Data-driven method and system for detecting abnormal high energy consumption of buildings at block scale

Abstract

The present invention provides a data-driven, block-scale, building-scale anomaly high energy consumption detection method and system. The method comprises the following steps: obtaining historical energy consumption data for each building in the tested block, determining whether the energy consumption is abnormal and labeling it, thereby obtaining labeled data on the energy consumption of each building; obtaining a dataset of architectural design parameters for each building in the tested block; obtaining a dataset of weather parameters for the city in which the building is located; constructing a detection model and training it based on the datasets of architectural design parameters and weather parameters, and evaluating the training of the detection model based on the labeled data to obtain a trained detection model; obtaining the architectural design parameters of the building to be tested and the weather parameters of the city in which it is located, quantifying them, and then using the detection model to obtain a predicted energy consumption value, and determining whether the energy consumption is abnormal. The present invention can accurately predict whether a building's energy consumption is abnormal, providing a reference for users.

Description

Technical Field

[0001] The present invention relates to the technical field of demand-side energy conservation in urban buildings, and more specifically, to a data-driven method and system for detecting high-energy consumption anomalies in block-scale buildings. Background Art

[0002] Therefore, improving building energy efficiency through energy management and energy conservation policies has become a priority, requiring strict enforcement of building energy codes for new buildings and effective prediction of energy demand for existing buildings. Building managers can use building energy demand monitoring systems to adjust how buildings are managed and arranged, prepare in advance for possible high electricity demand, and support renovation strategies. Providing relevant information to occupants can also encourage them to reduce energy use in buildings. Looking ahead, understanding a building's energy use will not only help identify buildings whose energy use may be higher than their rated demand as a warning sign, but also help determine whether energy use will increase with extreme weather; this knowledge can help city decision makers determine better energy management and effective strategies through monitoring systems.

[0003] A prior art document provides a method and system for diagnosing abnormal energy consumption based on a short-term building energy consumption prediction model (CN114169254A). The method includes Pearson correlation analysis and data preprocessing to construct an energy consumption sample set, which is then divided into a training set, a validation set, and a test set. It also proposes a multi-step-ahead short-term energy consumption prediction model using an LSTM network based on 10 input parameters, including historical energy consumption data, occupant behavior, meteorological factors, and time factors. It also uses a MIMO strategy to predict hourly building energy consumption using supervised learning. Based on the prediction model, a diagnostic method based on the mathematical relationship between predicted and actual values ​​is proposed on a smaller time scale. Problems with the prior art include: The urban energy consumption prediction method in Comparative Document 1 lacks urban morphological parameters that have a key impact on energy consumption; and it also lacks a classification of building energy consumption levels in cities and a warning diagnosis of abnormally high energy consumption. Summary of the Invention

[0004] In order to address the deficiencies in the prior art, the purpose of the present invention is to provide a data-driven method for detecting high-energy consumption anomalies in block-scale buildings. By combining the specific design parameters of the building and the meteorological parameters of the city in which it is located, the energy consumption characteristics of urban buildings are extracted, the design parameters and meteorological characteristics are quantified, and machine learning is integrated into the block-scale energy consumption detection to achieve high-energy consumption anomaly detection of block-scale buildings.

[0005] The present invention adopts the following technical solutions.

[0006] A data-driven method for detecting abnormal high energy consumption of buildings at the block scale, comprising the following steps:

[0007] Step 1: Obtain historical energy consumption data of each building in the measured block, determine whether the energy consumption is abnormal and mark it, and obtain the marked data of energy consumption of each building;

[0008] Step 2: Obtain the architectural design parameters of each building in the measured block and quantify them to obtain a dataset of architectural design parameters;

[0009] Step 3: Obtain the weather parameters of the city where the building is located, and quantify them to obtain a weather parameter dataset;

[0010] Step 4: Build a detection model and train it based on the dataset of building design parameters and weather parameters. Evaluate the training of the detection model based on the labeled data obtained in step 1 to obtain a trained detection model.

[0011] Step 5: Obtain the architectural design parameters of the building to be inspected and the weather parameters of the city where it is located, quantify them, obtain the predicted energy consumption value through the detection model, and determine whether its energy consumption is abnormal.

[0012] Preferably, the step 1 further comprises:

[0013] Step 1-1, obtaining historical energy consumption data of the detected block-scale building complex, the historical energy consumption data including electricity consumption data;

[0014] Step 1-2: Calculate the energy consumption intensity per unit area of ​​the building based on electricity consumption and building area and perform normalization.

[0015] Steps 1-3: setting energy consumption thresholds for each building based on the average energy consumption intensity per unit area of ​​the building;

[0016] Steps 1-4: determine whether the energy consumption of each building is abnormal based on the energy consumption threshold of each building, mark the monthly energy consumption intensity of each building, and obtain marked data.

[0017] Preferably, in step 1-2, the standardized calculation formula for energy consumption intensity is as follows:

[0018]

[0019] in, is the normalized energy intensity per unit area of ​​the i-th building in the j-th month, EUI i,j is the actual energy consumption intensity per unit area of ​​the i-th building in the j-th month, is the energy consumption intensity per unit area of ​​the building per month during the period of minimum value obtained for the i-th building, It is the maximum value of the energy consumption intensity per unit area of ​​the i-th building in each month during the acquired time period.

[0020] Preferably, in steps 1-3, the calculation formula for each building threshold is:

[0021]

[0022] in, is the energy consumption threshold of the i-th building, n is the total number of months during the study period, EUI i,j is the actual monthly energy consumption intensity per unit area of ​​the i-th building in the j-th month.

[0023] Preferably, in steps 1-4, if the energy consumption intensity per unit area of ​​the building per month is greater than a set threshold, the corresponding data point is marked as 1; otherwise, it is marked as 0.

[0024] Preferably, in step 2, the building design parameters include basic building form parameters, building interaction parameters and building function parameters;

[0025] The basic building morphology parameters also include: the total floor area of ​​the target building, the total number of floors of the target building, the height of the target building, the width of the target building, the length of the target building, the perimeter of the target building, the surface area of ​​the target building in the dataset, the volume of the target building, and the orientation of the target building.

[0026] The parameters of mutual influence between buildings also include: building attribute parameters and parameters of the influence of the surrounding environment on the building; the building attribute parameters also include the body coefficient of the target building, the ratio of the perimeter area of ​​the target building, and the ratio of the length and width of the target building, where the body coefficient refers to the ratio of the volume to the surface area of ​​the building; the parameters of the influence of the surrounding environment on the building also include the ratio of the south obstacle height to the street canyon width, the ratio of the west obstacle height to the street canyon width, the ratio of the north obstacle height to the street canyon width, and the ratio of the east obstacle height to the street canyon width;

[0027] Building function parameters also include: public services, education, industry, hospitals and hotels. The building functions of each building are quantified to obtain the building function data of each building.

[0028] Preferably, the quantifying of the building functions of each building further comprises:

[0029] If the building belongs to the public service category, the corresponding building function data is 0; if the building belongs to the education category, the corresponding building function data is 1; if the building belongs to the industrial category, the corresponding building function data is 2; if the building belongs to the hospital category, the corresponding building function data is 3; if the building belongs to the hotel category, the corresponding building function data is 4.

[0030] Preferably, the types of weather parameters obtained include temperature, felt temperature, humidity, air pressure, wind speed value, wind direction, rainfall per hour, snowfall per hour, cloud cover, number of weather condition descriptions, number of hours with uncomfortable temperature, number of hours with uncomfortable humidity and number of hours with uncomfortable temperature.

[0031] Preferably, the data value of the perceived temperature can be quantified using the following formula:

[0032] Feels_like=-42.379+(2.04901523×T)+(10.14333127×rh)-(0.22475541×T×rh)-(6.83783×10 -3 ×T 2 )-(5.481717×10 -2 ×rh 2 )+(1.22874×10 -3 ×T 2 ×rh)+(8.5282×10 -4 ×T×rh 2 )-(1.99×10 -6 ×T 2 ×rh 2 ),

[0033] Among them, Feels_like is the feeling temperature, that is, the heat index, T is the air temperature, and rh is the relative humidity. The air temperature and relative humidity can be directly obtained through weather software or the city meteorological bureau.

[0034] Preferably, the quantification of the number of hours with uncomfortable temperature, the number of hours with uncomfortable humidity and the number of hours with uncomfortable temperature also includes: temperatures outside the range of 20-27 degrees Celsius are considered uncomfortable, and humidity outside the range of 30-65% is considered uncomfortable.

[0035] Preferably, the step 4 further comprises:

[0036] Step 4-1: Combine the building design parameter dataset obtained in step 2 and the weather parameter dataset obtained in step 3 to construct a dataset for detection model training, and divide it into a training set and a test set;

[0037] Step 4-2: Build a detection model based on the support vector machine algorithm, and train and evaluate the detection model based on the training set, test set, and the building marker data obtained in step 1;

[0038] In step 4-3, the training of the detection model is determined to be complete based on the evaluation results, and the trained detection model is output.

[0039] Preferably, step 4-2 further includes: calculating the accuracy of the prediction result of the detection model based on the test set as an evaluation indicator to evaluate the detection model, including:

[0040] The detection model is tested using test set data. Based on the building energy consumption prediction value output by the detection model, the judgment method in step 1 is used to determine whether the energy consumption of each building is abnormal. Combined with the labeled data of each building's energy consumption in step 1, the labeled data is compared with the results judged based on the output value of the detection model. The prediction accuracy of the detection model for the energy consumption of each building is calculated, and the performance of the detection model on the test set is evaluated.

[0041] The present invention also provides a block-scale building high energy consumption anomaly detection system, comprising: a data acquisition module, a parameter processing module, a training module and a detection module;

[0042] Among them, the data acquisition module is used to collect energy consumption data, urban morphology data and meteorological data of the detected area;

[0043] The parameter processing module can process the data acquired by the data acquisition module to obtain the training set and test set required for training, as well as the data that can be directly input into the detection module in actual detection;

[0044] The training module can train the detection model constructed in the detection module based on the training set and the test set;

[0045] The detection module can detect abnormalities in building energy consumption in the detected area based on the trained detection model and obtain detection results.

[0046] The present invention also provides a terminal, comprising a processor and a storage medium;

[0047] The storage medium is used to store instructions;

[0048] The processor is used to operate according to the instructions to execute the steps of the data-driven block-scale building high energy consumption anomaly detection method.

[0049] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the data-driven block-scale building high energy consumption anomaly detection method.

[0050] The beneficial effect of the present invention is that, compared with the existing technology, the present invention can extract urban morphology based on building complex information at the community scale, create meteorological characteristics in combination with the meteorological profile of the target city, and combine it with the data-driven algorithm of machine learning to identify buildings on the demand side of urban building complexes that will generate abnormally high energy consumption demands in the next month of use. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is the overall flow chart of the data-driven block-scale building high energy consumption anomaly detection method in the present invention;

[0052] Figure 2 is the normalized energy consumption data of the example building complex;

[0053] Figure 3 It is a schematic diagram of the data flow of the detection system;

[0054] Figure 4 is a schematic diagram of the detection system results;

[0055] Figure 5 It is a structural diagram of the data-driven block-scale building high energy consumption anomaly detection system in the present invention. DETAILED DESCRIPTION

[0056] The present application will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and are not intended to limit the scope of protection of the present application.

[0057] like Figure 1 As shown, the present invention provides a data-driven method for detecting abnormal high energy consumption of buildings at a block scale, which specifically includes the following steps:

[0058] Step 1: Obtain historical energy consumption data of each building in the measured block, determine whether the energy consumption is abnormal and mark it, and obtain the marked data of energy consumption of each building;

[0059] Specifically, step 1 also includes:

[0060] Step 1-1, obtaining historical energy consumption data of the detected block-scale building complex, the historical energy consumption data including electricity consumption data;

[0061] The electricity consumption data of the block-scale building complex to be tested is obtained through the power supply bureau in the block's location. The obtained electricity consumption data can be accurate to monthly data, and the length of the obtained historical data is at least 3 years.

[0062] Step 1-2: Calculate the energy consumption intensity per unit area of ​​the building based on electricity consumption and building area and perform normalization.

[0063] Specifically, the energy intensity per unit area of ​​the building is obtained by dividing the electricity consumption by the building area. In order to more clearly understand the electricity consumption trend of each month during the study period, all energy intensity data per unit area of ​​the building need to be normalized according to the building. The normalized calculation formula for energy intensity is as follows:

[0064]

[0065] in, is the normalized energy intensity per unit area of ​​the i-th building in the j-th month, EUI i,j is the actual energy consumption intensity per unit area of ​​the i-th building in the j-th month, is the energy consumption intensity per unit area of ​​the building per month during the period of minimum value obtained for the i-th building, It is the maximum value of the energy consumption intensity per unit area of ​​the i-th building in each month during the acquired time period.

[0066] Steps 1-3: setting energy consumption thresholds for each building based on the average energy consumption intensity per unit area of ​​the building;

[0067] Specifically, a threshold is set for each building using the average energy intensity per unit area of ​​the target building plus the standard deviation of its energy intensity per unit area. The calculation formula for the threshold of each building is:

[0068]

[0069] in, is the energy consumption threshold of the i-th building, n is the total number of months during the study period, EUI i,j is the actual monthly energy consumption intensity per unit area of ​​the i-th building in the j-th month.

[0070] Steps 1-4: determine whether the energy consumption of each building is abnormal based on the energy consumption threshold of each building, mark the monthly energy consumption intensity of each building, and obtain marked data.

[0071] The monthly energy consumption intensity of each building is used as the target variable, and the target variable is converted into labeled data, including: judging whether the monthly energy consumption intensity per unit area of ​​each building is normal or abnormal based on the threshold, with normal marked as 1 and abnormal marked as 0. The labeling logic of the target variable is:

[0072]

[0073] ELSE label=0

[0074] That is, if the energy consumption intensity per unit area of ​​the building in each month is greater than the set threshold, the corresponding data point is marked as 1; otherwise, it is marked as 0.

[0075] Step 1 finally obtains the labeled data of energy consumption intensity of each building each month as 0 or 1.

[0076] Step 2: Obtain the architectural design parameters of each building in the measured block and quantify them to obtain a dataset of architectural design parameters;

[0077] Among them, step 2 obtains the data set of building design parameters in the detected block, and the building design parameters include basic building form parameters, building interaction parameters and building function parameters.

[0078] Specifically, the basic building form parameter dataset contains the physical parameters of the building, which can be obtained from digital online maps, urban GIS data, or open source 3D city models. These parameters are usually set during the design phase of the building and do not change as the building ages.

[0079] In conjunction with Table 1 below, the basic building morphology parameters obtained in the present invention include: the total floor area of ​​the target building, the total number of floors of the target building, the height of the target building, the width of the target building, the length of the target building, the perimeter of the target building, the surface area of ​​the target building in the data set, the volume of the target building and the orientation of the target building, where the orientation of the target building is the angle of the building, with the east orientation being 0 degrees and the south orientation being 90 degrees.

[0080] The obtained data sets of parameters of mutual influence between buildings include data sets of building attribute parameters and data sets of parameters of influence of surrounding environment on buildings;

[0081] Among them, the building attribute parameters are calculated based on the basic building design parameters and will not change over time. The building attribute parameters obtained by the present invention also include the body coefficient S_V of the target building, the ratio of the perimeter area of ​​the target building P_A, and the ratio of the length and width of the target building BAR, where the body coefficient S_V refers to the ratio of the volume to the surface area of ​​the building; the parameters affecting the surrounding environment on the building also include the ratio of the south obstacle height to the street canyon width HW_South, the ratio of the west obstacle height to the street canyon width HW_West, the ratio of the north obstacle height to the street canyon width HW_North, and the ratio of the east obstacle height to the street canyon width HW_East.

[0082] Furthermore, the calculation method for quantifying the parameters of the surrounding environment's impact on the building is as follows:

[0083]

[0084] Among them, i∈[South,West,North,East], H i，o，f Indicates the highest floor height of the building in the i-th direction, H i，f represents the height of the fth layer in the i-th direction, W i represents the street width in the i-th direction, and k represents the number of blocked layers.

[0085] Furthermore, the building functions in the present invention also include the following categories: public services, education, industry, hospitals and hotels. The building functions of each building are quantified to obtain the building function data of each building.

[0086] Specifically, the quantification of building functions also includes: if the building belongs to the public service category, the corresponding building function data is 0; if the building belongs to the education category, the corresponding building function data is 1; if the building belongs to the industrial category, the corresponding building function data is 2; if the building belongs to the hospital category, the corresponding building function data is 3; if the building belongs to the hotel category, the corresponding building function data is 4.

[0087] The architectural design parameters of each building obtained in step 2 are shown in Table 1 below.

[0088] Table 1. Architectural design parameters and meanings of each building

[0089]

[0090]

[0091] Step 3: Obtain the weather parameters of the city where the building is located, and quantify them to obtain a weather parameter dataset;

[0092] The weather parameter data includes historical weather data and raw data. The historical weather data can be obtained through weather software or city meteorological bureau. The present invention obtains historical weather data through OpenWeather application software. The raw data includes local hourly weather data.

[0093] Specific, directly measurable meteorological parameters include temperature

[0094] The weather parameters obtained by the present invention include temperature Temp, felt temperature Feels_like, humidity Humidity, air pressure Pressure, wind speed Wind_speed, wind direction Wind_deg, hourly rainfall Rain_1h, hourly snowfall Snow_1h, cloud cover Clouds_all, weather condition description number Weather_description, number of hours with uncomfortable temperature Temp_count, number of hours with uncomfortable humidity Humidity_count, and number of hours with uncomfortable temperature Feels_like_count.

[0095] Since building energy consumption data is recorded on a monthly basis, the present invention summarizes the acquired weather parameter data on a monthly basis, thereby ensuring the consistency of weather time granularity and energy consumption data.

[0096] Among them, the specific values ​​of temperature, humidity, air pressure, and rainfall can be directly obtained, and the data value of the perceived temperature can be quantified using the following formula:

[0097] Feels_like=-42.379+(2.04901523×T)+(10.14333127×rh)-(0.22475541×T×rh)-(6.83783×10 -3 ×T 2 )-(5.481717×10 -2 ×rh 2 )+(1.22874×10 -3 ×T 2 ×rh)+(8.5282×10 -4 ×T×rh 2 )-(1.99×10 -6 ×T 2 ×rh 2 ),

[0098] Among them, Feels_like is the feeling temperature, that is, the heat index, T is the air temperature, and rh is the relative humidity. The air temperature and relative humidity can be directly obtained through weather software or the city meteorological bureau.

[0099] The calculation of the feeling temperature Feels_like combines conventional temperature and humidity to better express the environmental state and determine the potential energy consumption demand under the environmental state.

[0100] In order to use weather data to better understand the impact of weather conditions on human activities, the number of hours of uncomfortable temperature and humidity per month is counted based on the hourly temperature and humidity in the area where the building is located; the number of hours of uncomfortable temperature Temp_count, the number of hours of uncomfortable humidity Humidity_count, and the number of hours of uncomfortable temperature Feels_like_count are quantified, including dividing the uncomfortable temperature range and humidity range, and counting the hours of corresponding parameters based on temperature Temp, temperature Feels_like, humidity, and the divided range.

[0101] The weather parameters and quantification methods of the city where the building is located in step 3 are shown in Table 2 below.

[0102] Table 2 Variables and summary methods of weather parameters

[0103]

[0104] Step 4: Build a detection model and train it based on the data set of building design parameters and weather parameters. Evaluate the training of the detection model based on the labeled data obtained in step 1 to obtain a trained detection model.

[0105] Specifically, step 4 also includes:

[0106] Step 4-1: Combine the building design parameter dataset obtained in step 2 and the weather parameter dataset obtained in step 3 to construct a dataset for detection model training, and divide it into a training set and a test set;

[0107] The building design parameter dataset obtained in step 2 and the weather parameter dataset obtained in step 3 are combined together to form a dataset that serves as the input data of the detection model; specifically, the input data points of the detection model are a combination of a set of building design parameters and monthly weather parameter summary data.

[0108] Preferably, all the input data of the detection model are labeled, including: first obtaining the basic data of different buildings, such as the length, width, height, orientation, function, and coordinates of the building, and then calculating the building design parameters in Table 1, and assigning the calculated parameters to different buildings (Bldg1, Bldg2, Bldg3, ..., Bldg n ) to form a static table (Staging Table) of the detection system, and label the city climate data according to DATA-XXX, such as DATA1-TEMP, DATA2-FEELLIKE, etc., to form a dynamic table (Operational Table) of the detection system, and then embed it into the energy consumption database (ENERGY META).

[0109] Step 4-2: Build a detection model based on the support vector machine algorithm, and train and evaluate the detection model based on the training set, test set, and the building marker data obtained in step 1;

[0110] Specifically, the detection of abnormally high energy consumption can be achieved using basic machine learning algorithms. The present invention applies the support vector machine (SVM) algorithm, which is easy to implement and has good predictive performance, to build a detection model. GridsearchCV is used for model tuning to help each model obtain the optimal prediction results.

[0111] Among them, the support vector machine constructs one or a set of hyperplanes in a high-dimensional or infinite-dimensional space, which can be used for classification, regression, or other tasks. The detection model built based on the SVM algorithm can be determined by solving the following equation:

[0112]

[0113] Among them, xi is labeled y i The training sample, inner product plus intercept <ω,x i >+b is the prediction of the sample, ε represents the threshold, and the value of ε can be any parameter.

[0114] In the present invention, the performance of the detection model on the training set is used to evaluate and compare the algorithms. In terms of parameter setting, the regularization parameter C, gamma, and calculation kernel (kernel value) of the detection model need to be set.

[0115] Furthermore, after the constructed detection model is trained with the data of the training set, the accuracy of the detection model prediction results is calculated based on the test set as an evaluation indicator to evaluate the detection model, specifically including:

[0116] The detection model is tested using test set data. Based on the building energy consumption prediction value output by the detection model, the judgment method in step 1 is used to determine whether the energy consumption of each building is abnormal. Combined with the labeled data of each building's energy consumption in step 1, the labeled data is compared with the results judged based on the output value of the detection model. The prediction accuracy of the detection model for the energy consumption of each building is calculated, and the performance of the detection model on the test set is evaluated.

[0117] The calculation formulas for accuracy are as follows:

[0118]

[0119] Among them, True Positive represents the abnormally high energy consumption buildings that are correctly predicted;

[0120] False Positive indicates an abnormally high energy consumption building that is incorrectly predicted;

[0121] Total indicates the total number of buildings detected.

[0122] In step 4-3, the training of the detection model is determined to be complete based on the evaluation results, and the trained detection model is output.

[0123] Specifically, according to the evaluation indicator Accuracy calculated in step 4-2, when the Accuracy is above 80%, it indicates that the training is completed, and the training of the detection model is ended at this time.

[0124] Step 5: Obtain the architectural design parameters of the building to be inspected and the weather parameters of the city where it is located, quantify them, obtain the predicted energy consumption value through the detection model, and determine whether its energy consumption is abnormal.

[0125] Specifically, the architectural design parameters of the building to be inspected and the weather parameters of the city where it is located are quantified to obtain a parameter data set, which is then input into the trained detection model. The detection model outputs the predicted building energy consumption value for the next month of the building to be inspected; the threshold of the building to be inspected is calculated based on the threshold calculation method in step 1, and the building energy consumption prediction value and the threshold of the building to be inspected are combined to determine whether its energy consumption is abnormal.

[0126] Furthermore, buildings with normal predicted energy consumption are marked as 1, and buildings with abnormal energy consumption are marked as 0 for user reference.

[0127] like Figure 5 As shown, the present invention also provides a data-driven block building high energy consumption anomaly detection system. The above-mentioned data-driven block building high energy consumption anomaly detection method can be implemented based on this system. Specifically, the system includes a data acquisition module, a parameter processing module, a training module and a detection module;

[0128] Among them, the data acquisition module is used to collect energy consumption data, urban morphology data and meteorological data of the detected area;

[0129] The parameter processing module can process the data acquired by the data acquisition module to obtain the training set and test set required for training, as well as the data that can be directly input into the detection module in actual detection;

[0130] The training module can train the detection model constructed in the detection module based on the training set and the test set;

[0131] The detection module can detect abnormalities in building energy consumption in the detected area based on the trained detection model and obtain detection results.

[0132] In order to verify the beneficial effects of the present invention, an actual city is selected as a case to illustrate the steps of the present invention:

[0133] Step 1: Select a case city and obtain its corresponding dataset

[0134] This study selected 71 buildings to ensure that we have the longest possible time series data and that all buildings have the same energy consumption as the model input data. In order to more clearly present the electricity consumption and its trend for each month over the four-year period, all EUI data are normalized to the building unit using Formula 1. Figure 2 As shown, Figure 2 The monthly electricity consumption trends for several randomly sampled buildings in the dataset are shown, along with the normalized monthly EUI distribution for all buildings in each month.

[0135] Step 2. Extract architectural design parameters for the selected cases

[0136] The design parameters of the urban building complex in the embodiment are quantitatively compared using the architectural design parameter table in Table 1. The quantitative results are shown in Table 3 below.

[0137] Table 3. Design parameter feature extraction of the building set of the embodiment

[0138]

[0139]

[0140] The width and height of the buildings are relatively concentrated, while the length distribution is relatively discrete. The shape coefficient and perimeter-to-area ratio of these buildings are mostly distributed between 0.28-0.40 and 0.2-0.27, respectively. The height and canyon width of the buildings in the other four directions can, to a certain extent, reflect the differences in the urban environment of the buildings. In terms of function, the 71 buildings in the example are mostly used for public services (50.6%), followed by education (21.1%), industry (15.5%), hospitals (9.9%), and hotels (2.8%).

[0141] Step 3. Extract meteorological parameters for the selected case city

[0142] The weather of the city where the embodiment is located is collected and analyzed using the environmental parameter table in Table 2. The statistical results of the embodiment are shown in Table 4, which generates monthly weather data characteristic values ​​summarized from the hourly data of the case study location.

[0143] Table 4. Design parameter feature extraction of the building set of the embodiment

[0144]

[0145]

[0146] Then the energy consumption characteristics, building design parameter characteristics, and meteorological parameter characteristic values ​​can be integrated and compiled into the database, such as Figure 3 shown.

[0147] Step 3. Prediction of abnormal high energy consumption in buildings

[0148] The detection model is evaluated by its performance on the test dataset, setting C to 10000, gamma to 0.01 and kernel to rbf. The algorithm achieves an F1 score and accuracy score of 0.569 and 0.854 respectively.

[0149] Then, the target building can be selected, its historical energy consumption data can be obtained, and its high energy consumption abnormality can be predicted and judged based on the detection model. After testing and inspection, the detection model can detect energy consumption abnormalities of the target building with an accuracy of more than 99%.

[0150] The beneficial effect of the present invention is that, compared with the existing technology, the present invention can extract urban morphology based on building complex information at the community scale, create meteorological characteristics in combination with the meteorological profile of the target city, and combine it with the data-driven algorithm of machine learning to identify buildings on the demand side of urban building complexes that will generate abnormally high energy consumption demands in the next month of use.

[0151] The present disclosure may be a system, method and / or computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0152] A computer-readable storage medium can be a tangible device that can hold and store instructions for use by an instruction execution device. A computer-readable storage medium can be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination thereof. As used herein, a computer-readable storage medium is not to be construed as a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through an electrical wire.

[0153] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions to be stored in the computer-readable storage medium in each computing / processing device.

[0154] The computer program instructions for performing the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, and conventional procedural programming languages ​​such as "C" language or similar programming languages. Computer-readable program instructions may be executed entirely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., utilizing an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be personalized by utilizing the state information of the computer-readable program instructions. The electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present disclosure.

[0155] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. A data-driven method for detecting abnormal high energy consumption in buildings at the block scale, characterized by: The steps include: Step 1: Obtain historical energy consumption data of each building in the measured block, determine whether the energy consumption is abnormal and mark it, and obtain the marked data of energy consumption of each building; The step 1 further comprises: Step 1-1, obtaining historical energy consumption data of the detected block-scale building complex, the historical energy consumption data including electricity consumption data; Step 1-2: Calculate the energy consumption intensity per unit area of ​​the building based on electricity consumption and building area and perform normalization. Steps 1-3: setting energy consumption thresholds for each building based on the average energy consumption intensity per unit area of ​​the building; Steps 1-4: determine whether the energy consumption of each building is abnormal based on the energy consumption threshold of each building, mark the monthly energy consumption intensity of each building, and obtain marked data; Step 2: Obtain the architectural design parameters of each building in the measured block and quantify them to obtain a dataset of architectural design parameters; Step 3: Obtain the weather parameters of the city where the building is located, and quantify them to obtain a weather parameter dataset; Step 4: Build a detection model and train it based on the dataset of building design parameters and weather parameters. Evaluate the training of the detection model based on the labeled data obtained in step 1 to obtain a trained detection model. Step 5: Obtain the architectural design parameters of the building to be inspected and the weather parameters of the city where it is located, quantify them, obtain the predicted energy consumption value through the detection model, and determine whether its energy consumption is abnormal.

2. The data-driven block-scale building high energy consumption anomaly detection method according to claim 1 is characterized in that: In step 1-2, the standardized calculation formula for energy consumption intensity is as follows: in, is the normalized energy intensity per unit area of ​​the i-th building in the j-th month, EUI i，j is the actual energy consumption intensity per unit area of ​​the i-th building in the j-th month, is the energy consumption intensity per unit area of ​​the building per month during the period of minimum value obtained for the i-th building, It is the maximum value of the energy consumption intensity per unit area of ​​the i-th building in each month during the acquired time period.

3. The data-driven block-scale building high energy consumption anomaly detection method according to claim 1 is characterized in that: In steps 1-3, the calculation formula for each building threshold is: in, is the energy consumption threshold of the i-th building, n is the total number of months during the study period, EUI i，j is the actual monthly energy consumption intensity per unit area of ​​the i-th building in the j-th month.

4. The data-driven block-scale building high energy consumption anomaly detection method according to claim 1 is characterized in that: In steps 1-4, if the energy consumption intensity per unit area of ​​the building per month is greater than the set threshold, the corresponding data point is marked as 1; otherwise, it is marked as 0.

5. The data-driven block-scale building high energy consumption anomaly detection method according to claim 1 is characterized in that: In step 2, the building design parameters include basic building form parameters, building interaction parameters and building function parameters; The basic building morphology parameters also include: the total floor area of ​​the target building, the total number of floors of the target building, the height of the target building, the width of the target building, the length of the target building, the perimeter of the target building, the surface area of ​​the target building in the dataset, the volume of the target building, and the orientation of the target building. The parameters of mutual influence between buildings also include: building attribute parameters and parameters of the influence of the surrounding environment on the building; the building attribute parameters also include the body coefficient of the target building, the ratio of the perimeter area of ​​the target building, and the ratio of the length and width of the target building, where the body coefficient refers to the ratio of the volume to the surface area of ​​the building; the parameters of the influence of the surrounding environment on the building also include the ratio of the south obstacle height to the street canyon width, the ratio of the west obstacle height to the street canyon width, the ratio of the north obstacle height to the street canyon width, and the ratio of the east obstacle height to the street canyon width; Building function parameters also include: public services, education, industry, hospitals and hotels. The building functions of each building are quantified to obtain the building function data of each building.

6. The data-driven block-scale building high energy consumption anomaly detection method according to claim 5 is characterized in that: The quantification of the building functions of each building also includes: If the building belongs to the public service category, the corresponding building function data is 0; if the building belongs to the education category, the corresponding building function data is 1; if the building belongs to the industrial category, the corresponding building function data is 2; if the building belongs to the hospital category, the corresponding building function data is 3; if the building belongs to the hotel category, the corresponding building function data is 4.

7. The data-driven block-scale building high energy consumption anomaly detection method according to claim 1 is characterized in that: The types of weather parameters obtained include temperature, feels-like temperature, humidity, air pressure, wind speed, wind direction, hourly rainfall, hourly snowfall, cloud cover, number of weather condition descriptions, number of hours with uncomfortable temperature, number of hours with uncomfortable humidity, and number of hours with uncomfortable temperature.

8. The data-driven block-scale building high energy consumption anomaly detection method according to claim 7 is characterized in that: The data value of the perceived temperature can be quantified using the following formula: Feels_like=-42.379+(2.04901523×T)+(10.14333127×rh)-(0.22475541×T×rh)-(6.83783×10 -3 ×T 2 )-(5.481717×10 -2 ×rh 2 )+(1.22874×10 -3 ×T 2 ×rh)+(8.5282×10 -4 ×T×rh 2 )-(1.99×10 -6 ×T 2 ×rh 2 ), Among them, Feels_like is the feeling temperature, that is, the heat index, T is the air temperature, and rh is the relative humidity. The air temperature and relative humidity can be directly obtained through weather software or the city meteorological bureau.

9. The data-driven block-scale building high energy consumption anomaly detection method according to claim 7 or 8, characterized in that: The quantification of hours of uncomfortable temperature, hours of uncomfortable humidity and hours of feeling uncomfortable temperature also includes: temperatures outside the range of 20-27 degrees Celsius are considered uncomfortable, and humidity outside the range of 30-65% is considered uncomfortable.

10. The data-driven block-scale building high energy consumption anomaly detection method according to claim 1, characterized in that: The step 4 further comprises: Step 4-1: Combine the building design parameter dataset obtained in step 2 and the weather parameter dataset obtained in step 3 to construct a dataset for detection model training, and divide it into a training set and a test set; Step 4-2: Build a detection model based on the support vector machine algorithm, and train and evaluate the detection model based on the training set, test set, and the building marker data obtained in step 1; In step 4-3, the training of the detection model is determined to be complete based on the evaluation results, and the trained detection model is output.

11. The data-driven block-scale building high energy consumption anomaly detection method according to claim 10, characterized in that: The step 4-2 further includes: calculating the accuracy of the prediction result of the detection model based on the test set as an evaluation indicator to evaluate the detection model, including: The detection model is tested using test set data. Based on the building energy consumption prediction value output by the detection model, the judgment method in step 1 is used to determine whether the energy consumption of each building is abnormal. Combined with the labeled data of each building's energy consumption in step 1, the labeled data is compared with the results judged based on the output value of the detection model. The prediction accuracy of the detection model for the energy consumption of each building is calculated, and the performance of the detection model on the test set is evaluated.

12. A data-driven block-scale building high energy consumption anomaly detection system using the data-driven block-scale building high energy consumption anomaly detection method according to any one of claims 1 to 11, characterized in that: include: Data acquisition module, parameter processing module, training module and detection module; Among them, the data acquisition module is used to collect energy consumption data, urban morphology data and meteorological data of the detected area; The parameter processing module can process the data acquired by the data acquisition module to obtain the training set and test set required for training, as well as the data that can be directly input into the detection module in actual detection; The training module can train the detection model constructed in the detection module based on the training set and the test set; The detection module can detect abnormalities in building energy consumption in the detected area based on the trained detection model and obtain detection results.

13. A terminal comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method according to any one of claims 1 to 11.

14. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 11 are implemented.

Images