Building group heat supply load prediction method considering personnel occupancy rate based on shoebox algorithm

Abstract

The invention belongs to the technical field of central heating load planning, and relates to a building group heating load prediction method based on a shoebox algorithm and considering a personnel occupancy rate, which comprises the following steps: collecting geometric feature information of a regional building group; intercepting regional non-geometric vector data; fusing multi-source vector data; building hot area division based on a shoebox algorithm; building a building-level personnel occupancy model theoretical framework; building-level typical occupation scene division; inhabitant activity chain depicting; solving a personnel state transition probability matrix based on a Markov chain; building a building prototype template library; performing prototype identification and template matching; according to the building-level personnel occupancy probability model, the building level is expanded from the single space / room scale by considering the original personnel occupancy rate, and the problem of complex and multi-level heat use scenes in the area caused by personnel activities and occupancy behaviors is fully considered by adopting the modeling thinking based on the main body; according to the method, building heat supply load prediction can be more careful and accurate.

Description

Technical Field

[0001] The present invention belongs to the technical field of central heating load planning, and relates to a method for predicting the heating load of a building complex considering the occupancy rate based on the shoe-box algorithm. Background Technique

[0002] In the northern regions of China, district heating is the most common heating form. Precise heating on demand is the basic path to ensure the heating comfort of users and achieve significant energy conservation and emission reduction at the same time. Due to the large system thermal inertia, building thermal inertia, and time-varying heat consumption patterns of users in district heating, the transient regulation response ability of the system is limited, and there are varying degrees of supply-demand mismatch and energy waste in the actual heating process. At present, in order to rationally configure and design the system during the regional energy planning stage, the prediction methods for the heating load of regional building complexes generally include the following three categories: 1. The load estimation method based on static thermal indexes; 2. The dynamic load calculation method with the help of building energy consumption simulation platforms; 3. The method of deducing historical energy consumption data using data-driven models.

[0003] However, for the first category of load estimation methods based on static thermal indexes, that is, the thermal index estimation method. It is mostly used to calculate the design load of district heating systems. The estimation results based on static indexes are often much higher than the actual heat consumption demand of users, and are numerically closer to the near-extreme load values in the actual situation. Therefore, using this method to predict the heating load of regional building complexes is likely to cause an overestimation of the load on the one hand, resulting in excessive redundant installed capacity and higher initial investment costs in the system. On the other hand, the static load results calculated by this method are difficult to guide the optimal design and actual operation regulation of the system. Therefore, the first category of methods is difficult to meet the load prediction requirements at the regional scale at the present stage.

[0004] Although the second category of dynamic load calculation methods with the help of building energy consumption simulation platforms shows high calculation efficiency and calculation accuracy at the single-building level. However, when the calculation scale is expanded to the regional level, with the addition of a large number of buildings with diverse functions, at this time, for the traditional modeling and energy simulation methods applicable to single buildings, in order to obtain load prediction results with the same time resolution, the modeling and calculation time costs will increase exponentially with the expansion of the scale, and a large amount of building detail information is required as the basis for model construction. Therefore, the second category of methods is difficult to show high calculation efficiency and strong applicability in the prediction of the energy demand of building complexes at the regional scale.

[0005] The third major category of historical energy consumption data deduction methods using data-driven models highly relies on long-term and high-quality historical load data and is only applicable to existing stock buildings. Therefore, it is difficult to directly apply the third category of methods to the energy design and planning stage of newly built building complexes. In addition, due to the limitation of the dataset stock, the energy consumption data statistical platforms of the vast majority of buildings are still incomplete, and the trained prediction models often lack adaptability to technological progress and condition changes.

[0006] As can be seen from the above, some of the existing heating load prediction methods for regional building complexes can only obtain static design loads. The total regional load obtained by using the simple superposition method ignores the potential asynchrony in time of energy consumption in different types of buildings, so it is very easy to overestimate the extreme load of the building complex. Some need to be improved in terms of modeling cost and calculation efficiency, either with an approximate black-box processing method, which is not popular and adaptable for newly built buildings. Moreover, no matter which category of heating load prediction method in the existing technology, there is a common problem that it does not consider the complex and multi-level heat consumption scenarios within the region caused by human activities and occupancy behaviors. For building complexes at the regional level, the climate conditions of the location, the spatial layout and geometric form of the buildings, the building functions, and the occupancy behaviors of the people will all affect their energy demands. The differences in energy consumption patterns among different types of buildings are the main reasons for the asynchronous characteristics of the heat loads of each building in the region in time.

[0007] Therefore, at the present stage, it is necessary to propose a set of universal heating dynamic load prediction methods to solve the above technical problems on the basis of considering the multi-level model input conditions of regional building complexes. Summary of the Invention

[0008] The method of the present invention is used to predict the heating dynamic load of building complexes at the regional scale. The overall working process of this method, the connection and logical sequence between each step, and the creation process of the regional building complex energy analysis model are divided into constructing a bottom-up building thermophysical sub-model based on the shoebox algorithm and developing a building-level occupant occupancy sub-model based on the Markov chain (MC-Markov Chain). By probabilistically characterizing the highly uncertain occupant transfer and occupancy behaviors, it reflects the complex and asynchronous heat consumption scenarios within the region, thereby improving the accuracy of model prediction.

[0009] The technical solution adopted by the present invention to solve the technical problem is: a heating load prediction method for building complexes considering the occupancy rate based on the shoebox algorithm, including the following steps:

[0010] Step 1: Collect geometric feature information of the regional building complex;

[0011] Step 2: Intercept regional non-geometric vector data;

[0012] Step 3, multi-source vector data fusion;

[0013] Step 4, building thermal zone division based on the shoe-box algorithm;

[0014] Step 5, construction of the theoretical framework of the building-level occupancy model;

[0015] Step 6, building-level typical occupancy scenario division;

[0016] Step 7, characterization of the occupant activity chain;

[0017] Step 8, solving the personnel state transition probability matrix based on the Markov chain;

[0018] Step 9, construction of the building prototype template library;

[0019] Step 10, prototype recognition and template matching.

[0020] Preferably, in the said Step 1, the collected geometric feature information of the regional building complex is converted into vector data; the geometric feature information of the regional building complex includes: the detailed information of each building in the region, and the detailed information includes: the bottom contour, floor area, relative spatial position, and height;

[0021] In Step 2, the vector data integration of the point of interest (POI) and the area of interest (AOI) is adopted, combined with the land use type data and the cross-section of the historical satellite cloud map, to jointly judge the use type and construction year of each building in the region; the regional non-geometric vector data includes: the point of interest (POI) and the area of interest (AOI); the point of interest (POI) represents the point-like vector data set of the building, and the area of interest (AOI) represents the regional geographical entity of the building on the map;

[0022] In Step 3, by introducing an attribute field for each closed building contour in the two-dimensional plane, the building footprints without main attributes in the region are converted into building footprints with main attributes, and the geometric and usage function information of each building in the region is summarized into the building attribute list, and a regional two-dimensional vector map is generated accordingly.

[0023] More preferably, in the said Step 4, the regional building complex vector map obtained in Step 3 is used to generate a three-dimensional model of the building complex space, and the shoe-box algorithm is used to process the three-dimensional model of the building complex space, and the three-dimensional model of the building complex space is converted into a regional building complex thermophysical model capable of performing dynamic energy simulation;

[0024] The core indicators used in the creation process of the shoe-box algorithm: the edge ratio er, elongation rate el, and shape factor sf are respectively:

[0025]

[0026] In formulas (2), (3), and (4), x, y, and z respectively represent the width, depth, and height of the core shoe box.

[0027] Preferably, in step 2, when vector data is insufficient or missing, national land type coverage data is used to assist in judgment and verification; for the construction year range of buildings, it is determined by comparing historical satellite images; data providers of national land type coverage data include Google Earth, Baidu Maps, AutoNavi Maps, and elevation maps.

[0028] Preferably, in step 5, a Markov chain probability transition model with personnel type and building usage function as the core is created to depict the temporal stochastic occupancy behavior of personnel within the region; starting from the personnel occupancy behavior at the building level, the prediction scale of the Markov chain model is extended from the single building level to the regional scope;

[0029] The classification of personnel includes: working people, students, and retired elderly people;

[0030] The classification of buildings includes: residential buildings, office buildings, and school buildings;

[0031] In step 5, a Markov chain model is used to abstract a temporal state transition matrix reflecting the random decision-making process of personnel, thereby obtaining a typical occupancy scenario with personnel as the main body; combined with the proportion of the number of each type of personnel in the region, a typical occupancy pattern with buildings as the main body is summarized; when conducting building energy consumption simulation, according to the unit area occupancy rate of personnel in each type of building, the corresponding total number of occupied people per unit time in each building in the region is calculated to estimate the heat gain of indoor personnel.

[0032] Preferably, in step 6, the division of typical occupancy scenarios considers the interaction and correlation between different personnel entities and different building types, and uses the form of sets and legends to represent all occupancy situations of different types of personnel in different functional buildings, and uses arrows of different colors to represent the potential transfer situations of personnel between buildings;

[0033] In step 7, the behavior rules of each type of personnel are described in the form of activity sequences, depicting the behavior and activity trajectories of personnel on weekdays and weekends, including going to work, going to school, eating, going out, and going home; the activity sequence of personnel in a day is successively divided into four parts: morning, noon, afternoon, and evening, and different colors are used to distinguish the occupied building scenarios; the activities of personnel are regarded as the transfer of occupancy states at the building level. Among them, school buildings only accommodate students, office buildings only accommodate working people, residential buildings contain all types of personnel entities, and the composition of various types of personnel in the region is evaluated in combination with the demographic information of the target area.

[0034] Preferably, in step 8, an ordered Markov chain is used to generate a random occupancy schedule for personnel; by creating a transition probability matrix, the time use survey data can be effectively converted into an occupancy dataset representing each building type.

[0035] Preferably, in step 8, the state probability of a person at time τ is jointly represented by the state probability at the initial state τ = 0 and the transition probability matrix as:

[0036]

[0037] In Equation (12), P 0 (τ) and P 1 (τ) represent the state probability matrix and the transition probability matrix respectively, and τ represents the unit time.

[0038] Preferably, in step 9, in the building energy consumption simulation, in addition to building a detailed building thermophysical model, simulation parameters required for calculation are also configured for the building models in the area; the simulation parameters include: setting the thermal parameters of the envelope structure, window-wall ratio, ventilation / air permeability, indoor design temperature, heating control strategy, occupancy rate of personnel, and equipment usage schedule for each building in the area;

[0039] By creating a building prototype template library and matching the building prototype template library with the buildings in the calculation area, batch configuration of the model simulation parameters is achieved; and core characteristic indicators are determined to identify and match the building models.

[0040] Preferably, in step 10, based on the building prototype template library, the simulation parameter configuration assigns the created building prototype templates to each building in the area to ensure the authenticity and effectiveness of the energy analysis model of the regional building complex;

[0041] The identification of building prototypes includes: assigning prototype templates according to the extracted building core characteristics; by determining the corresponding characteristic codes for each building in the area and combining the corresponding building design specifications, batch configuration of the calculation simulation parameters is completed.

[0042] The beneficial effects of the present invention are:

[0043] 1. By extracting building information in the regional environment from multi-source databases, including geometric and usage function type information of buildings, and using ArcGIS to fuse the multi-source information to construct a vector map of the region, the present invention reduces the time cost of modeling while ensuring building details, and the constructed regional spatial three-dimensional model can better reflect the building layout in the real scenario. Therefore, the method of the present invention is effective and practical in improving system capacity design, enhancing load prediction accuracy, and reducing the initial investment in system design.

[0044] 2. The present invention proposes to use the shoe-box algorithm to simplify the calculated heat zones of regional building groups, thereby improving the efficiency of modeling and calculation.

[0045] 3. For regional-level building groups, the present invention proposes a building-level personnel occupancy probability model, expanding the consideration of personnel occupancy rate from the single-space / room scale to the building level, and adopting an agent-based modeling thinking. Therefore, the method of the present invention can make the prediction of building heating load more detailed and accurate.

[0046] 4. The present invention proposes a method of using a building prototype template library to batch-assign simulation parameters for building models, so as to ensure the authenticity and effectiveness of the energy analysis model of regional building groups. Therefore, the method of the present invention can make the prediction of building heating load more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 is the flowchart of the method for predicting the heating load of building groups considering personnel occupancy rate based on the shoe-box algorithm of the present invention;

[0048] Figure 2 is the two-dimensional vector plane base map of the region of the present invention;

[0049] Figure 3 is the comparison diagram of the point of interest (POI) and the area of interest (AOI) of the present invention in the plane map;

[0050] Figure 4 is the recognition and matching process diagram of the building type and vector data of the present invention;

[0051] Figure 5 is the flowchart of the shoe-box algorithm of the present invention;

[0052] Figure 6 is the theoretical framework diagram of the building-level personnel occupancy model of the present invention;

[0053] Figure 7 is the diagram of the division of typical occupancy scenarios at the building level of the present invention;

[0054] Figure 8 is the diagram of the personnel activity chain and the corresponding building occupancy status of the present invention;

[0055] Figure 9 is the schematic diagram of the Markov chain state transition model of the present invention;

[0056] Figure 10 is the logic block diagram for determining the initial state probability of the present invention;

[0057] Figure 11 is the diagram of the hierarchical relationship of influencing variables of the present invention;

[0058] Figure 12 is the logic block diagram for the identification and matching of the building prototypes of the present invention;

[0059] Figure 13 is the diagram of the composition and distribution ratio of the main body of the personnel in the case area of the present invention;

[0060] Figure 14 is the diagram of the typical occupancy of the three types of buildings of the present invention on weekdays / weekends;

[0061] Figure 15 is the comparison diagram of the actual monitoring data in the embodiment of the present invention and the design load calculated by the heat index method per unit area. Detailed implementation manners

[0062] Next, the related technologies in the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present invention.

[0063] Refer to Figures 1 to 15 , in this implementation manner, the implementation process is first elaborated as a whole. The creation process of the model starts from the extraction of multi-source building vector data. Using the existing Geographic Information Service System (GIS), the extraction of building geometric profiles and usage type information is carried out, and the core geometric indicators, spatial layouts, and structural compositions of each building in the area are summarized and integrated in ArcGIS10.7; subsequently, based on the regional two-dimensional vector map, a three-dimensional spatial model of the regional building complex is constructed in the RHINOCERO 3D environment, and the shoe box algorithm is used to divide the calculated heat zones of the buildings into core shoe box areas and peripheral areas, thereby completing the construction process of the thermal physics sub-model of the regional building complex.

[0064] Then, according to the collected time use survey data of the occupants, an improved Markov chain mathematical model is adopted, and the state transition probability matrix of the occupants with an hourly resolution is solved through the Python programming language, and based on this, a building-level occupant occupancy sub-model is created. On this basis, by depicting the typical occupancy patterns of building-level occupants, a detailed occupancy schedule (distinguishing weekdays and weekends) corresponding to each functional building is obtained; in addition, taking the number of floors, functional type, and construction year of the building as the core feature indicators of the standard building, a building prototype template library is established, and the existing occupancy schedule is incorporated into it. Finally, the corresponding prototype template is assigned to each building model in the area through the JSON call code, and EnergyPlus is used as the simulation engine to perform dynamic load calculations and outputs with an hourly resolution on the umi (developed by the Massachusetts Institute of Technology) energy simulation platform.

[0065] The specific modeling and implementation methods for each step in the workflow will be elaborated separately below.

[0066] Step 1: Collection of geometric feature information of regional building complexes

[0067] The geometric feature information of regional building complexes includes detailed information such as the bottom contour, floor area, relative position, and height of each building within the specified area. The process of extracting building geometric information is divided into two stages: identifying and converting the attributes of buildings on the map. The method proposed in the present invention utilizes the open application programming interface (API) provided by the China National Geographic Information Public Service Platform (CN-GIP), and adopts a computer Python crawler tool to extract the regional sliced base map in vector form. Subsequently, the point coordinates on the bottom contour line of each building within the area and the metric height information are collected. The attribute information of all buildings is converted into surface elements and imported into the ArcGIS 10.7 software platform for additional processing, including loading regional sliced vector data, setting the band extraction function, raster reclassification, element fusion (merging the contour lines belonging to the same building), and surface simplification (reducing the jaggedness of the contour lines). As Figure 2 shown, the processed regional two-dimensional vector plane reflects the geometric characteristics of each building within the area in the form of vector building footprints. The attribute list on the left side of the figure represents the building height and building floor area corresponding to each footprint object.

[0068] Step 2: Interception of regional non-geometric vector data

[0069] Regional non-geometric vector data is mainly used to judge the usage function and construction year of each building within the area. Among them, POI (Point of Interest) and AOI (Area of Interest) vector data are helpful for analyzing and judging the usage type and function of each building within the area. When the data is insufficient or missing, the national land type coverage data can be used to assist in judgment and verification. For the construction year interval of buildings, it can be determined by comparing historical satellite images. From the perspective of data sources, data providers include Google Earth, Baidu Map, Amap, elevation map, etc. Among them, the vector data of Baidu Map and Amap can be directly accessed through their APIs. In recent years, AI Earth under Alibaba has also provided the latest building coverage data with a resolution of 10 meters, and supports viewing historical satellite cloud images in different year intervals. In addition, well-known domestic Internet platforms such as Meituan, Lianjia, and Ctrip also have the potential to provide extensive information such as building types and years. Most data sources may contain duplicate information, and combining multiple information acquisition channels can provide supplementation and support for the data.

[0070] A Point of Interest (POI) refers to a point - like vector data set in a Geographic Information System (GIS). It generally refers to the point - like data in an Internet electronic map platform and can include any entity abstracted as a point, such as a subway station, a gas station, or a restaurant. Each collected POI point usually contains various attributes related to the corresponding entity, including name, address, coordinates (latitude and longitude), and building category. Therefore, each building within a region should have corresponding POI data.

[0071] Correspondingly, an Area of Interest (AOI) belongs to a type of spatial carrier. Specifically, it represents a boundary range composed of many groups of coordinates and is mainly used to express regional - shaped geographical entities in a map, that is, various surface data in vector map data, such as the most basic administrative division surface data. Figure 3 The essential differences between POI and AOI vector data are clearly reflected on the map. The gray area in the figure represents the AOI data of a certain campus, while other point - like markings represent the POI data of some buildings. In comparison, AOI has better expressiveness, better computing power, and better stability (its change frequency is lower). Conversely, POI has a higher level of abstraction, which means that in a regional map of the same area, the information density carried by POI is higher, and it can transmit vector information at a higher resolution.

[0072] The present invention adopts a method of integrating POI and AOI vector data, combines land - use type data and historical satellite cloud map cross - sections, and jointly determines the use type and construction year of each building within the region. The POI and AOI vector data used in the present invention are respectively from open - application programming interfaces (APIs) provided by different map providers. Through the construction of application code, data is obtained from the website and the extracted information is stored in a database. The historical satellite cloud map cross - sections are taken from open - source data provided by an Internet platform. In addition, the yearbook reports of the target area also provide some information support.

[0073] Step 3: Multi - source vector data fusion

[0074] Another technical problem solved by this method is how to match and fuse the obtained multi - source geographical information and building vector data with the building footprints within the region on a two - dimensional plane to generate a regional two - dimensional vector map.

[0075] In steps 1 and 2, the vector data in the form of POIs and AOIs within the area are collected and integrated into a file in the Shapefile (.shp) format (ESRI Shapefile complies with the open spatial data format developed by the Environmental Systems Research Institute, abbreviated as Shapefile). To facilitate the connection and matching of point vector files and polygon vector files, the "Spatial Join" interactive function in ArcGIS is used to integrate the ESRI Shapefile (.shp) files in point and polygon forms with the building outline vector map obtained in step 1.

[0076] However, during the actual operation process, the geographical locations of some POI points or AOI polygons in the compiled POI / AOI vector data collected by the computer may not exactly match the building footprint coverage locations in the real world. Therefore, in this method, this problem is well solved by adopting the "Spatial Fusion" method. That is, POI points located within the perimeter range of any building are determined by the tolerance range and will be attributed to the building with the closest Euclidean calculated distance to it. The distance calculation formula is as follows:

[0077]

[0078] where x i and y i represent the planar position coordinates of the i-th POI status point; x j and y j represent the planar position coordinates of the geometric midpoint of the j-th building outline; D represents the Euclidean distance between the two, with the unit of meters.

[0079] In this process, the "Spatial Join" tool will determine the number of polygons to which a point within the area belongs. If a point cannot find its polygon, the tool will determine the number of the closest polygon. Figure 4 Shows the process of identifying and matching each building type within the area. Among them, the allocation tolerance range of POI vector points is set to 5m; the coverage rate threshold of AOI vector polygons is set to 70%, that is, if more than 70% of the footprint of a building is covered by a certain AOI polygon, then this building is classified as the corresponding type of this AOI polygon.

[0080] By introducing an "area number" attribute field for each point, the building footprints without main attributes within the area are transformed into building footprints with main attributes, and the geometric and usage function information of each building within the area is summarized into the building attribute list in ArcGIS, and a two-dimensional vector map of the area is generated accordingly.

[0081] Step 4: Division of building heat zones based on the shoebox algorithm

[0082] Import the regional building complex vector map obtained in step 3 (including attribute information fields such as the location coordinates, contour lines, height, and type of each building) into the RHINOCEROS 3D environment in graphical form. Use layer filtering and classification to generate a 3D model of the building complex space through footprint stretching. This model reflects the regional building space layout and building adjacency relationships in the real scene. In this step, this model will serve as the basic building framework to be further simplified.

[0083] To simulate the energy demand of building complexes at the regional scale, a reasonable simplification algorithm needs to be designed to improve the computational efficiency of the model. In previous studies, it was customary to use methods such as simplifying the heat balance process (R-C model) or reducing the computational physical domain to achieve a rough description of the details of the building energy consumption model. And this method proposes to use the verified shoebox algorithm to reduce the scale of the computational heat zone of the existing 3D model of the regional-level building complex space. In order to greatly improve the computational efficiency of the model on the basis of ensuring the reflection of the real building physical scene in the region.

[0084] The shoebox algorithm (jointly developed by Cornell University and the Massachusetts Institute of Technology) is an effective simplification method that can be used to handle complex building physical models and regional-level building complex distribution scenarios. This algorithm has been verified many times at the single-building and regional levels. It abstracts building entities of arbitrary shape into a set of simplified "shoeboxes" to represent the theoretical energy consumption model of the building. At the same time, this algorithm takes into account the influence of the external environment of regional buildings, self / mutual shading of buildings, and thermal disturbances of adjacent buildings on the energy consumption of the building itself. Therefore, it is fully applicable to simulating the dynamic loads of individual buildings in the regional scenario.

[0085] The simplification of the computational heat zone requires a trade-off. By sacrificing some geometric details of the building to a certain extent, higher computational efficiency can be obtained in return. The shoebox algorithm can transform buildings of any shape into standardized shoeboxes according to their storey heights. The shoebox represents a general approximation of the available space layout of the entire building. The workflow of the shoebox algorithm is divided into three main steps: (1) shoebox generation; (2) simulation of the solar radiation reception scenario; (3) analysis of thermal disturbances of adjacent buildings. Its operation framework is as Figure 5 shown.

[0086] In Figure 5 the shoebox creation process shown, three core indicators, namely the edge ratio (er), elongation ratio (el), and shape factor (sf), are used to describe the geometric characteristics of the actual building. The formulas corresponding to the indicators are as follows:

[0087]

[0088] Among them, x, y, and z represent the width, depth, and height of the core shoebox respectively.

[0089] These metrics are obtained after performing a sensitivity analysis on geometric-related factors that affect building energy consumption. Considering the different load conditions existing within a single building, it is necessary to further subdivide the thermal zones of all building units. This process divides a unit into multiple floors and the perimeter spaces of each floor, specifically including the core shoebox area and the perimeter area, according to the automatic zoning principle outlined in ASHRAE 90.1. Finally, the total floor area, core area, and perimeter area of each floor of each building unit are calculated and recorded.

[0090] Building exterior environment and self / mutual shading modeling: Considering the reduction in incident solar radiation due to the self-shading of the actual building and the surrounding environment, a portion of the windows of the building unit are replaced with non-transparent surfaces having the same thermal transmittance as the window. On this basis, the transparent envelope of the building is classified into two types: heat-transmitting and light-transmitting, and heat-transmitting and non-light-transmitting. To correctly demarcate the sizes of these two types of surfaces, a global radiation analysis is performed on the starting geometry and the reference shoebox respectively. The average solar incident radiation amount in each direction and floor is calculated to obtain the equivalent blocking ratio. For each window of the simplified model, the area of the heat-transmitting and non-light-transmitting type elements is calculated based on their corresponding equivalent blocking ratios. This surface is represented on the model as having the same window bottom as the heat-transmitting and light-transmitting type envelope structure, while the calculated height is matched to the equivalent area.

[0091] Adjacent building merging: For a building complex at the regional scale, it is crucial to consider the mutual thermal interaction between the exterior envelopes of adjacent building units. The thermal contact between adjacent buildings is regarded as an adiabatic surface. The shoebox algorithm calculates the equivalent adiabatic ratio (EAR ∈ (0,1)) of each surface by examining the potential adjacency relationships in all directions on each floor of the building, that is, the surface is between being fully exposed to the external environment or fully adiabatic. The equivalent adiabatic ratio is used to calculate the equivalent adiabatic area of the shoebox surface. The adiabatic surface has the same height as the initial surface, and the calculated bottom edge length is matched to the obtained equivalent area.

[0092] Step 4 uses the shoebox algorithm to process the three-dimensional spatial model of the regional building complex and transforms it into a thermophysical model of the regional building complex capable of performing dynamic energy simulation.

[0093] Step 5: Construction of the theoretical framework for the building-level occupancy model

[0094] The energy consumption and transfer behaviors of people are complex and random. It is crucial to describe the variation characteristics of space occupancy rate over time. For a single building, people mainly focus on the occupancy and transfer behaviors of people in different types of rooms; when the calculation scale expands to the regional level, the types and processes of people's behaviors should be reasonably simplified. In the regional building complex environment containing multiple building types, the focus of the problem should be shifted to depicting the state transition probabilities of different types of people in different functional buildings. In the method of the present invention, a Markov chain probability transfer model with the type of people as the core is proposed to depict the random occupancy behaviors of people in the region. This method sacrifices the complexity of people's behaviors in exchange for the typical occupancy situations of various types of buildings at the regional level. Starting from the people's occupancy behaviors at the building level, the prediction scale of the Markov chain model is extended from the single building level to the regional scope.

[0095] Figure 6 The theoretical framework of the building-level people occupancy model is shown. Starting from the people's occupancy behaviors themselves and combining the idea of agent interaction, the people agents with significant differences in occupancy and transfer behaviors are divided into three categories: working people, students, and retired elderly people. Three typical buildings that generally participate in central heating in practice are considered: residential buildings, office buildings, and school buildings. In addition, the behavioral differences between weekdays and weekends are also discussed. According to the characteristics and functions of the above agents, the transfer activity chains of different types of people between different functional buildings are depicted. By classifying, counting, and processing the time use information of regional people collected from the survey questionnaires, a time-sequence state transition matrix reflecting the random decision-making process of people is abstracted using the Markov chain mathematical model, and thus a typical occupancy scenario with people as the agent is obtained. Combining the proportion of the number of people of each type in the region, a typical occupancy pattern with buildings as the agent (i.e., the occupancy rate time schedule of the occupants) is summarized. Further, when conducting building energy consumption simulation, according to the per-unit area occupancy rate of people in each type of building, the corresponding total number of occupants in each building in the region per hour is calculated to estimate the heat gain of the people indoors.

[0096] It should be noted that public buildings such as shopping malls, hospitals, and hotels are not within the scope of consideration and calculation of this method. The specific reasons are as follows: First, through actual research, buildings of the above types usually do not participate in municipal central heating and often adopt the self-sufficient method of small heating stations / energy stations during the heating season, so they do not belong to the objects involved in this method. Second, the occupancy behaviors of people in the above buildings (such as shopping, seeing a doctor, traveling, etc.) are highly random and mobile. They neither belong to the probability places where people appear stably daily nor have corresponding "resident" people, and there is a complex time-dynamic correlation with the types of people. Therefore, it is impossible to summarize through the probabilistic description of the activity chains of individual subjects. Based on the above reasons, this method only targets four types of buildings: residential, office, and school, and constructs a building-level occupancy probability model applicable to the regional scale by combining the time usage habits of three typical types of people: working people, students, and the elderly with an hourly resolution.

[0097] Step 6: Division of typical occupancy scenarios at the building level

[0098] The division of typical occupancy scenarios should fully consider the interaction and correlation between different personnel subjects and different building types. That is, each type of person has its corresponding occupied building type. For example, for working people, the building types they may occupy are mainly: residential buildings and office buildings. There is a corresponding occupied building at different times of each day, and there are significant differences between weekdays and weekends. Therefore, according to the considered building types, personnel types, and date types, a total of 10 different typical occupancy scenarios are divided, as Figure 7 shown. From the perspectives of personnel subjects, building subjects, and date type subjects, the figure uses the form of sets and legends to represent all the occupancy situations of different types of people in different functional buildings, and uses arrows of different colors to represent the potential transfer situations of people between buildings.

[0099] Step 7: Characterization of the activity chains of occupants

[0100] Steps 5 and 6 clarify the hierarchical relationship and typical scenarios of the building-level occupancy model of personnel. The key is to describe the behavior rules of each type of person in the form of an activity sequence. At the building level, this is equivalent to the occupancy situations of different buildings.

[0101] Figure 8Characterizes the main behaviors and activity trajectories of various types of personnel on weekdays and weekends, including going to work, going to school, eating, going out, and going home. The activity sequence of personnel in a day is successively divided into four parts: morning, noon, afternoon, and evening, and the occupied building scenes are distinguished by different colors. The activities of personnel are regarded as the transfer of occupancy status at the building level. Among them, school buildings only accommodate students, office buildings are designated to only accommodate working personnel, while residential buildings contain all types of personnel entities. The composition of various types of personnel in the area needs to be further evaluated in combination with the demographic information of the target area.

[0102] Step 8: Solving the personnel state transition probability matrix based on the Markov chain

[0103] The purpose of the building-level personnel occupancy model is to generate random occupancy data with the same characteristics as the personnel time use survey data and transform it into the state transition probabilities of different types of personnel in different functional buildings. The Markov chain (MC-Markov-Chain) can describe the random decision-making process of personnel based on time series. Each state only depends on the previous state and the probability of state change at that moment, and it has the advantages of fewer state numbers and stable probability distribution in the mathematical model. Therefore, the present invention proposes to apply the ordered Markov chain technology as a method for generating the random occupancy schedule of personnel. By creating a transition probability matrix, the time use survey data can be effectively transformed into an occupancy data set representing each building type.

[0104] The determination process of the transition probability matrix includes determining the state probabilities of each point in the state space by processing the occupant time use survey data, jointly solving the time series state transition probability matrix, and using the convergence method to solve the initial state probability.

[0105] In this method, 0-1 variables are used to represent the occupancy status of each person in each functional building. "0" represents no occupancy, and "1" represents occupancy. Taking one day (24 hours) as a cycle, the time step of state conversion is set to one hour, denoted as:

[0106] τ = 0, 1, 2, …, 23 (5)

[0107] Therefore, the Markov chain mathematical model can predict the occupancy probability of personnel in the building at the hourly resolution. To describe the change of personnel occupancy status within a unit time step, the occupancy status sequence and the instantaneous state transition probability are defined as:

[0108] {X(τ)} = {X(τ), τ = 0, 1, 2, …, 23}, X(τ) ∈ S (6)

[0109] P ij (τ) = P{X(τ + 1) = j|X(τ) = i} (7)

[0110] In the formula, the value set S of X(τ) is called the state space, representing a finite number of indoor state points. Here, S ∈ {0, 1};

[0111] P ij (τ) represents the probability that for a certain type of building, a certain person transfers to state j after a time step under the condition of being in state i at time τ. Therefore, the magnitude of this value depends not only on the current state of the person but also on the current time node where the person is located. Figure 9 The following describes this state transition process:

[0112] From this, the transition probability second-order matrix (Equation 8) and the state probability second-order determinant (Equation 9) corresponding to time τ can be further obtained:

[0113]

[0114] X(τ) = [P 0 (τ) P 1 (τ)] (9)

[0115] The two constraint conditions included therein are respectively:

[0116]

[0117] The probabilities that a person is in state "0" and state "1" at time τ + 1 are respectively expressed as:

[0118]

[0119] The state probability of a person at time τ can be jointly expressed by the state probability at the initial state (τ = 0) and the transition probability matrix as:

[0120]

[0121] It is not difficult to see from the above formula that by determining the initial state probability and the transition probability matrix, the corresponding state probability of the occupancy of a person in a certain type of building at any time of a day can be calculated. Therefore, the two core problems to be solved are: (1) how to determine the transition probability matrix of various types of people at each moment of a day based on the time use survey data of various types of people; (2) how to determine the initial occupancy state probability of a person.

[0122] By analyzing the statistical results of the time use survey data, the transition probability matrix of a person is solved. The state transition probabilities between adjacent moments can be obtained by the formulas listed in Table 1. Considering the three types of buildings, three types of people, and two types of dates, there are a total of 432 transition probability matrices.

[0123] Table 1

[0124]

[0125] Note: In the above table represents the number of occupant samples with state i at time τ;

[0126] represents the number of occupant samples with state i at time τ and state transition to j at time τ + 1.

[0127] A calculation example is provided below to illustrate the solution method in the above table and help better understand the practical meaning of the parameters listed in the table.

[0128] Example: For an office building at 11 am on a weekday, assume that among the 420 office workers participating in the time - use survey, 402 respondents said they were in the office at 11 am, and 245 of these respondents said they were not in the office at 12 pm. This means that there are 157 office workers who were in the office at both 11 am and 12 pm. That is, for the office building:

[0129]

[0130] Therefore, the state transition probability within this time step (11:00 am - 12:00 pm) is:

[0131]

[0132] Among them:

[0133] P 10 (11), P 11 (11) < 1, and P 10 (11)+P 11 (11) = 1

[0134] That is, it satisfies the constraint conditions in Equation (10).

[0135] Repeat similar calculations for all subgroups, and then combine Equation (11) to organize all types of state transition probabilities into an ordered transition probability matrix.

[0136] Another core issue is to determine the initial state probabilities P 0 (0) and P 1 (0), Figure 10It shows the logical framework for determining the initial state probability. First, based on the survey results, the state transition probability matrix with an hourly resolution during a day is determined. Assuming a set of initial probabilities, the state probability at τ = 24 is calculated according to Equation (12), and this is used as the new initial value to repeat the above steps. Through repeated calculations until this initial value converges, the result obtained at this time is the actual value of the initial state probability. Finally, the actual hourly occupancy rates of the three types of personnel in various buildings are calculated by substituting the actual initial values. This method uses Python to compile loop iteration calculation code, and according to Figure 10 the calculation process shown, the in-room rate schedule under typical occupancy scenarios is determined.

[0137] Step 9: Construction of the building prototype template library

[0138] In building energy consumption simulation work, in addition to the need to build a detailed building thermophysical model, it is also a key step in constructing an energy analysis model for a regional building complex to configure the simulation parameters required for calculation for each building model in the region. This includes setting the thermal parameters of the building envelope, window-wall ratio, ventilation / air permeability, indoor design temperature, heating control strategy, occupancy rate, equipment usage schedule, etc. for each building in the region. When the calculation scale is a single building, these parameters can be set one by one. However, for a building complex at the regional scale, the method of setting them one by one will undoubtedly greatly increase the time cost of model building. By creating a building prototype template library and matching it with the buildings in the calculation region, batch configuration of the model simulation parameters can be achieved, significantly improving the modeling efficiency. And the accuracy of the model calculation results usually also depends on the integrity and generality of the prototype template. For this reason, this method creates a set of building prototype template libraries based on the hierarchical relationship between variable simulation parameters and assigns them to each building in the region.

[0139] For different regional forms, the building prototype template library contains all non-geometric model inputs of the prototype buildings within the specified building complex. It is a subdivision and generalization of various typical building characteristics within the specified region.

[0140] For a regional building complex, its overall energy demand is affected by various factors, including the composition of building types, stock scale, regional meteorological conditions, construction year, building area, thermal performance of transparent and non-transparent building envelope surfaces, building orientation, occupancy rate, heating schedule, indoor environmental temperature control strategy, etc. These factors can be macroscopically divided into geometry-related parameters and geometry-independent factors. Among them, the geometry-related parameters constitute the three-dimensional spatial model of the building complex, while the geometry-independent factors need to be further set as simulation parameters in the model.

[0141] The building prototype template library can batch-configure the simulation parameters of the model. However, the key point is to determine the core characteristic indicators to identify and match the building models. Therefore, first, a preliminary sensitivity analysis is conducted on the above influencing factors, and variable simulation parameters are used to describe the typical characteristics of different buildings.

[0142] Figure 11 It shows the hierarchical analysis of each influencing factor. The geometry-independent factors are further divided into known / inferable core characteristic indicators and unknown / to-be-determined variables. Among them, the core characteristic indicators representing typical buildings include building type (T-Type), construction year (Y-Year), and number of floors (F-Floor). Other unknown variables are restricted by the corresponding building body characteristics and can be determined individually or jointly by one or more core characteristic indicators. For example, the thermal performance of the building envelope is related to the usage type, construction year, and number of floors of the building. Therefore, this simulation variable is jointly determined by the above three core indicators. For the heating period of the building, it is only determined by the building type (residential, office building, or school) indicator alone.

[0143] By analyzing the mutual correlations among the variable simulation parameters, a complete building prototype template library and the corresponding model identification and matching logic are created.

[0144] Using the methods proposed in Step 1 and Step 2, the geometric information and non-geometric vector data of the regional building group can be obtained to master the core characteristics and geometry-related parameters of each building in the region. Furthermore, the building prototype template is established based on the known basic information and further classified based on the determined core characteristic indicators. Table 2 shows the building prototype feature coding system created based on the case area.

[0145] Table 2

[0146]

[0147] Note: Since the case area is located in Tibet Autonomous Region of China, the classification basis for the construction year column in the table is derived from the promulgation and update of local building HVAC design-related standards.

[0148] The simulation variables affecting building energy consumption include both the established modeling information that can be directly obtained and the modeling information that needs to be further defined but cannot be directly obtained. The existing modeling information can be obtained or preliminarily calculated through the methods detailed in Step 1 and Step 2, including building type, building height, and shape factor.

[0149] Modeling information that needs to be further defined cannot be directly extracted through existing technologies, but needs to be supplemented according to established modeling details (core feature indicators) and relevant HVAC industry codes and design standards for the country, region, and specific building types. This information includes the thermal properties of transparent / opaque envelope surfaces, the window-wall ratio (WWR) of the building, the occupancy rate, and the indoor environmental temperature control strategy. Among them, the occupancy rate is determined according to the building-level occupancy probability model construction method proposed in step

[0150] 8 and incorporated into the regional building complex energy analysis model in the form of an occupancy rate schedule.

[0151] Step 10: Prototype identification and template matching

[0152] Based on the complete building prototype template library developed in step 9, the key point of the simulation parameter configuration work is to accurately assign the created building prototype templates to each building in the region to ensure the authenticity and effectiveness of the regional building complex energy analysis model.

[0153] The identification work of building prototypes includes assigning prototype templates according to the extracted building core features. The specific identification and matching logic is as Figure 12 shown. By determining the corresponding feature codes for each building in the region and combining the corresponding building design codes, the batch configuration of calculation simulation parameters is completed.

[0154] The above ten steps have completely and elaborately described the construction process and method of the regional building complex energy analysis simulation model.

[0155] Embodiment

[0156] The implementation area of this embodiment is located in a certain city of an autonomous region. This area contains a total of 37 buildings, with a total heating area of 198,000 m 2 , and the calculated plot ratio is 1.63. The detailed information about the building composition is shown in Table 3. Table 4 lists the input parameters of the regional building complex energy simulation model related to the building prototype template library.

[0157] Table 3

[0158]

[0159] Table 4

[0160]

[0161] By distributing questionnaires on time use and daily routines to the residents in the case study area, the daily life and behavior patterns of local residents were understood. The information collected was divided into three parts: (a) basic information of the residents, including gender, age group, and occupation type; (b) the activity trajectories of residents on weekdays and weekends based on different occupation types and their time correlations; (c) occupancy state transition data at hourly resolution for establishing a daily cycle state transition matrix. During the entire research period, a total of 1000 questionnaires were distributed, and 648 valid questionnaires were recovered, with a valid recovery rate of 64.8%.

[0162] According to the population composition and the statistics of the number of on-the-job / retired personnel in various industries in the "XXX City Statistical Yearbook of XX Autonomous Region in 2022", the distribution ratios of occupant types in residential buildings were set at 62%, 26%, and 12%, corresponding to the groups of working people, students, and retired elderly respectively. Figure 13 The composition of the three types of occupant subjects is shown.

[0163] According to the mathematical models and methods proposed in step 8, the occupancy rate changes of the three building types on weekdays and weekends were obtained by solving, as Figure 14 shown.

[0164] Result rationality test: By comparing the calculation results of the model with the actual monitoring data of the heating station in the case area and the design load calculated by the unit area heat index method. As Figure 15 shown, the regional peak heat load calculated by the model is 5858 kW, while the result estimated by the heat index method is 11880 kW, and the peak load reflected by the actual test record is 5358 kW. Therefore, when using the traditional heat index method for system capacity design, the maximum load coefficient of the system is only 45.1%, indicating that there is a large deviation between the design load and the actual demand. Compared with the index method, the method proposed in the present invention increases the system load coefficient by 46.4%, and the error rate with the actual load is only 9.33%. This proves the effectiveness and practicality of the proposed method in improving system capacity design, enhancing load prediction accuracy, and reducing the initial investment in system design.

[0165] In summary, the present invention proposes a building-level personnel occupancy probability model for building complexes at the regional level, expanding the consideration of the personnel occupancy rate from the single space / room scale to the building level, and adopting an agent-based modeling thinking. The present invention fully considers the complex and multi-level heat use scenario problems within the region caused by personnel activities and occupancy behaviors; therefore, the method of the present invention can make the building heating load prediction more detailed and accurate.

[0166] It should be emphasized that the above are only the preferred embodiments of the present invention, and there is no restriction on the present invention in any form. Any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A method for predicting heating load of a building complex based on a shoebox algorithm and considering occupancy rate, characterized in that: The following steps are involved: Step 1: Collecting geometric feature information of regional building complexes; Step 2: regional non-geometric vector data interception; Step 3: Multi-source vector data fusion; Step 4: Building thermal zone division based on shoebox algorithm; Step 5: Construction of theoretical framework of building-level occupancy model; Step 6: Classification of typical occupancy scenarios at the building level; Step 7: Characterization of residents' activity chain; Step 8: Solve the personnel state transition probability matrix based on the Markov chain; Step 9: Building prototype template library construction; Step 10: Prototype recognition and template matching.

2. The method for predicting heating load of a building group based on the shoebox algorithm and considering occupancy rate according to claim 1 is characterized in that: In the step 1, the collected regional building complex geometric feature information is converted into vector data; The geometric feature information of the regional building complex includes: detailed information of the buildings in the region, the detailed information includes: bottom contour, floor area, relative position in space, and height; In the step 2, the point of interest POI and the area of ​​interest AOI vector data are integrated, and the land use type data and the historical satellite cloud map section are combined to jointly determine the use type and construction year of the buildings in the area; the regional non-geometric vector data includes: point of interest POI and area of ​​interest AOI; the point of interest POI represents a point vector data set of the building, and the area of ​​interest AOI represents a regional geographic entity of the building in the map; In step 3, by introducing attribute fields for each closed building outline in the two-dimensional plane, the building footprints without main attributes in the area are converted into building footprints with main attributes, and the geometry and functional information of each building in the area are summarized into a building attribute list, and a two-dimensional vector map of the area is generated.

3. The method for predicting heating load of a building group based on the shoebox algorithm and considering occupancy rate according to claim 2 is characterized in that: In the step 4, the regional building complex vector map obtained in step 3 is used to generate a three-dimensional model of the building complex space, and the shoebox algorithm is used to process the three-dimensional model of the building complex space to convert the three-dimensional model of the building complex space into a regional building complex thermal physical model capable of performing dynamic energy simulation; The core indicators used in the creation process of the shoebox algorithm: edge ratio er, elongation el and shape factor sf are: In formulas (2), (3), and (4), x, y, and z represent the width, depth, and height of the core shoe box, respectively.

4. The method for predicting heating load of a building group based on the shoebox algorithm and considering occupancy rate according to claim 1 is characterized in that: In step 2, when vector data is insufficient or missing, national land type coverage data is used to assist in judgment and verification; the construction year interval of the building is determined by comparing historical satellite images; the data providers of the national land type coverage data include Google Earth, Baidu Maps, Amap, and Elevation Maps.

5. The method for predicting heating load of a building group based on the shoebox algorithm and considering occupancy rate according to claim 1 is characterized in that: In step 5, a Markov chain probability transfer model with personnel type and building use function as the core is created to characterize the temporal random occupation behavior of personnel in the area; Starting from the occupancy behavior of people at the building level, the prediction scale of the Markov chain model is expanded from the single building level to the regional scope; The classification of the said personnel includes: working people, students and retired elderly people; The classification of said buildings includes: residential, office and school buildings; In step 5, a Markov chain model is used to abstract a time-series state transfer matrix that reflects the random decision-making process of personnel, thereby obtaining a typical occupancy scenario with personnel as the main body; combined with the proportion of each type of personnel in the area, a typical occupancy pattern with buildings as the main body is summarized; when performing building energy consumption simulation, the total number of occupants per unit time of each building in the area is calculated based on the unit area occupancy rate of personnel in each type of building, so as to estimate the heat gain of indoor personnel.

6. The method for predicting heating load of a building group based on the shoebox algorithm and considering occupancy rate according to claim 1 is characterized in that: In step 6, the division of typical occupancy scenarios takes into account the interactive associations between different personnel entities and different building types, and uses sets and legends to represent all occupancy situations of different types of personnel in buildings with different functions, and represents the potential transfer of personnel between buildings; In step 7, the behavior rules of each type of personnel are described in the form of activity sequences, the behavior and activity trajectory of the personnel are characterized, and the occupied building scenes are distinguished; the activities of the personnel are regarded as the transfer of occupancy status at the building level, and residential buildings contain all types of personnel. The composition of various types of personnel in the area is evaluated in combination with the demographic information of the target area.

7. The method for predicting heating load of a building group based on the shoebox algorithm and considering occupancy rate according to claim 6 is characterized in that: In step 8, an ordered Markov chain is used to generate a random occupancy schedule of personnel; and the time use survey data is converted into an occupancy data set representing each building type.

8. The method for predicting heating load of a building group based on the shoebox algorithm and considering occupancy rate according to claim 7 is characterized in that: In step 8, the state probability of the person at time τ is jointly expressed by the state probability at the initial state τ=0 and the transition probability matrix as follows: In formula (12), P0(τ) and P1(τ) represent the state probability matrix and transition probability matrix respectively, and τ represents the unit time.

9. The method for predicting heating load of a building group based on the shoebox algorithm and considering occupancy rate according to claim 1 is characterized in that: In step 9, in the building energy consumption simulation, in addition to building a detailed building thermal physical model, the simulation parameters required for calculation are also configured for the building models in the area; the simulation parameters include: setting the thermal parameters of the envelope structure of each building in the area, the window-to-wall ratio, the ventilation / air permeability, the indoor design temperature, the heating control strategy, the occupancy rate of personnel, and the equipment use schedule; By creating a building prototype template library and matching the building prototype template library with the buildings in the calculation area, batch configuration of model simulation parameters is achieved; and core feature indicators are determined to identify and match building models.

10. The method for predicting heating load of a building group based on the shoebox algorithm and considering occupancy rate according to claim 9 is characterized in that: In the step 10, based on the building prototype template library, the simulation parameter configuration assigns the created building prototype template to the buildings in the area; The identification of architectural prototypes includes: assigning prototype templates according to the extracted architectural core features; By determining the corresponding feature codes for the buildings in the area and combining them with the corresponding building design specifications, batch configuration of calculation and simulation parameters can be completed.

Images