.

[0108] In the embodiment of the present invention, by performing multimodal feature extraction and short-term load analysis, nonlinear load analysis can be achieved by combining multimodal sensor feature data, thereby performing short-term analysis on the load data of the central air conditioner and improving the accuracy of the load analysis.

[0109] S2. Extracting a significant lag coefficient group from the historical sensor data sequence, and performing multi-scale dilated convolution and bidirectional time series feature extraction on the historical sensor data sequence according to the significant lag coefficient group to obtain a bidirectional sensor time series feature sequence.

[0110] In detail, each significant lag coefficient in the significant lag coefficient group refers to a more significant lag coefficient in the historical sensor data sequence, and the lag coefficient refers to the span of each lag position. For example, lag coefficient 1 represents the relationship between adjacent points, and lag coefficient 2 represents the relationship every 1 time step. The bidirectional sensing timing feature sequence is a timing feature sequence obtained by performing forward and backward timing feature analysis on the historical sensor data sequence.

[0111] In the embodiment of the present invention, refer to Figure 3 As shown, the extracting of a significant lag coefficient group from the historical sensor data sequence includes:

[0112] S31, performing self-attention encoding on the historical sensor data sequence to obtain attention sensor weights;

[0113] S32, performing lag coefficient analysis on the historical sensor data sequence according to the attention sensor weight to obtain a sensor lag coefficient group;

[0114] S33, performing autocorrelation analysis on the historical sensor data sequence according to the sensor lag coefficient group to obtain a sensor autocorrelation array;

[0115] S34. Filter out a significant lag coefficient group from the sensor lag coefficient group according to the sensor self-correlation array.

[0116] In detail, the self-attention mechanism can be used to perform attention encoding to obtain attention sensor weights, which are the attention weights of each historical sensor data in the historical sensor data sequence. The lag coefficient analysis refers to screening out several serial numbers with larger values ​​in the attention sensor weights, calculating the sensor lag coefficient according to the difference between the serial numbers, and gathering all the sensor lag coefficients into a sensor lag coefficient group.

[0117] Specifically, the following sensor autocorrelation algorithm is used to perform autocorrelation analysis to obtain a sensor autocorrelation array:

[0118]

[0119] Among them, ρ r is the rth sensor autocorrelation coefficient in the sensor hysteresis coefficient group, r is the coefficient number, U r is the rth sensor hysteresis coefficient in the sensor hysteresis coefficient group, t is the data sequence number, T is the length of the historical sensor data sequence, x t is the tth historical sensor data in the historical sensor data sequence, x t+r It is the t+rth historical sensor data in the historical sensor data sequence.

[0120] In detail, the sensor autocorrelation algorithm can measure the correlation of the sequence itself according to the interdependence between the time steps in the historical sensor data sequence, so as to discover the periodicity and trend correlation of the data. The screening out of the significant lag coefficient group from the sensor lag coefficient group according to the sensor autocorrelation array means taking the sensor autocorrelation coefficients in the sensor autocorrelation array that are greater than a preset threshold as the significant autocorrelation array, and taking the sensor lag coefficients corresponding to the significant autocorrelation array in the sensor lag coefficient group as the significant sensor lag coefficient group.

[0121] Specifically, performing multi-scale dilated convolution and bidirectional time series feature extraction on the historical sensor data sequence according to the significant lag coefficient group to obtain a bidirectional sensor time series feature sequence includes:

[0122] Perform expansion rate configuration according to the significant hysteresis coefficient group to obtain an expansion rate group;

[0123] Initializing a multi-scale dilated convolution kernel group according to the dilation rate group;

[0124] Performing dilated convolution on the historical sensor data sequence using the multi-scale dilated convolution kernel group to obtain a dilated sensor feature sequence group;

[0125] Performing feature fusion on the expanded sensing feature sequence group to obtain a multi-scale sensing feature sequence;

[0126] Performing forward time series feature extraction on the multi-scale sensing feature sequence to obtain a forward sensing time series feature sequence;

[0127] Performing backward time series feature extraction on the multi-scale sensing feature sequence to obtain a backward sensing time series feature sequence;

[0128] The forward time series feature sequence and the backward time series feature sequence are feature concatenated to obtain a bidirectional sensing time series feature sequence.

[0129] In detail, the expansion rate configuration refers to configuring the corresponding expansion rate according to the numerical size of each significant lag coefficient in the significant lag coefficient group, using a smaller expansion rate at a point with a smaller numerical size, and a larger expansion rate at a point with a larger numerical size. For example, when the significant lag coefficient group is 1, 2, 4, 8, and 16, the expansion rate in the interval corresponding to 1, 2, and 4 is set to 2, and the expansion rate in the interval corresponding to 8 and 16 is set to 10. The expansion rate can also be set to a multiple of the numerical size of the significant lag coefficient.

[0130] Specifically, each multi-scale dilated convolution kernel in the multi-scale dilated convolution kernel group corresponds to a dilation rate in the dilation rate group, and the dilation rate corresponds to the spacing between values ​​when the multi-scale dilated convolution kernel processes data. When the dilation rate is 1, the dilated convolution degenerates into a standard convolution. When the dilation rate is greater than 1, a certain number of zeros are inserted between the elements of the convolution kernel, thereby increasing the size of the receptive field. For example, when the dilation rate of a 3x3 convolution kernel is 2, its receptive field is equivalent to that of a 5x5 convolution kernel, but the number of parameters and the amount of calculation remain unchanged.

[0131] In detail, the output features of convolution kernels with different expansion rates can be concatenated or added to achieve feature fusion of the expanded sensing feature sequence group, the forward long short-term memory layer of the bidirectional long short-term memory network (Bidirectional Long Short-Term Memory, referred to as BiLSTM) can be used to perform forward time series feature extraction, and the backward long short-term memory layer of the bidirectional long short-term memory network can be used to perform backward time series feature extraction.

[0132] In an embodiment of the present invention, by extracting a significant lag coefficient group, the correlation of the sequence itself can be measured according to the interdependence between the data time steps, thereby discovering the periodicity and trend correlation of the data. By performing multi-scale dilated convolution and bidirectional time series feature extraction, bidirectional time series feature extraction can be achieved while expanding the field of view, thereby improving the representation of the time series features and the accuracy of load analysis.

[0133] S3. Perform a long-term load analysis on the historical sensing data sequence according to the bidirectional sensing time series feature sequence to obtain long-term load data, and generate analysis load data according to the short-term load data and the long-term load data.

[0134] In detail, the long-term load data refers to the load data sequence of the central air conditioner in the future long-term period analyzed according to the bidirectional sensing time series feature sequence.

[0135] The long-term load analysis refers to performing a fully connected output prediction on the bidirectional sensing time series feature sequence, that is, multiplying the bidirectional sensing time series feature sequence by a preset weight of the fully connected layer, and adding the result to the bias of the fully connected layer to obtain long-term load data.

[0136] In the embodiment of the present invention, the long-term load analysis is performed on the historical sensor data sequence according to the bidirectional sensor time series feature sequence to obtain the long-term load data, including:

[0137] Performing self-attention encoding on the bidirectional sensing temporal feature sequence to obtain a bidirectional self-attention weight;

[0138] Using the bidirectional self-attention weight, a weighted load analysis is performed on the bidirectional sensing time series feature sequence to obtain attention long-term load data;

[0139] Performing autoregressive analysis on the bidirectional sensing time series feature sequence to obtain an autoregressive sensing data sequence;

[0140] Performing residual calculation on the historical sensor data sequence and the autoregressive sensor data sequence to obtain a residual sensor data sequence;

[0141] Perform multi-step decoding on the bidirectional sensing time series feature sequence according to the residual sensing data sequence to obtain decoded long-term load data;

[0142] The attention long-term load data and the decoding long-term load data are weightedly fused to obtain long-term load data.

[0143] In detail, the method of self-attention encoding is the same as the method of self-attention encoding in the above step S2, which will not be repeated here. The weighted load analysis refers to multiplying the bidirectional self-attention weight by the bidirectional sensing timing feature sequence, and adding the result to the preset bias parameter of the fully connected network to obtain the long-term attention load data.

[0144] Specifically, an autoregressive integrated moving average model (ARIMA) of a large number of bidirectional sensor time series feature sequences that are pre-labeled with long-term load data can be used for autoregressive analysis. The autoregressive sensor data sequence is a sensor data sequence corresponding to the historical sensor data sequence obtained by linear prediction. The multi-step decoding refers to using the residual sensor data sequence as a model loss value to adjust the model parameters of the autoregressive model, and using the iterated autoregressive model to step by step perform autoregressive analysis.

[0145] In detail, generating analytical load data based on the short-term load data and the long-term load data refers to sorting and splicing the short-term load data and the long-term load data in timestamp order to obtain analytical load data. For example, the short-term load data is the load data within the next three minutes, and the long-term load data is the load data within the next three to thirty minutes. The two are spliced ​​and combined to obtain the load data within the next thirty minutes.

[0146] In the embodiment of the present invention, by performing long-term load analysis, it is possible to combine autoregressive linear analysis and self-attention models to achieve more accurate long-term load data analysis, thereby improving the accuracy of load data analysis and the accuracy of subsequent load regulation.

[0147] S4, initializing the adjustment strategy set, performing fitness calculation and global optimal iteration on the adjustment strategy set according to the pre-acquired real-time sensor data and the analyzed load data, and obtaining a primary adjustment strategy.

[0148] In detail, each adjustment strategy in the adjustment strategy set corresponds to a control parameter sequence of the central air conditioner in a future time period. The control parameters include compressor power, fan speed, cooling system water pump flow and other parameters. The adjustment strategy set can be initialized by a random number model or a pseudo-random number model.

[0149] Specifically, the real-time sensor data is data obtained by recording sensor monitoring of data such as load data of the central air conditioner, environmental sensor data, and operating status data of the air conditioner at the current time.

[0150] In the embodiment of the present invention, the fitness calculation and global optimal iteration of the adjustment strategy set are performed according to the pre-acquired real-time sensor data and the analyzed load data to obtain the primary adjustment strategy, including:

[0151] Selecting the regulation strategies in the regulation strategy set one by one as the target regulation strategy, performing load analysis on the real-time sensor data according to the target regulation strategy, and obtaining regulation load data;

[0152] Extracting real-time temperature data from the real-time sensor data, performing thermodynamic analysis on the target adjustment strategy according to the real-time temperature data, and obtaining target cooling capacity data;

[0153] Calculating the fitness corresponding to the target adjustment strategy according to the adjustment load data, the target cooling capacity data and the analysis load data;

[0154] The fitness of all target adjustment strategies in the adjustment strategy set is aggregated into a fitness set, and the adjustment strategy set is globally optimally iterated according to the fitness set to obtain a primary adjustment strategy.

[0155] In detail, performing load analysis on the real-time sensor data according to the target adjustment strategy to obtain adjusted load data means extracting real-time load data from the real-time sensor data, performing load analysis according to the target adjustment strategy and various devices of the central air-conditioning, and obtaining the power consumption value of each device after adjustment according to the target adjustment strategy in a future time period.

[0156] Specifically, the thermodynamic analysis refers to calculating the total cooling value of the space where the central air conditioner is located based on the temperature at each moment corresponding to the target adjustment strategy and the temperature difference between the real-time temperature data, combined with the flow data of each device in the target adjustment strategy.

[0157] In detail, the fitness corresponding to the target adjustment strategy is calculated using the following fitness algorithm:

[0158]

[0159] Wherein, F refers to the fitness, w1 is the preset load weight coefficient, N is the total time length of the load data analysis, n is the time index, is the load value at the nth moment in the analyzed load data, M is the total number of devices corresponding to the regulated load data, m is the device index, It refers to the power consumption value of the mth device in the load regulation data at the nth time, v m refers to the efficiency coefficient of the mth device in the load regulation data, || is the absolute value symbol, w2 is the preset energy efficiency weight coefficient, Q m It refers to the cooling capacity value corresponding to the mth device in the target cooling capacity data.

[0160] Specifically, the fitness algorithm can determine the contribution of each regulation strategy in terms of flexible regulation and energy efficiency in combination with the load variation degree and energy efficiency, and reduce power consumption as much as possible while meeting the load demand to improve the energy efficiency of the system.

[0161] In detail, the globally optimal iteration of the adjustment strategy set according to the fitness set to obtain the primary adjustment strategy includes:

[0162] Performing minimum value screening on the fitness set to obtain extreme value fitness;

[0163] Determining whether the extreme value fitness is less than a preset fitness threshold;

[0164] If not, the adjustment strategy corresponding to the extreme fitness in the adjustment strategy set is used as the extreme adjustment strategy;

[0165] Performing fitness update on the adjustment strategy set according to the fitness set and the extreme value adjustment strategy to obtain an updated adjustment strategy set;

[0166] Performing crossover mutation on the update adjustment strategy set to obtain a mutation adjustment strategy set;

[0167] The adjustment strategy set is updated by using the variant adjustment strategy set, and the step of returning to selecting adjustment strategies in the adjustment strategy set one by one as target adjustment strategies is performed;

[0168] If so, the adjustment strategy corresponding to the extreme fitness in the adjustment strategy set is used as the primary adjustment strategy.

[0169] Specifically, updating the fitness of the adjustment strategy set according to the fitness set and the extreme value adjustment strategy to obtain the updated adjustment strategy set means using a tracking algorithm of a tuna school optimization algorithm to track and update the extreme value adjustment strategy according to the fitness of each adjustment strategy in the adjustment strategy set.

[0170] In detail, the crossover mutation refers to performing a crossover operation and a mutation operation, wherein the crossover operation refers to exchanging some policy parameters between two adjustment strategies, and the mutation operation helps the adjustment strategy to escape from the local optimum by randomly adjusting the policy parameters of the adjustment strategy.

[0171] In the embodiment of the present invention, a primary regulation strategy is obtained by performing fitness calculation and global optimal iteration, and the regulation strategy can be gradually adjusted to a strategy method that meets load demand and reduces power consumption, thereby achieving flexible load regulation and improving load regulation efficiency.

[0172] S5. Constructing the state space and updating the state feedback of the primary regulation strategy according to the real-time sensing data and the analyzed load data to obtain a standard regulation strategy, and using the standard regulation strategy to regulate the load of the central air conditioner.

[0173] In detail, the state space construction and state feedback update of the primary regulation strategy are performed according to the real-time sensor data and the analyzed load data to obtain a standard regulation strategy, including:

[0174] Constructing a state space according to the real-time sensor data and the primary adjustment strategy to obtain an air conditioning load state space;

[0175] Calculating a cost matrix of the air conditioning load state space according to the analyzed load data;

[0176] Solving the cost matrix for gain to obtain a feedback gain matrix;

[0177] The primary regulation strategy is feedback updated according to the gain feedback matrix to obtain a standard regulation strategy.

[0178] In detail, the state equation and the output equation can be used to construct the state space, that is, the parameters of the primary regulation strategy are used as control variables, and the load data in the real-time sensor data is used as the system state to generate the air conditioning load state space.

[0179] Specifically, the cost matrix of the air conditioning load state space can be calculated using a quadratic cost function, that is, the offset between the load data corresponding to the air conditioning load state space and the analysis load data can be calculated, and the discrete Riccati equation can be used to solve the gain to obtain a feedback gain matrix.

[0180] In detail, the feedback update refers to multiplying the gain feedback matrix by the system state to obtain the standard adjustment strategy, and using the standard adjustment strategy to load adjust the central air conditioner refers to controlling and adjusting the central air conditioner according to the control parameters such as compressor power, fan speed, cooling system water pump flow at each moment in the standard adjustment strategy.

[0181] In the embodiment of the present invention, by constructing the state space and updating the state feedback, the control state of the central air conditioner can be continuously converged to the optimal state, thereby realizing the flexible regulation of the load of the central air conditioner system and improving the efficiency of load regulation.

[0182] The present invention can realize nonlinear load analysis by combining multimodal sensor feature data through multimodal feature extraction and short-term load analysis, thereby performing short-term analysis on the load data of the central air conditioner and improving the accuracy of the load analysis. By extracting a significant lag coefficient group, the correlation of the sequence itself can be measured according to the mutual dependence between the data time steps, thereby discovering the periodicity and trend correlation of the data. By performing multi-scale dilated convolution and bidirectional time series feature extraction, bidirectional time series feature extraction can be realized under the premise of expanding the field of view, thereby improving the representation of the time series features and the accuracy of the load analysis.

[0183] By performing long-term load analysis, it is possible to combine autoregressive linear analysis and self-attention models to achieve more accurate long-term load data analysis, thereby improving the accuracy of load data analysis and the accuracy of subsequent load regulation. By performing fitness calculation and global optimal iteration, a primary regulation strategy is obtained, and the regulation strategy can be gradually adjusted to a strategy method that meets load requirements and reduces power consumption, thereby achieving flexible load regulation and improving the efficiency of load regulation. By performing state space construction and state feedback updates, the control state of the central air conditioner can continuously converge to the optimal state, achieving flexible regulation of the load of the central air conditioning system, and improving the efficiency of load regulation. Therefore, the central air conditioning refrigeration system load flexible regulation method proposed in the present invention can solve the problem of low efficiency when performing load regulation.

[0184] like Figure 4 , which is a functional module diagram of a central air-conditioning refrigeration system load flexible adjustment device provided by an embodiment of the present invention.

[0185] The central air-conditioning refrigeration system load flexible adjustment device 100 of the present invention can be installed in an electronic device. According to the functions to be implemented, the central air-conditioning refrigeration system load flexible adjustment device 100 can include a short-term analysis module 101, a two-way analysis module 102, a long-term analysis module 103, a global optimization module 104 and a load adjustment module 105. The module of the present invention can also be called a unit, which refers to a series of computer program segments that can be executed by an electronic device processor and can complete fixed functions, which are stored in the memory of the electronic device.

[0186] In this embodiment, the functions of each module / unit are as follows:

[0187] The short-term analysis module 105 is used to perform multimodal feature extraction and short-term load analysis on the pre-acquired historical sensor data sequence to obtain short-term load data, wherein the multimodal feature extraction and short-term load analysis on the pre-acquired historical sensor data sequence to obtain short-term load data includes: performing format unification and data normalization operations on the pre-acquired historical sensor data sequence to obtain a standard sensor data sequence; performing multimodal feature extraction on the standard sensor data sequence to obtain a historical load feature sequence group; and calculating the load correlation coefficient between each historical load feature sequence in the historical load feature sequence group using the following load correlation algorithm:

[0188]

[0189] Among them, p(C i ,C j ) refers to the historical load characteristic sequence C in the historical load characteristic sequence group i and historical load characteristic sequence C j The load correlation coefficient, i, j are the serial numbers of the historical load characteristic sequence in the historical load characteristic sequence group, K i is the historical load characteristic sequence C i The total number of features, K j is the historical load characteristic sequence C j The total number of features, k and l are the feature numbers, cov() is the covariance operator, σ() is the standard deviation operator, x ik Refers to the historical load characteristic sequence C i The kth feature in w ik is the historical load characteristic sequence C i The prior weight of the kth feature in , x jl Refers to the historical load characteristic sequence C j The lth feature in w jl is the historical load characteristic sequence C ja priori weight of the lth feature in; performing correlation dimension reduction on the historical load feature sequence group according to the load correlation coefficient to obtain a reduced dimension load feature sequence group; performing principal component dimension reduction and short-term load analysis on the reduced dimension load feature sequence group to obtain short-term load data;

[0190] The bidirectional analysis module 102 is used to extract a significant lag coefficient group from the historical sensor data sequence, and perform multi-scale dilated convolution and bidirectional time series feature extraction on the historical sensor data sequence according to the significant lag coefficient group to obtain a bidirectional sensor time series feature sequence;

[0191] The long-term analysis module 103 is used to perform long-term load analysis on the historical sensing data sequence according to the bidirectional sensing time series feature sequence to obtain long-term load data, and generate analysis load data according to the short-term load data and the long-term load data;

[0192] The global optimization module 104 is used to initialize the adjustment strategy set, perform fitness calculation and global optimal iteration on the adjustment strategy set according to the pre-acquired real-time sensor data and the analyzed load data, and obtain a primary adjustment strategy;

[0193] The load regulation module 105 is used to construct the state space and update the state feedback of the primary regulation strategy according to the real-time sensor data and the analyzed load data, obtain the standard regulation strategy, and use the standard regulation strategy to regulate the load of the central air conditioner.

[0194] In detail, each module described in the central air-conditioning refrigeration system load flexible adjustment device 100 in the embodiment of the present invention is used in the same manner as described above. Figure 1 The technical means are the same as the central air-conditioning refrigeration system load flexible adjustment method described in, and can produce the same technical effects, so they will not be repeated here.

[0195] like Figure 5 FIG. 1 is a schematic diagram of the structure of an electronic device for implementing a method for flexible load adjustment of a central air-conditioning refrigeration system provided by an embodiment of the present invention.

[0196] The electronic device 1 may include a processor 10, a memory 11, a communication bus 12 and a communication interface 13, and may also include a computer program stored in the memory 11 and executable on the processor 10, such as a central air conditioning refrigeration system load flexibility adjustment program.

[0197] In some embodiments, the processor 10 may be composed of an integrated circuit, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and combinations of various control chips. The processor 10 is the control core (ControlUnit) of the electronic device, and uses various interfaces and lines to connect various components of the entire electronic device, and executes or executes programs or modules stored in the memory 11 (for example, executing a flexible load adjustment program for a central air-conditioning refrigeration system, etc.), and calls data stored in the memory 11 to execute various functions of the electronic device and process data.

[0198] The memory 11 includes at least one type of readable storage medium, and the readable storage medium includes a flash memory, a mobile hard disk, a multimedia card, a card-type memory (for example, SD or DX memory, etc.), a magnetic memory, a disk, an optical disk, etc. In some embodiments, the memory 11 may be an internal storage unit of an electronic device, such as a mobile hard disk of the electronic device. In other embodiments, the memory 11 may also be an external storage device of an electronic device, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device. Further, the memory 11 may also include both an internal storage unit of the electronic device and an external storage device. The memory 11 can not only be used to store application software and various types of data installed in the electronic device, such as the code of the load flexible adjustment program of the central air-conditioning refrigeration system, but also be used to temporarily store data that has been output or is to be output.

[0199] The communication bus 12 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The bus is configured to realize connection and communication between the memory 11 and at least one processor 10, etc.

[0200] The communication interface 13 is used for communication between the above-mentioned electronic device and other devices, including a network interface and a user interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device and other electronic devices. The user interface may be a display (Display), an input unit (such as a keyboard (Keyboard)), and optionally, the user interface may also be a standard wired interface, a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an OLED (Organic Light-Emitting Diode, organic light-emitting diode) touch device, etc. Among them, the display may also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device and to display a visual user interface.

[0201] The figure only shows an electronic device with components. Those skilled in the art will understand that the structure shown in the figure does not constitute a limitation on the electronic device, and may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0202] For example, although not shown, the electronic device may also include a power source (such as a battery) for supplying power to each component. Preferably, the power source may be logically connected to the at least one processor 10 through a power management device, so that the power management device can realize functions such as charging management, discharging management, and power consumption management. The power source may also include one or more DC or AC power sources, recharging devices, power failure detection circuits, power converters or inverters, power status indicators, and other arbitrary components. The electronic device may also include a variety of sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.

[0203] It should be understood that the embodiment is for illustration only and the scope of the patent application is not limited to this structure.

[0204] The central air-conditioning refrigeration system load flexible adjustment program stored in the memory 11 of the electronic device 1 is a combination of multiple instructions. When running in the processor 10, it can achieve:

[0205] Perform multimodal feature extraction and short-term load analysis on the pre-acquired historical sensor data sequence to obtain short-term load data, wherein the multimodal feature extraction and short-term load analysis on the pre-acquired historical sensor data sequence to obtain short-term load data includes: performing format unification and data normalization operations on the pre-acquired historical sensor data sequence to obtain a standard sensor data sequence; performing multimodal feature extraction on the standard sensor data sequence to obtain a historical load feature sequence group; and calculating the load correlation coefficient between each historical load feature sequence in the historical load feature sequence group using the following load correlation algorithm:

[0206]

[0207] Among them, p(C i ,C j ) refers to the historical load characteristic sequence C in the historical load characteristic sequence group i and historical load characteristic sequence C j The load correlation coefficient, i, j are the serial numbers of the historical load characteristic sequence in the historical load characteristic sequence group, K i is the historical load characteristic sequence C i The total number of features, K j is the historical load characteristic sequence C j The total number of features, k and l are the feature numbers, cov() is the covariance operator, σ() is the standard deviation operator, x ik Refers to the historical load characteristic sequence C i The kth feature in w ik is the historical load characteristic sequence C i The prior weight of the kth feature in , x jl Refers to the historical load characteristic sequence C j The lth feature in w jl is the historical load characteristic sequence C j a priori weight of the lth feature in; performing correlation dimension reduction on the historical load feature sequence group according to the load correlation coefficient to obtain a reduced dimension load feature sequence group; performing principal component dimension reduction and short-term load analysis on the reduced dimension load feature sequence group to obtain short-term load data;

[0208] Extracting a significant lag coefficient group from the historical sensor data sequence, and performing multi-scale dilated convolution and bidirectional time series feature extraction on the historical sensor data sequence according to the significant lag coefficient group to obtain a bidirectional sensor time series feature sequence;

[0209] Performing a long-term load analysis on the historical sensing data sequence according to the bidirectional sensing time series feature sequence to obtain long-term load data, and generating analytical load data according to the short-term load data and the long-term load data;

[0210] Initializing a regulation strategy set, performing fitness calculation and global optimal iteration on the regulation strategy set according to the pre-acquired real-time sensor data and the analyzed load data, and obtaining a primary regulation strategy;

[0211] The state space of the primary regulation strategy is constructed and the state feedback is updated according to the real-time sensing data and the analyzed load data to obtain a standard regulation strategy, and the standard regulation strategy is used to regulate the load of the central air conditioner.

[0212] Specifically, the specific implementation method of the processor 10 for the above instructions can refer to the description of the relevant steps in the corresponding embodiment of the accompanying drawings, which will not be repeated here.

[0213] Furthermore, if the module / unit integrated in the electronic device 1 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. The computer-readable storage medium can be volatile or non-volatile. For example, the computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, and a read-only memory (ROM).

[0214] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0215] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0216] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0217] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0218] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the appended claims rather than the above description, so it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any attached figure mark in the claims should not be regarded as limiting the claims involved.

[0219] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application device that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0220] In addition, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. The multiple units or devices stated in the device embodiment can also be implemented by one unit or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any particular order.

[0221] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for flexible load adjustment of a central air-conditioning refrigeration system, characterized in that: The method comprises: Perform multimodal feature extraction and short-term load analysis on the pre-acquired historical sensor data sequence to obtain short-term load data, wherein the multimodal feature extraction and short-term load analysis on the pre-acquired historical sensor data sequence to obtain short-term load data includes: performing format unification and data normalization operations on the pre-acquired historical sensor data sequence to obtain a standard sensor data sequence; performing multimodal feature extraction on the standard sensor data sequence to obtain a historical load feature sequence group; and calculating the load correlation coefficient between each historical load feature sequence in the historical load feature sequence group using the following load correlation algorithm: Among them, p(C i ,C j ) refers to the historical load characteristic sequence C in the historical load characteristic sequence group i and historical load characteristic sequence C j The load correlation coefficient, i, j are the serial numbers of the historical load characteristic sequence in the historical load characteristic sequence group, K i is the historical load characteristic sequence C i The total number of features, K j is the historical load characteristic sequence C j The total number of features, k and l are the feature numbers, cov() is the covariance operator, σ() is the standard deviation operator, x ik Refers to the historical load characteristic sequence C i The kth feature in w ik is the historical load characteristic sequence C i The prior weight of the kth feature in , x jl Refers to the historical load characteristic sequence C j The lth feature in w jl is the historical load characteristic sequence C j a priori weight of the lth feature in; performing correlation dimension reduction on the historical load feature sequence group according to the load correlation coefficient to obtain a reduced dimension load feature sequence group; performing principal component dimension reduction and short-term load analysis on the reduced dimension load feature sequence group to obtain short-term load data; Extracting a significant lag coefficient group from the historical sensor data sequence, and performing multi-scale dilated convolution and bidirectional time series feature extraction on the historical sensor data sequence according to the significant lag coefficient group to obtain a bidirectional sensor time series feature sequence; Performing a long-term load analysis on the historical sensing data sequence according to the bidirectional sensing time series feature sequence to obtain long-term load data, and generating analytical load data according to the short-term load data and the long-term load data; Initializing a regulation strategy set, performing fitness calculation and global optimal iteration on the regulation strategy set according to the pre-acquired real-time sensor data and the analyzed load data, and obtaining a primary regulation strategy; The state space of the primary regulation strategy is constructed and the state feedback is updated according to the real-time sensing data and the analyzed load data to obtain a standard regulation strategy, and the standard regulation strategy is used to regulate the load of the central air conditioner.

2. The method for flexible load adjustment of a central air-conditioning refrigeration system according to claim 1, characterized in that: The performing principal component dimensionality reduction and short-term load analysis on the dimensionality reduction load feature sequence group to obtain short-term load data includes: Performing a matrix operation on the reduced-dimensionality load feature sequence group to obtain a reduced-dimensionality load feature matrix; Performing covariance matrix calculation on the reduced-dimensional load characteristic matrix to obtain a load characteristic relationship matrix; Performing eigenvalue decomposition on the load characteristic relationship matrix to obtain a load characteristic vector group and a load characteristic value group; Performing principal component screening on the load characteristic vector group according to the load characteristic value group to obtain a principal component load characteristic group; Performing radial basis calculation on the principal component load feature group to obtain a radial basis load data group; Performing activation response on the principal component load feature group to obtain an activation response weight group; The radial basis load data set is weightedly fused according to the activation response weight set to obtain short-term load data.

3. The method for flexible load adjustment of a central air-conditioning refrigeration system according to claim 1, characterized in that: The step of extracting a significant lag coefficient group from the historical sensor data sequence comprises: Performing self-attention encoding on the historical sensor data sequence to obtain attention sensor weights; Performing hysteresis coefficient analysis on the historical sensor data sequence according to the attention sensor weight to obtain a sensor hysteresis coefficient group; The following sensor autocorrelation algorithm is used to perform autocorrelation analysis on the historical sensor data sequence according to the sensor lag coefficient group to obtain a sensor autocorrelation array: Among them, ρ r is the rth sensor autocorrelation coefficient in the sensor hysteresis coefficient group, r is the coefficient number, U r is the rth sensor hysteresis coefficient in the sensor hysteresis coefficient group, t is the data sequence number, T is the length of the historical sensor data sequence, x t is the tth historical sensor data in the historical sensor data sequence, x t+r is the t+rth historical sensor data in the historical sensor data sequence; A significant lag coefficient group is screened out from the lag coefficient group according to the lag coefficient array.

4. The method for flexible load adjustment of a central air-conditioning refrigeration system according to claim 1, characterized in that: The method of performing multi-scale dilated convolution and bidirectional time series feature extraction on the historical sensor data sequence according to the significant lag coefficient group to obtain a bidirectional sensor time series feature sequence includes: Perform expansion rate configuration according to the significant hysteresis coefficient group to obtain an expansion rate group; Initializing a multi-scale dilated convolution kernel group according to the dilation rate group; Performing dilated convolution on the historical sensor data sequence using the multi-scale dilated convolution kernel group to obtain a dilated sensor feature sequence group; Performing feature fusion on the expanded sensing feature sequence group to obtain a multi-scale sensing feature sequence; Performing forward time series feature extraction on the multi-scale sensing feature sequence to obtain a forward sensing time series feature sequence; Performing backward time series feature extraction on the multi-scale sensing feature sequence to obtain a backward sensing time series feature sequence; The forward time series feature sequence and the backward time series feature sequence are feature concatenated to obtain a bidirectional sensing time series feature sequence.

5. The method for flexible load adjustment of a central air-conditioning refrigeration system according to claim 1, characterized in that: The performing long-term load analysis on the historical sensor data sequence according to the bidirectional sensor time series feature sequence to obtain long-term load data includes: Performing self-attention encoding on the bidirectional sensing temporal feature sequence to obtain a bidirectional self-attention weight; Using the bidirectional self-attention weight, a weighted load analysis is performed on the bidirectional sensing time series feature sequence to obtain attention long-term load data; Performing autoregressive analysis on the bidirectional sensing time series feature sequence to obtain an autoregressive sensing data sequence; Performing residual calculation on the historical sensor data sequence and the autoregressive sensor data sequence to obtain a residual sensor data sequence; Perform multi-step decoding on the bidirectional sensing time series feature sequence according to the residual sensing data sequence to obtain decoded long-term load data; The attention long-term load data and the decoding long-term load data are weightedly fused to obtain long-term load data.

6. The method for flexible load adjustment of a central air-conditioning refrigeration system according to claim 1, characterized in that: The step of performing fitness calculation and global optimal iteration on the adjustment strategy set according to the pre-acquired real-time sensor data and the analyzed load data to obtain a primary adjustment strategy includes: Selecting the regulation strategies in the regulation strategy set one by one as the target regulation strategy, performing load analysis on the real-time sensor data according to the target regulation strategy, and obtaining regulation load data; Extracting real-time temperature data from the real-time sensor data, performing thermodynamic analysis on the target adjustment strategy according to the real-time temperature data, and obtaining target cooling capacity data; The fitness corresponding to the target adjustment strategy is calculated according to the adjustment load data, the target cooling capacity data and the analysis load data using the following fitness algorithm: Wherein, F refers to the fitness, w1 is the preset load weight coefficient, N is the total time length of the load data analysis, n is the time index, P n is the load value at the nth moment in the analyzed load data, M is the total number of devices corresponding to the regulated load data, m is the device index, It refers to the power consumption value of the mth device in the load regulation data at the nth time, v m refers to the efficiency coefficient of the mth device in the load regulation data, || is the absolute value symbol, w2 is the preset energy efficiency weight coefficient, Q m It refers to the cooling capacity value corresponding to the mth device in the target cooling capacity data; The fitness of all target adjustment strategies in the adjustment strategy set is aggregated into a fitness set, and the adjustment strategy set is globally optimally iterated according to the fitness set to obtain a primary adjustment strategy.

7. The method for flexible load adjustment of a central air-conditioning refrigeration system according to claim 6, characterized in that: The globally optimal iteration of the adjustment strategy set according to the fitness set to obtain a primary adjustment strategy includes: Performing minimum value screening on the fitness set to obtain extreme value fitness; Determining whether the extreme value fitness is less than a preset fitness threshold; If not, the adjustment strategy corresponding to the extreme fitness in the adjustment strategy set is used as the extreme adjustment strategy; Performing fitness update on the adjustment strategy set according to the fitness set and the extreme value adjustment strategy to obtain an updated adjustment strategy set; Performing crossover mutation on the update adjustment strategy set to obtain a mutation adjustment strategy set; The adjustment strategy set is updated by using the variant adjustment strategy set, and the step of returning to selecting adjustment strategies in the adjustment strategy set one by one as target adjustment strategies is performed; If so, the adjustment strategy corresponding to the extreme fitness in the adjustment strategy set is used as the primary adjustment strategy.

8. The method for flexible load adjustment of a central air-conditioning refrigeration system according to claim 1, characterized in that: The step of constructing the state space and updating the state feedback of the primary regulation strategy according to the real-time sensor data and the analyzed load data to obtain a standard regulation strategy includes: Constructing a state space according to the real-time sensor data and the primary adjustment strategy to obtain an air conditioning load state space; Calculating a cost matrix of the air conditioning load state space according to the analyzed load data; Solving the cost matrix for gain to obtain a feedback gain matrix; The primary regulation strategy is feedback updated according to the gain feedback matrix to obtain a standard regulation strategy.

9. A central air conditioning refrigeration system load flexible adjustment device, characterized in that: The device comprises: The short-term analysis module is used to perform multimodal feature extraction and short-term load analysis on the pre-acquired historical sensor data sequence to obtain short-term load data, wherein the multimodal feature extraction and short-term load analysis on the pre-acquired historical sensor data sequence to obtain short-term load data includes: performing format unification and data normalization operations on the pre-acquired historical sensor data sequence to obtain a standard sensor data sequence; performing multimodal feature extraction on the standard sensor data sequence to obtain a historical load feature sequence group; and calculating the load correlation coefficient between each historical load feature sequence in the historical load feature sequence group using the following load correlation algorithm: Among them, p(C i ,C j ) refers to the historical load characteristic sequence C in the historical load characteristic sequence group i and historical load characteristic sequence C j The load correlation coefficient, i, j are the serial numbers of the historical load characteristic sequence in the historical load characteristic sequence group, K i is the historical load characteristic sequence C i The total number of features, K j is the historical load characteristic sequence C j The total number of features, k and l are the feature numbers, cov() is the covariance operator, σ() is the standard deviation operator, x ik Refers to the historical load characteristic sequence C i The kth feature in w ik is the historical load characteristic sequence C i The prior weight of the kth feature in , x jl Refers to the historical load characteristic sequence C j The lth feature in w jl is the historical load characteristic sequence C j a priori weight of the lth feature in; performing correlation dimension reduction on the historical load feature sequence group according to the load correlation coefficient to obtain a reduced dimension load feature sequence group; performing principal component dimension reduction and short-term load analysis on the reduced dimension load feature sequence group to obtain short-term load data; A bidirectional analysis module, used to extract a significant lag coefficient group from the historical sensor data sequence, and perform multi-scale dilated convolution and bidirectional time series feature extraction on the historical sensor data sequence according to the significant lag coefficient group to obtain a bidirectional sensor time series feature sequence; A long-term analysis module, configured to perform long-term load analysis on the historical sensing data sequence according to the bidirectional sensing time series feature sequence to obtain long-term load data, and generate analysis load data according to the short-term load data and the long-term load data; A global optimization module is used to initialize a set of adjustment strategies, perform fitness calculation and global optimal iteration on the set of adjustment strategies according to the real-time sensor data acquired in advance and the analyzed load data, and obtain a primary adjustment strategy; The load regulation module is used to construct the state space and update the state feedback of the primary regulation strategy according to the real-time sensor data and the analyzed load data, obtain the standard regulation strategy, and use the standard regulation strategy to regulate the load of the central air conditioner.

10. An electronic device, characterized in that: The electronic device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the central air-conditioning refrigeration system load flexible adjustment method as described in any one of claims 1 to 8.

Images