Building carbon emission prediction method and system based on optimization learning

Abstract

The invention discloses a building carbon emission prediction method and system based on optimization learning, and the method comprises the steps: recognizing first influence factors of a first object, and sorting the first influence factors; and performing first correction processing on the first influence factor. And establishing a first neural network model, and inputting the first influence factor for training and testing. According to the building carbon emission prediction method and system based on optimization learning provided by the invention, a building full-life-cycle carbon emission prediction model based on time series data is constructed based on external environment factors such as building personnel characteristics, building characteristics, electric equipment characteristics and air temperature characteristics, so that accurate prediction of carbon emission in a building full-life cycle is realized; and a foundation is laid for energy conservation and emission reduction of buildings.

Description

Technical Field

[0001] The present invention relates to the technical field, and in particular to a method and system for predicting building carbon emissions based on optimization learning. Background Art

[0002] In 2020, the total carbon emissions from the entire process of buildings in my country reached 5.08 billion tons of CO2, accounting for 50.9% of the national carbon emissions. The carbon emissions during the building operation phase were 2.16 billion tons of CO2, accounting for 21.7% of the total national carbon emissions. As a result, building energy conservation has become one of the key response strategies for my country to break through the bottleneck of energy development. It is also the current strategic policy for my country to reduce energy burden, ease energy pressure, protect the environment, and achieve sustainable development. The accurate prediction of building carbon emissions is the basis for building energy conservation and emission reduction.

[0003] Foreign countries have carried out research on carbon emissions from buildings relatively early. Among them, Karatasou used a relatively advanced artificial neural network method to conduct deep learning of energy consumption data, and then built a building energy consumption prediction model, and finally achieved the improvement of prediction ability; A. Azadeh used a random program to combine genetic engineering and artificial neural networks to predict electricity consumption; Tugce Kazanasmaz (2009) and others tested the lighting energy consumption data of an office building within three months, and used the circumvallate network method to analyze and predict it.

[0004] Early domestic scholars mostly used regression analysis for prediction, but with the development of emerging technologies, some scholars introduced deep learning methods such as neural networks to predict building energy consumption and carbon emissions. In addition, there are research methods that combine multiple prediction methods to build models, all of which have achieved predictions within a certain range of errors.

[0005] In recent years, deep learning methods based on recurrent neural networks have been widely used in many fields due to their strong nonlinearity and high precision. In particular, project operation data has significant time-space characteristics, and has periodic and lag characteristics in time. Based on the above research, a time series prediction model based on periodic neural networks and reverse diffusion feedforward networks is established to fully explore the time series characteristics of building operation status and improve the accuracy of building energy consumption and carbon emission forecasts. Summary of the invention

[0006] In view of the above-mentioned problems, the present invention is proposed.

[0007] Therefore, the technical problem solved by the present invention is that the existing research is mostly focused on the reconstruction of buildings and technical energy saving, but less on energy saving in management. From the perspective of research level, current research is mainly concentrated on major industries such as the national, provincial, municipal, county, and city levels, while there are relatively few studies on carbon emissions at the building level. At present, there is no unified measurement system for the measurement of building carbon emissions at home and abroad. The traditional building carbon emission estimation and prediction method based on the multivariate regression model, because it uses linear regression, cannot be completely close to the actual data, so the accuracy is usually not high, and there are also some limitations in practice, and it is also necessary to try to dig out various influencing factors for prediction. In actual operation, it is more difficult to implement.

[0008] To solve the above technical problems, the present invention provides the following technical solutions: a method for predicting building carbon emissions based on optimization learning, comprising: identifying a first influencing factor of a first object, and ranking the first influencing factor.

[0009] A first correction process is performed on the first influencing factor.

[0010] A first neural network model is established, and the first influencing factors are input for training and testing.

[0011] As a preferred solution of the building carbon emission prediction method based on optimization learning described in the present invention, wherein: the identification of the first influencing factor of the first object includes using a literature survey method to sort out existing research results to obtain the first influencing factor.

[0012] As a preferred solution of the building carbon emission prediction method based on optimization learning described in the present invention, wherein: the sorting of the first influencing factors includes identifying and screening the first influencing factors to determine the priority of the first objects.

[0013] As a preferred solution of the building carbon emission prediction method based on optimization learning described in the present invention, the first correction processing includes missing value and outlier processing and normalization processing.

[0014] As a preferred solution of the building carbon emission prediction method based on optimization learning described in the present invention, the missing value and outlier processing includes filling and deleting the missing data. The normalization processing includes standardizing the original data with the standard deviation and the mean value.

[0015] As a preferred solution of the building carbon emission prediction method based on optimization learning described in the present invention, the first object includes but is not limited to the carbon emission of the building, and the first influencing factors include personnel characteristics, building characteristics, electrical equipment characteristics and temperature characteristics.

[0016] As a preferred solution of the building carbon emission prediction method based on optimization learning described in the present invention, wherein: the establishment of the first neural network model includes establishing an RNN time series model, classifying the first influencing factors at the input layer, and dividing them into a training set and a test set according to a proportion. The training set is normalized and a standardized training set is output.

[0017] After completing the training of the recurrent neural network, prediction is performed. The method of selection algebra is used in the prediction process to obtain the prediction results.

[0018] A building carbon emission prediction system based on optimization learning, characterized by: including:

[0019] The sorting module identifies a first influencing factor of the first object and sorts the first influencing factor.

[0020] A correction module performs a first correction process on the first influencing factor.

[0021] The prediction module establishes a first neural network model and inputs the first influencing factors for training and testing.

[0022] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0023] A computer-readable storage medium stores a computer program, which implements the steps of the method described above when executed by a processor.

[0024] The beneficial effects of the present invention are as follows: based on external environmental factors such as building personnel characteristics, building characteristics, electrical equipment characteristics, air temperature characteristics, etc., a building life cycle carbon emission prediction model based on time series data is constructed to achieve accurate prediction of carbon emissions throughout the building life cycle, laying the foundation for achieving building energy conservation and emission reduction. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0026] Figure 1 An overall flow chart of a building carbon emission prediction method based on optimization learning provided in the first embodiment of the present invention.

[0027] Figure 2A comparison chart of predicted values ​​and actual values ​​of building 1 according to a building carbon emission prediction method based on optimization learning provided in the second embodiment of the present invention.

[0028] Figure 3 A comparison chart of predicted values ​​and actual values ​​of Building 2 according to a building carbon emission prediction method based on optimization learning provided in the second embodiment of the present invention.

[0029] Figure 4 A comparison chart of predicted values ​​and actual values ​​of building 3 according to a building carbon emission prediction method based on optimization learning provided in the second embodiment of the present invention. DETAILED DESCRIPTION

[0030] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0031] Example 1, reference Figure 1 , which is an embodiment of the present invention, provides a building carbon emission prediction method based on optimization learning, comprising:

[0032] S1: Identify first influencing factors of a first object, and rank the first influencing factors.

[0033] In the present invention, the first object is the carbon emission of the building, and the first influencing factors are personnel characteristics, building characteristics, electrical equipment characteristics, and temperature characteristics.

[0034] It should be noted that by identifying and screening the factors affecting building carbon emissions, we can quickly grasp and determine the priority of building carbon emissions. Therefore, this paper adopts the method of literature survey, sorts out the existing research results, and obtains the following four factors that have a significant impact on building carbon emissions, which are ranked in order of importance:

[0035] Personnel characteristics: Personnel characteristics are an important factor affecting carbon emissions from public buildings. Existing studies have shown that population density is an important factor affecting building compactness, and in low-density areas, its operating energy consumption and carbon emissions remain at a certain level, resulting in high per capita energy consumption and carbon emissions, which in turn exacerbates the adverse effects of low-carbon development in buildings. With the continuous expansion of my country's construction scale, my country's building energy consumption has increased year by year. It can be inferred that during my country's construction process, the number of residents will have a certain impact on total energy consumption and total carbon emissions.

[0036] Building characteristics: Generally, the description of building volume can be divided into two aspects: area and height. The height of a building is the elevation from the roof to the outdoor floor. Building height refers to the height difference between two adjacent floors of a building. Generally, as the number of floors increases, the carbon emissions per unit building area should increase. When the number of floors is small, the growth rate is large, and then gradually slows down until the height reaches a certain limit and remains unchanged. In addition, there are differences in carbon emissions between the above-ground area and the underground area of ​​a building.

[0037] Characteristics of electrical equipment: Building environmental systems mainly include air conditioning, lighting, computers, elevators and other energy-consuming systems that affect building energy consumption and carbon emissions. Through energy-saving renovations of most buildings, it is found that energy conservation and emission reduction can be achieved to a certain extent by using new energy sources, energy-saving lamps and other energy-saving means. Therefore, electrical equipment in buildings is particularly important.

[0038] Temperature characteristics: Climate conditions are an important factor affecting building energy consumption and carbon emissions. Meteorological factors such as sunshine, rainfall, temperature, and wind direction will directly affect the building's insulation, heat preservation, dehumidification and other needs, while temperature is the main factor affecting building energy conservation and emission reduction. For example, in cold areas in the north, heating is required in winter. In hot places, air conditioning is required in summer. On the same land, with the same number of people and the same living standards, as the temperature around the building rises, the building's energy consumption and carbon dioxide emissions will also increase.

[0039] The first object may also be carbon emissions from transportation vehicles and carbon emissions from industrial production, and the corresponding first influencing factors are also different from the influencing factors of carbon emissions from buildings.

[0040] In an optional embodiment of the present invention, the first object is the carbon emission of a vehicle. The carbon emission of a vehicle mainly comes from fuel consumption, and there are many factors affecting its carbon emission. The corresponding influencing factors include:

[0041] Number of passengers: The carbon emissions of a vehicle are directly related to the number of passengers it carries. When the number of passengers increases, the carbon emissions per unit of transportation of the vehicle decreases (i.e., the carbon emissions per person per kilometer decreases), and vice versa. Therefore, knowing the average number of passengers per trip can effectively predict the carbon emissions of a vehicle.

[0042] Travel mode: Different travel modes (such as public transportation, private cars, electric vehicles, etc.) have different impacts on carbon emissions. Electric vehicles reduce carbon emissions because they use electricity, while fuel vehicles rely on fossil fuels and have higher carbon emissions.

[0043] Vehicle type: Different types of vehicles (such as cars, buses, trucks, motorcycles, etc.) have different impacts on carbon emissions. For example, cars have low carbon emissions, but large trucks have more significant carbon emissions.

[0044] Fuel type: The type of fuel used by a vehicle (such as gasoline, diesel, electricity, natural gas, etc.) directly determines its carbon emission level. For example, electric vehicles have lower carbon emissions, while traditional fuel vehicles emit higher carbon dioxide.

[0045] Emission standards: Emission standards for modern vehicles (such as Euro 6, National V, etc.) have a significant impact on carbon emissions. Stricter emission standards require vehicles to emit fewer pollutants, thereby reducing carbon emissions.

[0046] In an optional embodiment of the present invention, the first object is carbon emissions from industrial production, which is one of the main sources of carbon emissions, especially heavy industry and manufacturing. Its carbon emissions are closely related to factors such as energy used in the production process, production technology, and equipment efficiency. The corresponding influencing factors include:

[0047] Production scale: The larger the production scale, the higher the energy consumption and carbon emissions. Large-scale production requires more raw materials and energy consumption, and also generates more waste and pollution. On the contrary, small-scale production can better control energy use in some cases, thereby reducing carbon emissions.

[0048] Production process: Different production processes have a huge impact on carbon emissions. For example, the carbon emissions of traditional coal-fired power generation processes are very high, while production processes that use clean energy (such as solar energy, wind energy, etc.) significantly reduce carbon emissions.

[0049] Product type: Different products have different carbon emissions from their production process. For example, the production process of high-energy-consuming industrial products such as steel, fertilizers, and chemicals will generate a lot of carbon emissions, while the production process of light consumer goods may be relatively low-carbon.

[0050] Energy type: The type of energy used in industrial production directly affects carbon emissions. The use of fossil fuels such as coal, natural gas and oil will produce higher carbon emissions, while renewable energy (such as wind power and solar energy) can significantly reduce carbon emissions.

[0051] Energy efficiency: The more efficient the energy use, the lower the carbon emissions per unit product. Modern equipment and intelligent management can improve energy efficiency and thus reduce carbon emissions.

[0052] Energy supply: The way energy is supplied (such as central heating, distributed energy, etc.) will also affect carbon emissions. If a central heating system is used, it may be necessary to transport fuel, which itself will bring additional carbon emissions.

[0053] S2: Perform a first correction process on the first influencing factor.

[0054] The treatment of missing values ​​and outliers includes, for missing data, including but not limited to the use of two methods to deal with completion and deletion. For example, other statistics such as mean, mode, median, etc. can be used to fill in, or the average value before and after can be calculated by interpolation method. In addition, it is also possible to determine whether to delete all samples or completely delete all features based on the number of missing values. In addition, for missing data, the "type" method can also be used to fill in. Taking the electricity carbon emission data during the construction and operation period as an example, each sample is an independent time series data. Starting from extracting the valid data interval of each sampling point, the first number of non-zero values ​​in the sequence is used as the starting point, and the number of the last non-zero value is used as the end. Secondly, at each column sampling point, the ratio of its failure value to the maximum sample length should be less than a certain threshold, so as to ensure the security of the data. First, the Excel method is used to statistically analyze the basic data of carbon emissions of the power system in the operation of my country's buildings, and the electricity consumption of the same period of other years is used to replace some of the missing or abnormal data caused by statistical data (more than one order of magnitude change compared with last year).

[0055] Data normalization includes, normalization is to standardize a single sample and convert it into a single vector. In view of the current situation that the measurement units of my country's building carbon emission time series data and its influencing factors are not unified and chaotic, it is necessary to standardize them to achieve the purpose of standardization. The existing data normalization methods mainly include Min-Max normalization and Z-score normalization. Maximum value normalization is to convert the original data into the range of [0,1]. Time series data is non-observable. In the prediction process, the maximum or minimum value after adding data does not match the original data, so that the normalized prediction result fluctuates within [0,1]. Therefore, in the present invention, the Z-score normalization method is mainly used, that is, the standard deviation and mean are used to standardize the original data. The following is the specific calculation formula:

[0056]

[0057] Among them, μ represents the sample size mean of the data, and S represents the sample size standard deviation of the data. Since the research is about time series data, there is the possibility of predicting multiple time series at the same time in the time series data. Therefore, for each group of samples, a logarithmic transformation is first performed, and then the mean and variance of each sampling sequence are calculated. In addition, this method can also normalize the test cases according to the number of samples and the number of variances obtained in the training set.

[0058] S3: Establish a first neural network model, and input the first influencing factors for training and testing.

[0059] The recurrent neural network time series model divides data processing into two stages: training and testing. Training is concentrated on the hidden layer. First, at the input level, this project intends to rank the carbon emissions of electricity during the original building operation as F n ={f1,f2,…f n} and 17 factors with important effects {X1,X2…X 17}, then, divide them into two groups according to 80% and 20%, that is, F tr ={f1,f2,…f m}, F te ={f m +1,f m +2,…f n Then, the training set is normalized using the traditional Z-score (mean 0 and standard deviation 1), and the standardized training set can be expressed as:

[0060]

[0061] Among them, Fir′ represents the training data set F tr The data after standardization. fi′ represents the i-th sample data after standardization. i represents the i-th sample data in the original data set. n represents the data set F tr The total number of samples in .

[0062] Model input after data segmentation:

[0063] X={X1,X2,...X t}

[0064] X p ={f p ',f p+1 ',...f m-L+p-1 '}

[0065] Where X represents the input data set of the neural network. p represents a data segment with a time step length of L selected from the training data set, including p+1 ′ to f m-L+p ′ is the normalized data. p represents the starting position of the current data segment. m represents the total number of samples of training data. L represents the time step used for training.

[0066] The corresponding theoretical output is:

[0067] Y={Y1,Y2,...Y t}

[0068] Y P ={f p+1 ',f p+2 ',...f m-L+p '}

[0069] Where Y represents the theoretical output data set of the neural network. Yp represents the output data set from f p+1 ′ to f m-L+p ′’s standardized data as the output of the neural network.

[0070] After completing the training of the recurrent neural network, it is predicted and the method of selection algebra is selected in the prediction process.

[0071] Y f ={f m-L+1 ',f m-L+2 ',...f m '}

[0072] Among them, Y f Represents the predicted output data sequence. m-Z+1 ′ represents the first value predicted from the normalized training data. Z represents the length of the predicted data.

[0073] For the neural network model, the output can be expressed as:

[0074]

[0075] Among them, P f Represents the prediction result of the recurrent neural network output. LSTM m+1 * Indicates that the LSTM model performs Y at the m+1th moment f The predicted result. m-L +2,P m-L+3 ,…,P m+1 Represents the prediction results from step m-L+2 to step m+1.

[0076] At time m+1, the predicted value is P m+1 , then compare the L-1 most recent data points of Y with P m+1 Merge into:

[0077] P m+1 = {P m+2 ,P m+3 ,...P m}

[0078] Among them, P m+1 Represents the dataset after merging the predicted values.

[0079] In the recurrent neural network model, Ym+1 Input it, and the predicted value at time m+2 is P m+2 , and so on, the predicted sequence data is:

[0080] P m = {P m+1 ,P m+2 ,...P m}

[0081] Among them, P m represents the complete forecast sequence.

[0082] Next, for P m Perform Zcore denormalization to obtain the final prediction sequence for the corresponding test set F:

[0083]

[0084] Among them, de_score(P m ) indicates that the predicted value P m Denormalize to restore it to the scale of the original data. i ′ represents the standardized i-th predicted value. i represents the i-th predicted value after denormalization. i Represents the i-th sample in the original data. n represents the total number of samples in the data set.

[0085] The carbon emission time series data in the actual power generation process is divided into different time periods according to the ratio of 80% and 20%, and training samples and test samples are constructed respectively. The model is trained using the Bayesian regularization method provided by MATLAB.

[0086] The computer device may be a server. The computer device includes a processor, a memory, an input / output interface (I / O for short) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data cluster data of a power monitoring system. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a building carbon emission prediction method based on optimization learning is implemented.

[0087] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0088] Example 2, reference Figures 2 to 4 , which is an embodiment of the present invention, provides a building carbon emission prediction method and system based on optimization learning. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through simulation experiments.

[0089] Based on the carbon emission measurement data of three buildings in an office park, the method proposed by the present invention is used for prediction. The results are as follows: Figure 2 , 3 , 4. The black dotted line represents the predicted value, and the red solid line represents the actual value.

[0090] After 1000 trainings, the RNN model was applied to the carbon emission data of the four teaching buildings, and the mean RMSE (RMSE) was obtained. The performance of RMSE in the test set was 3.11e+03, which is the average prediction error of the model of about 3000kg, which can more accurately analyze the carbon emission level. Most of the carbon emissions in the power industry are around 40.78kg. In general, the prediction results of the recurrent neural network have some deviations at the sharp corners of the actual data, but in general, it can better reflect the changing trend of the entire carbon emission.

[0091] Embodiment 3 is an embodiment of the present invention, including a building carbon emission prediction system based on optimization learning, specifically:

[0092] The sorting module identifies a first influencing factor of the first object and sorts the first influencing factor.

[0093] A correction module performs a first correction process on the first influencing factor.

[0094] The prediction module establishes a first neural network model and inputs the first influencing factors for training and testing.

[0095] 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 preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should be included in the scope of the claims of the present invention.

Claims

1. A building carbon emission prediction method based on optimization learning, characterized in that: include: Identify a first influencing factor of the first object, and rank the first influencing factors; performing a first correction process on the first influencing factor; A first neural network model is established, and the first influencing factors are input for training and testing.

2. The method for predicting building carbon emissions based on optimization learning according to claim 1, characterized in that: The identifying the first influencing factor of the first object includes using a literature survey method to sort out existing research results to obtain the first influencing factor.

3. The building carbon emission prediction method based on optimization learning as claimed in claim 2, characterized in that: The sorting of the first influencing factors includes identifying and screening the first influencing factors to determine the priority of the first objects.

4. The method for predicting building carbon emissions based on optimization learning as claimed in claim 3, characterized in that: The first correction process includes missing value and outlier processing and normalization processing.

5. The method for predicting building carbon emissions based on optimization learning as claimed in claim 4, characterized in that: The missing value and outlier processing includes filling in and deleting the missing data; the normalization processing includes standardizing the original data with the standard deviation and mean value.

6. The method for predicting building carbon emissions based on optimization learning as claimed in claim 5, characterized in that: The first object includes but is not limited to carbon emissions of buildings, and the first influencing factors include personnel characteristics, building characteristics, electrical equipment characteristics and temperature characteristics.

7. The method for predicting building carbon emissions based on optimization learning according to claim 6, characterized in that: The establishing of the first neural network model includes establishing an RNN time series model, classifying the first influencing factors at the input layer and dividing them into a training set and a test set according to a ratio; normalizing the training set and outputting a standardized training set; After completing the training of the recurrent neural network, prediction is performed. The method of selection algebra is used in the prediction process to obtain the prediction results.

8. A building carbon emission prediction system based on optimization learning using the method according to any one of claims 1 to 7, characterized in that: A sorting module, identifying a first influencing factor of the first object and sorting the first influencing factor; A correction module, performing a first correction process on the first influencing factor; The prediction module establishes a first neural network model and inputs the first influencing factors for training and testing.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images