An energy-saving control method and system based on spatial form of building group

Abstract

The application discloses an energy-saving control method and system based on building group space form, relates to the technical field of intelligent building energy-saving control, and comprises the following steps: inputting a building form feature parameter matrix into a multi-scale generative adversarial network, respectively extracting overall layout features, local spacing features and surface detail features of the building group, and generating an energy consumption probability distribution heat map corresponding to the space distribution of the building group; inputting the energy consumption probability distribution heat map into a discriminator of the multi-scale generative adversarial network, calculating the optimization gradient of each building form parameter, and combining a dynamic constraint function to generate a form adjustment instruction set. The application not only improves the accuracy of energy consumption prediction, but also identifies high energy consumption areas and provides targeted energy-saving measures. The problems that the energy consumption prediction accuracy is limited and it is difficult to efficiently convert into operable energy-saving measures in the prior art are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent building energy-saving control, and in particular to an energy-saving control method and system based on spatial form of building group. BACKGROUND

[0002] With the acceleration of global urbanization, building energy consumption has become one of the key factors affecting energy consumption and environmental protection. Early research mainly focused on the optimization of single building design, such as reducing energy consumption by improving building materials and optimizing building shape design. In recent years, with the development of computer technology, especially the progress of artificial intelligence algorithms, the research focus has gradually shifted to the analysis of the overall spatial layout and form characteristics of building groups, aiming to achieve more efficient energy-saving control strategies through intelligent means.

[0003] The prior art has the following problems: 1. Traditional methods often rely on static data for energy consumption evaluation, which is difficult to reflect the dynamic influence of changes in the internal and external environment of the building group in real time, resulting in limited energy consumption prediction accuracy. 2. In addition, current technical solutions are still insufficient in terms of how to efficiently convert complex building form characteristics into operable energy-saving measures. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides an energy-saving control method based on the spatial form of a building group to solve the problems of limited energy consumption prediction accuracy and difficulty in efficiently converting complex building form characteristics into operable energy-saving measures in the prior art.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the present application provides an energy-saving control method based on the spatial form of a building group, which includes collecting the three-dimensional spatial coordinates, volume, orientation angle and compound curvature parameters of each building in the building group, and constructing a building form feature parameter matrix; inputting the building form feature parameter matrix into a multi-scale generative adversarial network, respectively extracting the overall layout features, local spacing features and surface detail features of the building group, and generating an energy consumption probability distribution heat map corresponding to the spatial distribution of the building group; inputting the energy consumption probability distribution heat map into the discriminator of the multi-scale generative adversarial network, calculating the optimization gradient of each building form parameter, and generating a form adjustment instruction set in combination with a dynamic constraint function; distributing the form adjustment instruction set to each control terminal in the building group through a low-latency communication protocol, and dynamically adjusting the operating parameters of the equipment based on a multi-scale adaptive control algorithm.

[0008] As a preferred scheme of the building group space form-based energy-saving control method, the composite curvature parameter is the result of fusing micro-surface curvature, meso-distance curvature and macro layout curvature through a multi-scale fusion mechanism.

[0009] The micro-surface curvature is the ratio relationship of the ratio of the Gaussian curvature to the average curvature of the triangular mesh of the building surface.

[0010] The meso-distance curvature is the statistical dispersion characteristic of the length of the adjacent building Delaunay triangular partition edge.

[0011] The macro layout curvature is the eigenvalue difference degree relationship of the coordinate covariance matrix of the building group.

[0012] As a preferred scheme of the building group space form-based energy-saving control method, the generator of the multi-scale generative adversarial network is composed of a three-level pyramid hierarchical attention mechanism.

[0013] The macro layer adopts a decaying spatial attention mechanism.

[0014] The meso layer adopts a windowed local attention mechanism.

[0015] The micro layer adopts a curvature modulation multi-head attention mechanism.

[0016] As a preferred scheme of the building group space form-based energy-saving control method, the energy consumption probability distribution heat map is input into the discriminator of the multi-scale generative adversarial network to calculate the optimization gradient of each building form parameter, and the specific steps are as follows,

[0017] The energy consumption probability distribution heat map is normalized and processed by superimposing the building position mask.

[0018] The processed energy consumption probability distribution heat map is subjected to multi-scale convolution feature extraction and a confidence matrix is generated.

[0019] Based on the confidence matrix, the optimization gradient of each building form parameter is identified.

[0020] As a preferred scheme of the building group space form-based energy-saving control method, the form adjustment instruction set is generated in combination with a dynamic constraint function, and the specific steps are as follows,

[0021] The optimization gradient of the building form parameter is processed through the dynamic constraint function application and gradient correction.

[0022] Based on the processed building form parameter, the form adjustment instruction set is generated through the Adam optimizer.

[0023] As a preferred scheme of the energy-saving control method based on the spatial form of the building group, the multi-scale adaptive control algorithm comprises macro-level overall orientation adjustment of the building group, meso-level ventilation coordination between adjacent buildings, and micro-level angle adjustment of the facade sunshade board.

[0024] As a preferred scheme of the energy-saving control method based on the spatial form of the building group, the low-delay communication protocol adopts a zero message queue architecture to realize end-to-end transmission, and dynamically selects a master and standby transmission path through a discriminator to evaluate the confidence of the energy consumption probability distribution thermal map.

[0025] In a second aspect, the present application provides an energy-saving control system based on the spatial form of a building group, comprising: a data acquisition module for acquiring three-dimensional spatial coordinates, volume, orientation angle and composite curvature parameters of each building in the building group, and constructing a building form feature parameter matrix; an energy consumption analysis module for inputting the building form feature parameter matrix into a multi-scale generative adversarial network, extracting overall layout features, local spacing features and surface detail features of the building group, and generating an energy consumption probability distribution thermal map corresponding to the spatial distribution of the building group; an optimization adjustment module for inputting the energy consumption probability distribution thermal map into a discriminator of the multi-scale generative adversarial network, calculating optimization gradients of each building form parameter, and generating a form adjustment instruction set in combination with a dynamic constraint function; and a control execution module for distributing the form adjustment instruction set to each control terminal in the building group through a low-delay communication protocol, and dynamically adjusting operation parameters of equipment based on a multi-scale adaptive control algorithm.

[0026] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, any step of the energy-saving control method based on the spatial form of the building group according to the first aspect of the present application is implemented.

[0027] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, any step of the energy-saving control method based on the spatial form of the building group according to the first aspect of the present application is implemented.

[0028] The present application has the following beneficial effects: by accurately collecting three-dimensional spatial coordinates, volume, orientation angle and composite curvature parameters of the building group, a detailed building form feature parameter matrix is constructed, and the overall layout, local spacing and surface detail features of the building group are extracted by using a multi-scale generative adversarial network to generate an energy consumption probability distribution thermal map. Not only the accuracy of energy consumption prediction is improved, but also high energy consumption areas can be identified and targeted energy-saving measures can be provided. Through intelligent data processing and analysis, the goal of significantly improving the energy efficiency management of the building group is achieved. BRIEF DESCRIPTION OF DRAWINGS

[0029] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description only constitute some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0030] Fig. 1 Flow chart for the building group space form-based energy saving control method in embodiment 1;

[0031] Fig. 2 Flow chart for the composite curvature parameter calculation in embodiment 1;

[0032] Fig. 3 Flow chart for the multi-scale generative adversarial network in embodiment 1;

[0033] Fig. 4 Flow chart for the form adjustment instruction generation in embodiment 1. DETAILED DESCRIPTION

[0034] In order to make the above-mentioned objects, features and advantages of the present application more apparent and comprehensible, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0035] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application. However, the present application can be practiced in other ways different from those described herein without departing from the scope of the present application, and those skilled in the art can make similar extensions without departing from the concept of the present application, so the present application is not limited to the specific embodiments disclosed below.

[0036] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an independent or alternative embodiment that excludes other embodiments.

[0037] Embodiment 1, with reference to Figs. 1-4 The embodiment provides a building group space form-based energy saving control method, comprising the following steps:

[0038] S1: Collecting three-dimensional space coordinates, volume, orientation angle and composite curvature parameters of each building in the building group, and constructing a building form feature parameter matrix;

[0039] S1.1: Collecting three-dimensional coordinates, volume, orientation angle and composite curvature parameters of each building in the building group;

[0040] The unmanned aerial vehicle flies along the cross-circulating path, obtains multi-view images of the building group through RTK positioning, and arranges ground control points;

[0041] It should be noted that the unmanned aerial vehicle is equipped with a five-lens tilt photography module, and a cross-circulating flight path is set. The RTK module built in the unmanned aerial vehicle receives the Qianxun position network differential signal to realize real-time dynamic positioning, and synchronously records the POS data of each image. The image acquisition sequence is sorted according to the time stamp, the original sequence index of the aerial photography path is generated, and the spatial topology sequence index is generated through spatial proximity reordering. A ground control point is arranged every 200 square meters in the building group range. The ground control point adopts a 30*30cm cross-shaped coded mark, and the three-dimensional coordinates of the ground control point are measured using a Leica TS16 total station. Each ground control point is covered by at least three images from different angles. After the flight is completed, the images and POS data are exported and imported into the ContextCapture software together with the ground control point coordinate table as the input source for subsequent dense point cloud generation.

[0042] The dense point cloud is generated using ContextCapture, and the dense point cloud is registered with the ground control point through the ICP algorithm (iterative closest point);

[0043] It should be noted that the multi-view images taken by the unmanned aerial vehicle are imported into the ContextCapture software, and the image matching parameters including image overlap rate, feature point matching accuracy and point cloud resolution are set. The initial dense point cloud is generated through the multi-view stereo matching algorithm. Based on the three-dimensional coordinates of the ground control point measured by the total station, the corresponding ground control point positions in the dense point cloud are manually labeled in ContextCapture. Based on the dense point cloud and the ground control point, the ICP algorithm (iterative closest point) is called to calculate the rigid transformation matrix. Through the minimization of the Euclidean distance error between the dense point cloud and the ground control point, iterative optimization is performed until convergence. The registered dense point cloud is output.

[0044] The registered dense point cloud is used to generate a building triangular mesh model through a Poisson surface reconstruction algorithm, and the geometric parameters of each building are extracted, including three-dimensional coordinates, volume, orientation angle and composite curvature parameters.

[0045] It should be noted that the specific process of generating a building triangular mesh model from the registered dense point cloud through the Poisson surface reconstruction algorithm is as follows: based on the registered dense point cloud data, the Poisson surface reconstruction algorithm is used to calculate the point cloud normal vector and construct an implicit surface function, and the building triangular mesh model is generated through octree spatial division and isosurface extraction.

[0046] In the triangular mesh model, the three-dimensional coordinates are obtained by calculating the geometric center point coordinates of each building triangular mesh model, the building triangular mesh model is decomposed into non-overlapping tetrahedral units by the Delaunay tetrahedral division algorithm, and the volume is obtained by accumulating all the tetrahedral volumes; the covariance matrix of the building bottom triangular mesh projection point set is calculated by the principal component analysis method, and the characteristic vector direction corresponding to the maximum eigenvalue is extracted, and the orientation angle is calculated by combining the angle between the north direction;

[0047] The specific process of the composite curvature parameter for fusing the micro-surface curvature, the meso-distance curvature and the macro-layout curvature through the multi-scale fusion mechanism is as follows: based on the Gaussian curvature and the average curvature of the building surface triangular mesh, the ratio relationship of the Gaussian curvature and the average curvature is processed by the normalized arctangent function to generate the micro-surface curvature; based on the statistical characteristics of the Delaunay triangular division edge length between adjacent buildings, the dispersion degree of the distance distribution is represented by the ratio of the square of the standard deviation of the triangular division edge length to the mean value to generate the meso-distance curvature; based on the eigenvalue difference degree of the covariance matrix of the main axis direction of the building group, the directional intensity of the layout is represented by the ratio of the absolute difference value of the first two large eigenvalues of the covariance matrix to the sum to generate the macro-layout curvature; the micro-surface curvature, the meso-distance curvature and the macro-layout curvature are summed by weighting to generate the composite curvature parameter.

[0048] S1.2: Based on the geometric parameters of each building including three-dimensional coordinates, volume, orientation angle and composite curvature parameter, a building form parameter matrix is constructed.

[0049] S2: The building form feature parameter matrix is input into a multi-scale generative adversarial network to extract the overall layout feature, the local distance feature and the surface detail feature of the building group, and a thermal map of energy consumption probability distribution corresponding to the spatial distribution of the building group is generated;

[0050] S2.1: The building form parameter matrix is normalized and high-dimensional embedded;

[0051] The three-dimensional coordinates, volume, orientation angle and composite curvature parameter in the building form parameter matrix are subjected to parameter normalization processing;

[0052] It should be noted that the three-dimensional coordinate parameter in the building form feature parameter matrix is converted into relative coordinates with the geometric center of the building group as the origin and normalized to the [-1, 1] interval; the volume parameter is divided by the maximum volume value in the building group and constrained to the [0, 1] interval; the orientation angle parameter (radian system) is divided by π and constrained to the [-1, 1] interval.

[0053] The normalized building form parameter matrix is mapped into a high-dimensional embedding matrix through a learnable weight matrix.

[0054] Further, the normalized 6-dimensional parameters are mapped to a 512-dimensional vector through a learnable weight matrix in the multi-scale generative adversarial network to generate a high-dimensional embedding matrix.

[0055] It should be noted that the normalized building form parameter matrix includes 6 columns of parameters such as three-dimensional coordinates, volume, orientation angle, and compound curvature. The normalized building form parameter matrix is input into the multi-scale generative adversarial network, and a linear transformation is performed on the normalized building form parameter matrix through a learnable weight matrix in the multi-scale generative adversarial network to map the 6-dimensional parameters of each building to a 512-dimensional vector to generate an initial high-dimensional embedding building form parameter matrix.

[0056] Based on the spatial topology sequence index of the UAV aerial path, a sine and cosine position encoding is generated, and the sine and cosine position encoding is added to the high-dimensional embedding matrix element by element.

[0057] It should be noted that based on the spatial topology sequence (i.e., the image collection sequence) of the UAV aerial path, a unique position index value is assigned to each building. According to the Transformer position encoding rule, a sine function value is filled in the even dimensions of the 512-dimensional embedding vector, and a cosine function value is filled in the odd dimensions. In the specific calculation, the position index is divided by the power of 10000, and then the trigonometric function operation is performed. The generated sine and cosine position encoding is added to the high-dimensional embedding matrix element by element, so that the high-dimensional embedding matrix contains building form features and spatial sequence information. The high-dimensional embedding matrix is used as the input of the three-level pyramid Transformer generator, and is used for subsequent multi-scale feature extraction and heat map generation.

[0058] S2.2: Extract the overall layout features, local spacing features, and surface detail features of the building group through the three-level pyramid Transformer generator.

[0059] The overall layout feature extraction in the macro layer is combined with the logarithmic decay term of the building spacing in the Softmax function. The correlation weight of the distant buildings is reduced by 50 meters for each increase in building spacing. The overall layout features of the macro layer are finally output, representing the global layout correlation of the building group.

[0060] The local spacing feature extraction in the meso layer is to divide the buildings into 7x7 grid windows according to the normalized plan coordinates through window division and orientation constraint. The self-attention weight between buildings in each window is calculated, and an exponential decay penalty is applied to the buildings with an orientation angle difference of more than 30° to reduce their interaction weight. The meso feature matrix is output, representing the local building spacing and shadow blocking effect.

[0061] It should be noted that the window division is to project the three-dimensional space coordinates of each building in the building group to a plane, ignore the height dimension, generate a two-dimensional plane coordinate set, normalize the two-dimensional plane coordinate set, and divide a uniform grid; calculate the absolute value of the orientation angle difference of each pair of buildings in the grid, and combine the exponential decay penalty term to constrain the orientation;

[0062] The specific steps of calculating the self-attention weight between buildings are: based on the high-dimensional embedding matrix, generate a query matrix, a key matrix and a value matrix through linear transformation; for each building in the grid window, calculate the dot product of the query matrix and the transpose matrix of the key matrix, and divide the result by the scaling factor to obtain an initial attention score matrix; for each pair of buildings in the window, if the difference in the orientation angle exceeds 30°, multiply the initial attention score by an exponential decay weight to reduce the interaction intensity of the building pair; perform Softmax function normalization on the decayed attention score matrix to generate a normalized attention weight matrix; multiply the normalized attention weight matrix with the value matrix to generate the feature representation of the buildings in the window; traverse all grid windows, and concatenate the feature representations output by each window in the original building position order to form a complete mesoscopic feature matrix, representing the local building spacing and shadow blocking effect.

[0063] The surface detail feature extraction of the micro layer is to input the composite curvature parameter into a two-layer fully connected network through curvature modulation attention calculation to generate a modulation coefficient in the interval [0, 1]; in the three independent attention heads, the modulation coefficient is used as the scaling factor of the key-value pair to enhance the feature contribution of the high-curvature building; output a micro feature matrix representing the local influence of surface curvature on energy consumption.

[0064] S2.3: Generate an energy consumption probability distribution heat map corresponding to the spatial distribution of the building group;

[0065] Through hierarchical weight calculation and feature weighted fusion, dynamic gated multi-scale feature fusion is performed to generate a multi-scale fusion feature matrix;

[0066] Further, the macro feature matrix, the mesoscopic feature matrix and the micro feature matrix are respectively subjected to global average pooling to generate scalar weight values; the weights are normalized by the Softmax function to ensure that the sum of the weights of the three is 1;

[0067] The macro feature matrix is multiplied by the macro weight, the mesoscopic feature matrix is multiplied by the mesoscopic weight, and the micro feature matrix is multiplied by the micro weight, and the element-wise addition is performed to obtain a fusion feature matrix; the fusion feature matrix is spatially aligned by a 3x3 convolution kernel to eliminate the resolution difference between different scale features.

[0068] Based on the deconvolution network structure and spatial coordinate mapping, deconvolution upsampling and spatial mapping are performed to generate a single-channel heat map matrix;

[0069] Further, based on the deconvolution network structure, a single-channel heat map matrix is generated;

[0070] First layer deconvolution: input multi-scale fusion feature matrix (dimension n x 512), use 3 x 3 convolution kernel (step 2), output resolution is raised to n x 256 x 256.

[0071] Second layer deconvolution: repeat 3 x 3 convolution (step 2), resolution is gradually raised to 64 x 64, 128 x 128.

[0072] Channel adjustment: the last layer of deconvolution uses a 1 x 1 convolution kernel to reduce the number of channels from 512 to 1, generating a single-channel heat map matrix.

[0073] Generate heat map pixels through spatial coordinate mapping;

[0074] Align the normalized coordinates of the building with the heat map pixel grid, and map the discrete building features in the multi-scale fusion feature matrix to the continuous space through bilinear interpolation; in the deconvolution process, according to the building position mask (generated by three-dimensional coordinates), the non-building area (such as green land, road) of the heat map is constrained to be a low energy consumption value.

[0075] Combine the Sigmoid activation function and physical constraint correction to perform heat map probability normalization and generate an energy consumption probability distribution heat map;

[0076] Further, input the single-channel heat map matrix into the Sigmoid function to constrain the original value to the [0, 1] interval to generate a probability distribution heat map; according to the heat transfer equation (Fourier's law), the heat map gradient is forced to match the material thermal conductivity in the building surface area, and L2 regularization constraint is applied to the pixels outside the mask (non-building area) to avoid noise interference.

[0077] S3: input the energy consumption probability distribution heat map into the discriminator of the multi-scale generative adversarial network, calculate the optimization gradient of each building form parameter, and generate a form adjustment instruction set combined with the dynamic constraint function;

[0078] S3.1: preprocess the energy consumption probability distribution heat map;

[0079] Linearly map the pixel value of the energy consumption probability distribution heat map from [0, 1] to [-1, 1] interval to adapt the input range of the discriminator of the multi-scale generative adversarial network; use bicubic interpolation method to adjust the heat map resolution from 128 x 128 to 256 x 256, improve the sensitivity of the discriminator to detailed features.

[0080] A binary mask is generated based on the collected three-dimensional coordinates of the building, with the building projection area marked as 1 and the non-building area (road, green land) marked as -1; the binary mask is multiplied with the normalized energy consumption probability distribution heat map pixel by pixel to suppress the noise interference in the non-building area.

[0081] S3.2: Multi-scale convolution feature extraction is performed on the processed energy consumption probability distribution heat map to generate a confidence matrix, and a target function is calculated;

[0082] The processed energy consumption probability distribution heat map is input into the discriminator of the multi-scale generative adversarial network, and is sequentially subjected to 5 layers of convolution operation, with the resolution of the feature map output by each layer being reduced by half; a channel attention module is inserted after the third layer of convolution, and the weights are dynamically adjusted according to the importance of the feature map channels; the last layer of convolution outputs an 8x8 feature map, which is mapped to a 128x128 confidence matrix through a fully connected layer, with each pixel value ∈ [-1, 1], representing the matching degree of the position heat map with the real energy consumption distribution; the mean value of the confidence matrix is calculated as the global optimization target.

[0083] S3.3: Based on the target function, the optimization gradient of each building shape parameter is calculated;

[0084] Further, the confidence mean value is taken as the optimization target, and the gradient of the building shape parameter matrix is calculated through the PyTorch automatic differentiation framework; the gradient of each building shape parameter is calculated independently; the building shape parameters are sorted according to the absolute value of the gradient, and the high sensitivity parameters are optimized preferentially; the gradient of the composite curvature parameter is subjected to a sign constraint; the composite curvature parameter whose gradient value exceeds the range of [-0.5, 0.5] is truncated to avoid optimization shock.

[0085] S3.4: Shape adjustment instruction set is generated in combination with the dynamic constraint function;

[0086] The dynamic constraint function is applied and the gradient is corrected through minimum distance constraint correction, curvature safety range constraint, and volume cost constraint correction;

[0087] Further, all building pairs are traversed, and if the building distance is less than the planning fire distance, the coordinate gradient is subjected to a reverse correction. For every 1 meter reduction in building distance, the corresponding coordinate gradient is reduced by a correction coefficient; if the composite curvature parameter exceeds the material safety range, the curvature gradient is set to zero to prohibit further adjustment; according to the volume adjustment cost model, the cost penalty term is deducted from the optimization gradient of the building shape parameter. The volume gradient is linearly reduced by a cost weight (for example, 0.01);

[0088] It should be noted that the volume adjustment cost model collects building energy saving reconstruction economic parameters, including unit volume adjustment cost (yuan / m3) and material strength limit unit price (yuan / MPa), and stores them into a dynamic constraint database. According to the economic parameters in the dynamic constraint database, the volume adjustment cost model is constructed

[0089] When initializing the Adam optimizer, set the learning rate, momentum decay parameter, and second moment estimation decay parameter, and initialize the first moment estimation vector and second moment estimation vector to zero matrix. Sort the gradient of the building form parameter matrix according to the absolute value of the gradient, and the priority order is volume parameter > composite curvature parameter > three-dimensional coordinate parameter. Update each parameter in turn: extract the current parameter gradient value from the optimization gradient of the building form parameter, multiply it by the learning rate, obtain the superimposed momentum term based on the exponentially decaying weighted average of the first moment estimation vector and the momentum decay parameter, and obtain the second moment estimation correction term based on the weighted average of the second moment estimation vector and the second moment estimation decay parameter, and finally generate the parameter update amount.

[0090] Input the adjusted form parameter matrix into the three-level pyramid Transformer generator to generate a new energy consumption probability distribution heat map and calculate the confidence improvement amount. Call the collision detection interface of the BIM model to check whether the adjusted building spacing and curvature violate safety specifications. If there is a conflict, roll back to the last valid parameter version;

[0091] It should be noted that the adjusted form parameter matrix is input into the generator of the pre-trained multi-scale generative adversarial network to generate a new energy consumption probability distribution heat map. The new heat map is input into the discriminator to output a new confidence matrix and calculate the mean, which is compared with the original confidence mean to obtain the improvement amount. If the improvement amount is less than 0.1, trigger the conflict detection process. Call the collision detection interface of the BIM model to traverse the three-dimensional coordinates of all building pairs, calculate the plane Euclidean distance, and if there is a building spacing less than the fire spacing specification value (e.g. 6 meters), mark it as a spacing conflict. Check if the composite curvature parameter exceeds the material safety range. If it exceeds, mark it as a curvature conflict. If there is any conflict, retrieve the last conflict-free form parameter matrix from the historical version library and replace the current parameter matrix. Re-input the rolled-back parameter matrix into the generator to generate a heat map and calculate the confidence. If the improvement amount is greater than or equal to 0.1 and there is no conflict, save it as the latest valid version.

[0092] Generate a list of instructions in the form of a three-tuple of "building ID, parameter type, and adjustment amount", and arrange them in ascending order of building ID to generate a set of form adjustment instructions.

[0093] S4: Distribute the form adjustment instruction set to each control terminal in the building group through a low-latency communication protocol, and dynamically adjust the operating parameters of the equipment based on a multi-scale adaptive control algorithm.

[0094] Further, the morphology adjustment instruction set in JSON format is transmitted to the edge gateway in the building group through the MQTT protocol;

[0095] It should be noted that when the morphology adjustment instruction set in JSON format (including building ID, parameter type, adjustment amount, and timestamp field) is transmitted to the edge gateway in the building group through the MQTT protocol, the morphology adjustment instruction set in JSON format is first UTF-8 encoded to generate a binary data stream; the MQTT client publishes a message to the parameterized topic path, wherein the topic path parameter of the MQTT protocol is replaced by the building group; the edge gateway subscribes to the topic path parameter of the MQTT protocol, parses the JSON field after receiving the message, verifies the legality of the building ID (for example, checks whether the ID is in the pre-registration list), and if it is legal, distributes the morphology adjustment instruction set to the corresponding control terminal according to the building ID; if no message is received, the message is retransmitted according to the QoS=1 rule after 50 ms, and the maximum retransmission is 3 times; after the transmission is completed, the edge gateway generates a transmission log (including message ID, timestamp, and building ID list) and stores it in the local database.

[0096] The edge gateway routes the instruction to the corresponding control terminal according to the building ID;

[0097] It should be noted that after the edge gateway receives the morphology adjustment instruction set in JSON format, it parses the building ID field and queries the pre-defined building ID-IP address mapping table; it verifies whether the building ID exists in the registration list, and if it exists, it extracts the target IP address; it publishes the instruction message to the exclusive topic of the target control terminal through the MQTT protocol; the control terminal subscribes to the topic corresponding to its building ID, receives and parses the instruction content; the edge gateway records the transmission log, including the building ID, target IP address, and timestamp, and stores it in the local SQLite database.

[0098] After the control terminal receives the instruction, it parses the parameter type and adjustment amount, and dynamically adjusts the equipment operating parameters through the multi-scale adaptive control algorithm;

[0099] Further, the expansion percentage of the building facade sunshade component is adjusted in proportion according to the volume adjustment amount at the macro time scale (hour level);

[0100] At the meso time scale (minute level), the opening degree and air volume of the ventilation system are adjusted by linear interpolation according to the composite curvature parameter adjustment amount;

[0101] At the micro time scale (second level), the temperature set value of the air conditioning equipment is dynamically corrected by the PID control algorithm according to the orientation angle adjustment amount combined with real-time illumination intensity and temperature sensor data;

[0102] All device parameter adjustment results are fed back to the edge gateway through the Modbus-TCP protocol, real-time operation logs are generated, and the real-time operation logs are synchronized to the building group management platform.

[0103] The embodiment also provides an energy-saving control system based on building group space form, comprising: a data acquisition module, configured to acquire three-dimensional space coordinates, volume, orientation angle and compound curvature parameters of each building in the building group, and construct a building form feature parameter matrix; an energy consumption analysis module, configured to input the building form feature parameter matrix into a multi-scale generative adversarial network, extract overall layout features, local spacing features and surface detail features of the building group respectively, and generate an energy consumption probability distribution heat map corresponding to the space distribution of the building group; an optimization adjustment module, configured to input the energy consumption probability distribution heat map into a discriminator of the multi-scale generative adversarial network, calculate optimization gradients of each building form parameter, and generate a form adjustment instruction set in combination with a dynamic constraint function; and a control execution module, configured to distribute the form adjustment instruction set to each control terminal in the building group through a low-delay communication protocol, and dynamically adjust operation parameters of devices based on a multi-scale adaptive control algorithm.

[0104] The embodiment also provides a computer device suitable for the energy-saving control method based on building group space form, comprising: a memory and a processor; the memory is configured to store computer executable instructions, and the processor is configured to execute the computer executable instructions to implement the energy-saving control method based on building group space form proposed in the above embodiment.

[0105] The computer device can be a terminal, and the computer device comprises a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for running of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is configured to perform wired or wireless communication with external terminals. The wireless communication can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or can be a key, trackball or touchpad arranged on the shell of the computer device, or can be an external keyboard, touchpad or mouse, etc.

[0106] The embodiment also provides a storage medium on which a computer program is stored, the program being executed by a processor to implement the energy-saving control method based on the spatial form of the building group as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk, or an optical disk.

[0107] To sum up, the present application achieves the following effects: by accurately collecting the three-dimensional spatial coordinates, volume, orientation angle and compound curvature parameters of the building group, a detailed building form feature parameter matrix is constructed, and the overall layout, local spacing and surface detail features of the building group are extracted by using a multi-scale generative adversarial network to generate an energy consumption probability distribution thermal map. Not only the accuracy of energy consumption prediction is improved, but also high energy consumption areas can be identified and targeted energy-saving measures can be provided. Through intelligent data processing and analysis, the goal of significantly improving the energy efficiency management of the building group is achieved, and the problems of limited energy consumption prediction accuracy and difficulty in efficiently converting into operable energy-saving measures in the prior art are solved.

[0108] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application but not limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, and all of them should be covered in the scope of the claims of the present application.

Claims

1. A method for energy saving control based on spatial form of a building group, characterized in that: The application relates to a building form optimization method based on multi-scale adaptive control. The method comprises the following steps: collecting three-dimensional space coordinates, volume, orientation angle and composite curvature parameters of each building in a building group, and constructing a building form feature parameter matrix; inputting the building form feature parameter matrix into a multi-scale generative adversarial network, respectively extracting overall layout features, local spacing features and surface detail features of the building group, and generating an energy consumption probability distribution heat map corresponding to the spatial distribution of the building group; inputting the energy consumption probability distribution heat map into a discriminator of the multi-scale generative adversarial network, calculating the optimization gradient of each building form parameter, and generating a form adjustment instruction set in combination with a dynamic constraint function; distributing the form adjustment instruction set to each control terminal in the building group through a low-delay communication protocol, and dynamically adjusting the operation parameters of the equipment based on a multi-scale adaptive control algorithm; the composite curvature parameter is the result of fusing micro surface curvature, mesoscopic spacing curvature and macro layout curvature through a multi-scale fusion mechanism; the micro surface curvature is the ratio relationship between the Gaussian curvature and the average curvature of the building surface triangular mesh; the mesoscopic spacing curvature is the statistical dispersion degree feature of the Delaunay triangular partition edge length of adjacent buildings; the macro layout curvature is the eigenvalue difference degree relationship of the coordinate covariance matrix of the building group; the specific fusion process is as follows, based on the Gaussian curvature and the average curvature of the building surface triangular mesh, the ratio relationship between the Gaussian curvature and the average curvature is processed through a normalized arctangent function to generate the micro surface curvature; based on the statistical features of the Delaunay triangular partition edge length between adjacent buildings, the dispersion degree of the spacing distribution is represented by the ratio of the square of the standard deviation and the average of the triangular partition edge length to generate the mesoscopic spacing curvature; based on the eigenvalue difference degree of the principal axis direction covariance matrix of the building group, the directional intensity of the layout is represented by the ratio of the absolute difference value and the sum of the first two eigenvalues of the covariance matrix to generate the macro layout curvature; the micro surface curvature, the mesoscopic spacing curvature and the macro layout curvature are summed through weighting to generate the composite curvature parameter.

2. The building group space form based energy saving control method according to claim 1, wherein: the generator of the multi-scale generative adversarial network is composed of a three-level pyramid type layered attention mechanism; the macro layer adopts an attenuation type space attention mechanism; the mesoscopic layer adopts a windowed local attention mechanism; the micro layer adopts a curvature modulation multi-head attention mechanism.

3. The building group space form based energy saving control method according to claim 2, wherein: the specific steps of inputting the energy consumption probability distribution heat map into the discriminator of the multi-scale generative adversarial network to calculate the optimization gradient of each building form parameter are as follows, normalizing the energy consumption probability distribution heat map and performing building position mask superposition processing; performing multi-scale convolution feature extraction on the processed energy consumption probability distribution heat map, and generating a confidence matrix; based on the confidence matrix, the optimization gradient of each building form parameter is identified.

4. The building group space form based energy saving control method according to claim 1, wherein: the specific steps of generating the form adjustment instruction set in combination with the dynamic constraint function are as follows, processing the optimization gradient of the building form parameter through dynamic constraint function application and gradient correction; based on the processed building form parameter, the form adjustment instruction set is generated through an Adam optimizer. 5.The building group space form based energy-saving control method according to claim 1, wherein: the multi-scale adaptive control algorithm comprises macro-level overall orientation adjustment of the building group, mesoscopic-level ventilation coordination between adjacent buildings and micro-level facade sunshade board angle adjustment.

6. The building group space form based energy saving control method according to claim 5, wherein: The low-delay communication protocol adopts a zero message queue architecture to realize end-to-end transmission, and dynamically selects a main and backup transmission path through confidence evaluation of an energy consumption probability distribution heat map by a discriminator.

7. A building group space form-based energy saving control system based on the building group space form-based energy saving control method according to any one of claims 1 to 6, characterized in that: The method comprises the steps of: a data acquisition module is configured to acquire three-dimensional spatial coordinates, volume, orientation angle and compound curvature parameters of each building in the building group, and construct a building form feature parameter matrix; an energy consumption analysis module is configured to input the building form feature parameter matrix into a multi-scale generative adversarial network, extract overall layout features, local spacing features and surface detail features of the building group respectively, and generate an energy consumption probability distribution heat map corresponding to the spatial distribution of the building group; an optimization adjustment module is configured to input the energy consumption probability distribution heat map into a discriminator of the multi-scale generative adversarial network, calculate optimization gradients of each building form parameter, and generate a form adjustment instruction set in combination with a dynamic constraint function; a control execution module is configured to distribute the form adjustment instruction set to each control terminal in the building group through a low-delay communication protocol, and dynamically adjust operation parameters of equipment based on a multi-scale adaptive control algorithm.

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that: The processor executes the computer program to realize the steps of the energy-saving control method based on the spatial form of the building group according to any one of claims 1-6.

9. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to realize the steps of the energy-saving control method based on the spatial form of the building group according to any one of claims 1-6.

Images