A method, device, equipment and storage medium for predicting building power load

Abstract

The present invention discloses a method, device, equipment and storage medium for predicting building power load. The method comprises: obtaining historical building power load values ​​and additional characteristic data of target building objects at each unit time in a historical time period; comparing the historical building power load value of each unit time with a preset threshold value, and determining a working characteristic value for characterizing the work intensity type at each unit time based on the comparison result; using the additional characteristic data and working characteristic value of each unit time in the historical time period as model input, and using the historical building power load value of each unit time in the historical time period as model output, training the preset model, and obtaining a prediction model for predicting building power load; the present invention not only has high practicality, but also can more efficiently and accurately predict building power load.

Description

Technical Field

[0001] The present invention relates to the technical field of power load forecasting, and in particular to a method, device, equipment and storage medium for forecasting building power load. Background Art

[0002] As a safe, efficient and clean energy, electric energy plays a vital role in the country's current energy substitution and environmental protection. Since electric energy cannot be stored on a large scale, the dispatching department needs to change the power generation in real time according to the dynamic demand changes on the load side. Otherwise, the quality of electric energy may deteriorate, thereby undermining the safe and stable operation of the power system. In order to effectively avoid the above problems, it is necessary to predict the power load. In the field of construction, the prediction of building power demand with high accuracy is of great significance for the optimal control of building systems and the reduction of building operation energy consumption. Small improvements in the prediction method can produce huge economic benefits. Although there are many building energy consumption simulation software now, the existing building energy consumption simulation software only relies on single data for power load prediction, which makes it difficult to portray the complex electricity consumption behavior of buildings, and the prediction accuracy needs to be improved. Therefore, how to predict the building power load more efficiently and accurately has become a technical problem that technicians in this field need to solve urgently. Summary of the invention

[0003] The purpose of the present invention is to provide a method, device, equipment and storage medium for predicting building power load, which are not only highly practical, but also can predict building power load more efficiently and accurately.

[0004] According to one aspect of the present invention, a method for predicting building power load is provided, the method comprising:

[0005] Obtain the historical building power load value and additional characteristic data of the target building object at each unit time in the historical time period;

[0006] Compare the historical building power load value of each unit time with the preset threshold value, and determine the work characteristic value for characterizing the work intensity type of each unit time based on the comparison result;

[0007] The additional characteristic data and working characteristic values ​​of each unit time in the historical time period are used as model inputs, and the historical building power load values ​​of each unit time in the historical time period are used as model outputs. The preset model is trained to obtain a prediction model for predicting building power load.

[0008] Optionally, the additional characteristic data includes at least one of the following: weather forecast data, holiday schedule data, and air conditioning cooling and heating loads.

[0009] Optionally, when the additional characteristic data includes air conditioning cooling and heating loads, the method further includes:

[0010] Acquire the building characteristic data of the target building object at each unit time in the historical time period; wherein the building characteristic data includes at least one of the following: building structure data, enclosure structure material data, heating and air conditioning mode and system distribution data, indoor personnel activity law data and building lighting condition data;

[0011] The building characteristic data of each unit time in the historical time period and the preset indoor temperature value are input into the energy consumption calculation platform, so as to calculate the air conditioning cooling and heating load of the target building object in each unit time in the historical time period through the energy consumption calculation platform.

[0012] Optionally, the preset threshold is determined in the following manner:

[0013] The K-means clustering algorithm is used to cluster the power load values ​​of the historical buildings at each unit time in the historical time period into two subclasses;

[0014] Determine the size value of the center point of each subclass respectively, and determine the first critical point and the second critical point based on the size values ​​of the two center points; wherein the first critical point is smaller than the second critical point;

[0015] The first critical point and the second critical point are set as the preset threshold value.

[0016] Optionally, the comparing the historical building power load value of each unit time with a preset threshold value, and determining the work characteristic value for characterizing the work intensity type of each unit time based on the comparison result, includes:

[0017] Traverse the power load values ​​of historical buildings at each unit time in turn. If the power load value of the historical building currently traversed is less than the first critical point, set the first characteristic value for characterizing weak working intensity for the currently traversed unit time. If the power load value of the historical building currently traversed is greater than the first critical point and less than the second critical point, set the second characteristic value for characterizing medium working intensity for the currently traversed unit time. If the power load value of the historical building currently traversed is greater than the second critical point, set the third characteristic value for characterizing high working intensity for the currently traversed unit time.

[0018] The first eigenvalue is smaller than the second eigenvalue, and the second eigenvalue is smaller than the third eigenvalue.

[0019] Optionally, the preset model is a TCN-GRU model.

[0020] Optionally, after the preset model is trained to obtain a prediction model for predicting building power load, the method further includes:

[0021] The additional characteristic data and the working characteristic value of the target building object at the time to be predicted are input into the prediction model, and the output of the prediction model is used as the predicted building power load value.

[0022] Optionally, when the additional feature data includes weather forecast data, inputting the additional feature data and the working feature value of the target building object at the time to be predicted into the prediction model, and using the output of the prediction model as the predicted building power load value, comprises:

[0023] Based on the weather forecast data for the time to be predicted, a weather range value is formed based on a preset weather forecast error range;

[0024] Selecting a plurality of Monte Carlo sample values ​​within the weather range using a Monte Carlo algorithm;

[0025] Substituting each Monte Carlo sample value for the weather forecast data in the additional feature data in turn and inputting the Monte Carlo sample value into the prediction model together with the working feature value at the time to be predicted, so as to obtain the corresponding predicted building power load value;

[0026] A predicted building power load interval is formed based on all the predicted building power load values ​​obtained.

[0027] In order to achieve the above object, the present invention also provides a device for predicting building power load, the device comprising:

[0028] An acquisition module, used to acquire the historical building power load value and additional characteristic data of the target building object at each unit time in the historical time period;

[0029] A determination module, used to compare the historical building power load value of each unit time with a preset threshold value, and determine a work characteristic value for characterizing the work intensity type of each unit time based on the comparison result;

[0030] The training module is used to take the additional characteristic data and working characteristic values ​​of each unit time in the historical time period as model input, and take the historical building power load value of each unit time in the historical time period as model output, train the preset model, and obtain a prediction model for predicting the building power load.

[0031] In order to achieve the above-mentioned purpose, the present invention also provides a computer device, which specifically includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the method for predicting building power load introduced above when executing the computer program.

[0032] In order to achieve the above object, the present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the method for predicting building power load introduced above are implemented.

[0033] The method, device, equipment and storage medium for predicting building power load provided by the present invention comprehensively consider the influence of employee work intensity, weather factors, building air-conditioning equipment, etc. on the building power load, and design a building power load prediction method based on the combination of physical calculation software and data drive, which is not only highly practical, but also can predict the building power load more efficiently and accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0035] Figure 1 An optional flow chart of the method for predicting building power load provided in Embodiment 1;

[0036] Figure 2 Another optional flow chart of the method for predicting building power load provided in the first embodiment;

[0037] Figure 3 A schematic diagram of an optional composition structure of the device for predicting building power load provided in the second embodiment;

[0038] Figure 4 A schematic diagram of an optional hardware architecture of a computer device provided in Example 3. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical scheme and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0040] Embodiment 1

[0041] The embodiment of the present invention provides a method for predicting building power load, such as Figure 1 As shown, the method specifically comprises the following steps:

[0042] Step S101: Obtain the historical building power load value and additional characteristic data of the target building object at each unit time in the historical time period.

[0043] The target building object may be an entire building, such as an office building, a teaching building, a hospital building, or a shopping mall, or may be one or more floors of a building, such as the ground floor area of ​​an office building or all the dining floors of a shopping mall. In addition, the unit time is the minimum time period for counting the building power load, such as the unit time is a day or a unit time is an hour. Preferably, the unit time is an hour.

[0044] The additional characteristic data is parameter data associated with the building power load and used to assist in predicting the building power load; preferably, the additional characteristic data includes at least one of the following: weather forecast data, holiday schedule data, and air conditioning heating and cooling load. Among them, the weather forecast data is the weather data of each unit time in the historical time period obtained from the preset weather forecast platform, for example, the weather forecast data includes: temperature data, humidity data. The holiday schedule data is used to describe the holiday schedule in the historical time period, for example, the holiday schedule data includes: working days, weekends, holidays. The air conditioning heating and cooling load is used to describe the energy consumption of the air conditioning equipment of the target building object in each unit time in the historical time period, and the air conditioning heating and cooling load includes: air conditioning heat load and air conditioning cooling load. Of course, in practical applications, the additional characteristic data may also include parameter data of other dimensions for assisting in predicting the building power load, which is not specifically limited here.

[0045] Step S102: Compare the historical building power load value of each unit time with the preset threshold value, and determine the work characteristic value of each unit time for characterizing the work intensity type based on the comparison result.

[0046] Among them, the power load value of the historical building is proportional to the working characteristic value, that is, the power load value of the historical building is proportional to the working intensity; if the power load value of the historical building is larger, it means that the working intensity is greater, and then the working characteristic value is larger; if the power load value of the historical building is smaller, it means that the working intensity is smaller, and then the working characteristic value is smaller.

[0047] The work characteristic value is used to characterize the work intensity of the staff in the target building object within a unit time; the work intensity types can be divided into N categories, and each category corresponds to a work characteristic value, that is, there are N work characteristic values. At this time, it is necessary to set (N-1) preset thresholds in advance to determine the work characteristic value of each unit time based on the (N-1) preset thresholds.

[0048] Preferably, the work intensity types include: weak, medium, and strong, and the work intensity values ​​include: a first eigenvalue characterizing weak work intensity, a second eigenvalue characterizing medium work intensity, and a third eigenvalue characterizing strong work intensity, and the first eigenvalue is smaller than the second eigenvalue, and the second eigenvalue is smaller than the third eigenvalue.

[0049] Since it is impossible to obtain the employee work intensity data in the historical time, the work characteristic value used to characterize the work intensity is estimated in this embodiment through the historical building power load value. Specifically, the preset thresholds for classifying the work intensity are set in advance. For example, when the work intensity type includes three characteristic values, two corresponding preset thresholds are set in advance, so that the corresponding work characteristic value is determined based on the building power load value through the set preset thresholds.

[0050] Step S103: Using the additional characteristic data and working characteristic values ​​of each unit time in the historical time period as model input, and using the historical building power load values ​​of each unit time in the historical time period as model output, training the preset model to obtain a prediction model for predicting building power load.

[0051] Since the daily work intensity of staff members fluctuates and they may work overtime at irregular intervals, this brings great uncertainty to the power load forecast. If a fixed employee time sheet is directly used, it cannot truly reflect the working status of employees. In this embodiment, in order to improve the prediction accuracy, the work characteristic value is first determined based on the historical building power load value, and then the preset model is trained based on the work characteristic value and additional feature data to obtain a prediction model with higher prediction accuracy. Since the work intensity of employees will directly affect the power consumption of the building, the greater the work intensity, the greater the required power load. Therefore, by using the work characteristic value that characterizes the work intensity to predict the building power load, the prediction accuracy can be improved.

[0052] Specifically, when the additional characteristic data includes air conditioning cooling and heating loads, the method further includes:

[0053] Step A1: Acquire the building characteristic data of the target building object at each unit time in the historical time period; wherein the building characteristic data includes at least one of the following: building structure data, enclosure material data, heating and air conditioning mode and system distribution data, indoor personnel activity law data and building lighting condition data;

[0054] Step A2: input the building characteristic data of each unit time in the historical time period and the preset indoor temperature value into the energy consumption calculation platform, so as to calculate the air conditioning cooling and heating load of the target building object at each unit time in the historical time period through the energy consumption calculation platform.

[0055] In this embodiment, the air conditioning cooling and heating load can be calculated in the manner of steps A1 to A2 above. In the case where the building feature data is incomplete, reference can be made to existing standards and specifications, such as: "General Specification for Energy Conservation and Renewable Energy Utilization in Buildings GB55015-2021", by using the data in the standards and specifications to complete the missing building feature data, and input the completed building feature data into the energy consumption calculation platform. Preferably, the energy consumption calculation platform uses EnergyPlus as the calculation core, performs year-round load calculation and energy consumption simulation analysis on the building and its air-conditioning system, so as to obtain hourly energy consumption data of different equipment. Since the energy consumption calculation platform is a prior art, the specific process of calculating the air conditioning cooling and heating load based on the building feature data and the preset indoor temperature value will not be repeated here.

[0056] Existing building power load prediction methods mainly focus on pure data-driven methods; however, pure data-driven methods lack interpretability in real physical scenarios because they cannot capture actual measurement characteristics such as building structure, heating and ventilation; therefore, in this embodiment, an energy consumption calculation platform (i.e., energy consumption simulation software) is used to model real physical scenarios and enhance the interpretability of physical scenarios for the prediction of building power load.

[0057] Specifically, the preset threshold is determined as follows:

[0058] Step B1: clustering the historical building power load values ​​of each unit time in the historical time period into two subclasses using a K-means clustering algorithm;

[0059] Among them, the K-means clustering algorithm is the most commonly used clustering algorithm based on Euclidean distance. It believes that the closer the distance between two targets, the greater the similarity. The specific steps of step B1 are as follows: first, set the number of clusters to 2 (that is, the K value is 2), and randomly assign K samples as initial cluster centers; second, assign all samples (that is, the power load values ​​of historical buildings) to the nearest cluster according to the distance, and recalculate the cluster center based on the assigned cluster samples; finally, reclassify all samples according to the new cluster center, and repeat the above steps until the error is minimized.

[0060] Step B2: Determine the size value of the center point of each subclass respectively, and determine the first critical point and the second critical point based on the size values ​​of the two center points; wherein the first critical point is smaller than the second critical point.

[0061] Preferably, when the value of the first center point is smaller than the value of the second center point, step B2 specifically includes:

[0062] The product of the value of the first center point and the first coefficient is set as the first critical point; wherein the first coefficient is a value greater than 1 and less than 2;

[0063] The product of the value of the second center point and the second coefficient is set as the second critical point; wherein the second coefficient is a value greater than 0 and less than 1.

[0064] In practical applications, the first coefficient may be set to 1.4, and the second coefficient may be set to 0.6. The specific values ​​of the first coefficient and the second coefficient may be modified as required and are not specifically limited here.

[0065] Step B3: setting the first critical point and the second critical point as the preset threshold value.

[0066] At this time, the step S102 specifically includes:

[0067] Traverse the power load values ​​of historical buildings at each unit time in turn. If the power load value of the historical building currently traversed is less than the first critical point, set the first characteristic value for characterizing weak working intensity for the currently traversed unit time. If the power load value of the historical building currently traversed is greater than the first critical point and less than the second critical point, set the second characteristic value for characterizing medium working intensity for the currently traversed unit time. If the power load value of the historical building currently traversed is greater than the second critical point, set the third characteristic value for characterizing high working intensity for the currently traversed unit time.

[0068] The first eigenvalue is smaller than the second eigenvalue, and the second eigenvalue is smaller than the third eigenvalue.

[0069] In practical applications, the first eigenvalue may be set to 0, the second eigenvalue may be set to 1, and the third eigenvalue may be set to 2. However, the first eigenvalue, the second eigenvalue and the third eigenvalue may be modified as required and are not specifically limited here.

[0070] In this embodiment, features can be extracted from the historical building power load data by using the K-means clustering algorithm, and the work intensity status of the employees can be determined. The preset threshold determined by the above steps B1 to B3 can flexibly determine the work intensity of the employees, rather than simply referring to the employee rest work table.

[0071] Specifically, the preset model is a TCN-GRU model, where TCN is a temporal convolutional network and GRU is a gated recurrent unit. Compared with the traditional TCN model, the TCN-GRU model has a newly added GRU block that further captures the temporal dependency between the outputs of the TCN residual blocks, which is conducive to capturing long-term dependencies in time series. Building power load forecasting is a time series forecasting task, which predicts future building power loads based on historical building power load record values ​​on different partitioned floors. The building power load forecasting task can be expressed by the following equation:

[0072] (X(i+1),…,X(i+t))=f(X(i-T+1),…,X(i));

[0073] The building power load forecasting model aims to learn a function f so that the information X (i -T+1……i) , predict the power load X in the next t time steps (i+1……i+t) .

[0074] Among them, the TCN model is a process of dilated causal convolution, which can capture local temporal features. Compared with other baseline models, it has stronger stability. The TCN-GRU model can capture longer temporal correlations brought by the expanded receptive field of the dilated convolution. The TCN structure allows for finer control of the input information, extracting features from the time series and capturing local patterns in the sequence. In addition, through the convolution operation, the TCN model can effectively capture short-range dependencies in the time series. By adding a dilation factor to the dilated convolution, TCN is able to capture information at different time scales simultaneously, thereby enhancing its ability to simulate long-term dependencies without increasing network parameters. The use of residual connections solves the problem of gradient disappearance during training. By stacking multiple layers, TCN can extract features at different levels to better capture complex patterns in the input sequence. The receptive field represents the range of the input sequence that can be captured by the convolution layer. By adjusting the dilation factor and kernel size, TCN can flexibly control the receptive field of each layer at different time scales.

[0075] The main steps of TCN include:

[0076] Convolution operation: y t =σ(W*x t + b);

[0077] Among them, x t is the input time series. During the model training phase, x tis a time series formed by the additional characteristic data and working characteristic values ​​of each unit time in the historical time period; t is the output prediction result. During the model training phase, y t is the historical building power load value of each unit time in the historical time period; σ is the dilated convolution operation; W is the convolution kernel (or filter); b is the bias term;

[0078] Dilated Convolution:

[0079] Among them, X,s,k,d,sd,i represent the past time series input, sequence element, filter size, dilation factor and direction respectively;

[0080] Residual connection: y(t)=F(x(t))+x(t);

[0081] Among them, x(t) is the input, also known as "skip connection" or "identity mapping"; F is the residual function, that is, the nonlinear transformation after the convolution operation; y(t) is the final output.

[0082] Convolutional layer stacking:

[0083] Among them, Fi(X) represents the transformation operation of the i-th layer, which may include convolution, pooling, etc.;

[0084] Receptive field: R i =(2·dilation rate i kernel size i )-dilation rate i +1

[0085] Among them, Ri is the receptive field size of the neurons in the i-th layer;

[0086] dilation rate i is the convolution kernel expansion rate of the i-th layer;

[0087] kernel size i is the convolution kernel size of the i-th layer;

[0088] GRU is a variant of LSTM, so it can also solve the long dependency problem in RNN networks; in addition, GRU has one less gate function than LSTM, so the number of parameters is also less than LSTM, so overall the training speed of GRU is faster than that of LSTM.

[0089] The main steps of GRU include:

[0090] Update Gate: Z t =σ(Wz ·[h t-1 ,x t ]);

[0091] Among them, Z t is the reset gate, controlled by the σ function;

[0092] W z is the weight matrix;

[0093] h t-1 : is the hidden state of the previous moment (previous unit time);

[0094] x t is the input of the current moment (current unit time);

[0095] Reset gate: r t =σ(W r ·[h t-1 ,x t ]);

[0096] Among them, r t is the reset gate, also controlled by the σ function;

[0097] W r is the weight matrix;

[0098] h t-1 is the hidden state at the previous moment;

[0099] x t is the input at the current moment;

[0100] Candidate hidden states:

[0101] in, is the new candidate hidden state, obtained by the hyperbolic tangent function tanh;

[0102] W h is the weight matrix;

[0103] r t To reset the gate;

[0104] h t-1 is the hidden state at the previous moment;

[0105] x t is the input at the current moment;

[0106] Hidden state:

[0107] Among them, h t is the updated hidden state;

[0108] z tTo update the door;

[0109] h t-1 is the hidden state at the previous moment;

[0110] is the new candidate hidden state.

[0111] The reset gate determines whether to ignore the previous hidden state. It controls the amount of past information that is reset when calculating the candidate hidden state. The update gate determines how much of the previous hidden state information should be retained. It determines which parts of the past hidden state should be combined with the candidate hidden state to form the current hidden state. The candidate hidden state is a combination of the past hidden state and the current input, which determines whether to retain the information from the previous time step. It is used to update the current hidden state. The hidden state determines how much of the previous step information to retain and how much new information to accept.

[0112] Furthermore, after step S103, the method further includes:

[0113] Step C1: obtaining weather forecast data, holiday schedule data, building feature data and working feature values ​​of the target building object at the time to be predicted; wherein the building feature data includes at least one of the following: building structure data, enclosure material data, heating and air conditioning mode and system distribution data, indoor personnel activity pattern data and building lighting condition data;

[0114] It should be noted that in the process of using the trained prediction model to predict the building power load, the work characteristic values ​​used to characterize the current work intensity type of the employees can be directly obtained;

[0115] Step C2: inputting the building characteristic data and the preset indoor temperature value at the time to be predicted into the energy consumption calculation platform, so as to calculate the air conditioning cooling and heating load of the target building object at the time to be predicted through the energy consumption calculation platform;

[0116] Step C3: The weather forecast data, holiday schedule data and air conditioning cooling and heating loads of the target building object at the time to be predicted are used to form additional feature data of the target building object at the time to be predicted;

[0117] Step C4: inputting the additional characteristic data and the working characteristic value of the target building object at the time to be predicted into the prediction model, and using the output of the prediction model as the predicted building power load value.

[0118] Furthermore, when the additional characteristic data includes weather forecast data, the step C4 specifically includes:

[0119] Step C41: Based on the weather forecast data of the time to be predicted, a weather range value is formed based on a preset weather forecast error range;

[0120] Step C42: using a Monte Carlo algorithm to select a plurality of Monte Carlo sample values ​​within the weather range;

[0121] Step C43: replacing the weather forecast data in the additional feature data with each Monte Carlo sample value in turn and inputting the resultant Monte Carlo sample value into the prediction model together with the working feature value at the time to be predicted, so as to obtain the corresponding predicted building power load value;

[0122] Step C44: forming a predicted building power load interval based on all the obtained predicted building power load values.

[0123] Since weather forecast data may have certain errors, and this error will affect the final prediction result, in this embodiment, in order to eliminate the deviation or error caused by the weather forecast data, a weather range value is first formed based on the weather forecast error range. For example, the temperature value in the weather forecast data is 25°C, and the weather forecast error range is ±1, then the weather range value is 24°C to 26°C; then, a plurality of sample values ​​are randomly selected within the weather range value using the Monte Carlo algorithm; then, the plurality of sample values ​​are used to obtain a plurality of prediction results through the prediction model; finally, the final building power load interval is formed based on the value range of all prediction results.

[0124] In this embodiment, since the prediction model depends on weather forecast data, the lag and deviation between the weather forecast data and the actual value may cause errors; these errors will cause deviations between the predicted value and the actual value; a prediction interval is required during prediction to ensure that the true value is within this range. In this embodiment, the focus is mainly on using the error interval of weather forecast data, and the prediction interval based on weather forecast errors rather than residuals is obtained through the Monte Carlo algorithm, which further enhances the practicality of this embodiment.

[0125] like Figure 2As shown, it is an overall flow chart of predicting building power load provided by this embodiment, which includes a data collection stage, a data preparation stage, and a deep learning model stage; in the data collection stage, historical building power load data, weather forecast data, holiday schedule data, and building feature data (i.e., building structure data) are collected; in the data preparation stage, on the one hand, the building feature data are input into the energy consumption calculation platform to calculate the air conditioning cooling and heating load, and on the other hand, the historical building power load data are input into the K-means clustering algorithm to obtain a critical value for judging the work intensity type (i.e., working state), and the working feature value is determined based on the critical value; finally, at the deep learning model node, the weather forecast data, holiday schedule data, air conditioning cooling and heating load data, and working feature value are used as model inputs and the historical building power load is used as output to train the TCN-GRU model to obtain a prediction model; in addition, when using the prediction model to predict the building power load, the Monte Carlo algorithm can be used to form a prediction interval based on the prediction result.

[0126] The effectiveness of this embodiment is verified by using specific cases as follows:

[0127] The experimental data in this embodiment comes from the measured building power load data of an office building in a certain area from April 17 to November 7, and the data sampling frequency is 1 hour (that is, the unit time is 1 hour). It is planned to use the data from April 17 to October 20 as training data to predict the building power load data from October 20 to November 7.

[0128] First, the K-means clustering algorithm is used to divide all the power load data of historical buildings into two categories, and the dividing lines are 2.3 and 5.8. Therefore, the working characteristic value of the power load less than 2.3 is set to 0; the working characteristic value of the power load greater than 2.3 and less than 5.8 is set to 1, and the working characteristic value of the power load greater than 5.8 is set to 2.

[0129] Then, according to the building characteristic data (including building structure, enclosure material, heating and air conditioning mode and system distribution, indoor personnel activity pattern, lighting conditions, etc.) and the requirements of indoor set temperature value, the energy consumption calculation platform is input to dynamically calculate the building's annual energy consumption and obtain the air conditioning cooling and heating load. The holiday schedule, weather forecast data, air conditioning cooling and heating load, working characteristic value are used as characteristic value X, and the historical building power load is used as target value Y to bring into the TCN-GRU model for training and prediction.

[0130] When using the trained model for prediction, multiple temperature scenarios can be generated based on the possible temperature error range of the weather forecast data, and the predicted building power load range can be further obtained through the Monte Carlo algorithm.

[0131] Through calculation, the MSE value of the prediction model is 0.0063, MAE is 0.0609, SMAPE is 0.2416, and R2 can reach above 0.95, indicating that the prediction accuracy is extremely high. In summary, the rationality and effectiveness of this embodiment are effectively verified, and a certain prediction effect is achieved.

[0132] Embodiment 2

[0133] The embodiment of the present invention provides a device for predicting building power load, such as Figure 3 As shown, the device specifically includes the following components:

[0134] An acquisition module 301 is used to acquire the historical building power load value and additional characteristic data of the target building object at each unit time in the historical time period;

[0135] The determination module 302 is used to compare the historical building power load value of each unit time with a preset threshold value, and determine the work characteristic value for characterizing the work intensity type of each unit time based on the comparison result;

[0136] The training module 303 is used to use the additional characteristic data and working characteristic values ​​of each unit time in the historical time period as model input, and use the historical building power load value of each unit time in the historical time period as model output, train the preset model, and obtain a prediction model for predicting the building power load.

[0137] Specifically, the additional characteristic data includes at least one of the following: weather forecast data, holiday schedule data, and air conditioning cooling and heating loads.

[0138] When the additional characteristic data includes air conditioning cooling and heating loads, the device further includes:

[0139] A calculation module is used to obtain the building characteristic data of the target building object at each unit time in the historical time period; wherein the building characteristic data includes at least one of the following: building structure data, envelope material data, heating and air-conditioning mode and system distribution data, indoor personnel activity law data and building lighting condition data; the building characteristic data of each unit time in the historical time period and the preset indoor temperature value are input into the energy consumption calculation platform, so as to calculate the air-conditioning cooling and heating load of the target building object at each unit time in the historical time period through the energy consumption calculation platform.

[0140] Furthermore, the device also includes:

[0141] A setting module is used to cluster the historical building power load values ​​of each unit time in the historical time period into two subclasses using a K-means clustering algorithm; determine the size value of the center point of each subclass respectively, and determine a first critical point and a second critical point based on the size values ​​of the two center points; wherein the first critical point is smaller than the second critical point; and set the first critical point and the second critical point as the preset threshold value.

[0142] Furthermore, the determining module 302 is specifically configured to:

[0143] Traverse the power load values ​​of historical buildings at each unit time in turn. If the power load value of the historical building currently traversed is less than the first critical point, set the first characteristic value for characterizing weak working intensity for the currently traversed unit time. If the power load value of the historical building currently traversed is greater than the first critical point and less than the second critical point, set the second characteristic value for characterizing medium working intensity for the currently traversed unit time. If the power load value of the historical building currently traversed is greater than the second critical point, set the third characteristic value for characterizing high working intensity for the currently traversed unit time.

[0144] The first eigenvalue is smaller than the second eigenvalue, and the second eigenvalue is smaller than the third eigenvalue.

[0145] Furthermore, the preset model is a TCN-GRU model.

[0146] Furthermore, the device also includes:

[0147] The prediction module is used to input the additional characteristic data and working characteristic values ​​of the target building object at the time to be predicted into the prediction model, and use the output of the prediction model as the predicted building power load value.

[0148] Furthermore, when the additional feature data includes weather forecast data, the prediction module is specifically used to:

[0149] Based on the weather forecast data for the time to be predicted, a weather range value is formed based on a preset weather forecast error range;

[0150] Selecting a plurality of Monte Carlo sample values ​​within the weather range using a Monte Carlo algorithm;

[0151] Substituting each Monte Carlo sample value for the weather forecast data in the additional feature data in turn and inputting the Monte Carlo sample value into the prediction model together with the working feature value at the time to be predicted, so as to obtain the corresponding predicted building power load value;

[0152] A predicted building power load interval is formed based on all the predicted building power load values ​​obtained.

[0153] Embodiment 3

[0154] This embodiment also provides a computer device, such as a smart phone, tablet computer, laptop computer, desktop computer, rack server, blade server, tower server or cabinet server (including an independent server or a server cluster composed of multiple servers) that can execute programs. Figure 4 As shown, the computer device 40 of this embodiment includes at least but not limited to: a memory 401 and a processor 402 that can communicate with each other via a system bus. It should be noted that Figure 4 Only computer device 40 is shown with components 401 - 402 , but it should be understood that implementing all of the components shown is not a requirement, and more or fewer components may alternatively be implemented.

[0155] In this embodiment, the memory 401 (i.e., a readable storage medium) includes a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., an SD or DX memory, etc.), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory 401 may be an internal storage unit of the computer device 40, such as a hard disk or a memory of the computer device 40. In other embodiments, the memory 401 may also be an external storage device of the computer device 40, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the computer device 40. Of course, the memory 401 may also include both the internal storage unit and the external storage device of the computer device 40. In this embodiment, the memory 401 is generally used to store the operating system and various application software installed on the computer device 40. In addition, the memory 401 may also be used to temporarily store various types of data that have been output or are to be output.

[0156] In some embodiments, the processor 402 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips. The processor 402 is generally used to control the overall operation of the computer device 40 .

[0157] Specifically, in this embodiment, the processor 402 is used to execute the program of the method for predicting the building power load stored in the memory 401. When the program of the method for predicting the building power load is executed, the following steps are implemented:

[0158] Obtain the historical building power load value and additional characteristic data of the target building object at each unit time in the historical time period;

[0159] Compare the historical building power load value of each unit time with the preset threshold value, and determine the work characteristic value for characterizing the work intensity type of each unit time based on the comparison result;

[0160] The additional characteristic data and working characteristic values ​​of each unit time in the historical time period are used as model inputs, and the historical building power load values ​​of each unit time in the historical time period are used as model outputs. The preset model is trained to obtain a prediction model for predicting building power load.

[0161] The specific implementation process of the above method steps can be found in Example 1, and this embodiment will not be repeated here.

[0162] Embodiment 4

[0163] This embodiment also provides a computer-readable storage medium, such as a flash memory, a hard disk, a multimedia card, a card-type memory (for example, an SD or DX memory, etc.), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a disk, an optical disk, a server, an App application store, etc., on which a computer program is stored. When the computer program is executed by a processor, the following method steps are implemented:

[0164] Obtain the historical building power load value and additional characteristic data of the target building object at each unit time in the historical time period;

[0165] Compare the historical building power load value of each unit time with the preset threshold value, and determine the work characteristic value for characterizing the work intensity type of each unit time based on the comparison result;

[0166] The additional characteristic data and working characteristic values ​​of each unit time in the historical time period are used as model inputs, and the historical building power load values ​​of each unit time in the historical time period are used as model outputs. The preset model is trained to obtain a prediction model for predicting building power load.

[0167] The specific implementation process of the above method steps can be found in Example 1, and this embodiment will not be repeated here.

[0168] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0169] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0170] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method.

[0171] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for predicting building power load, characterized in that: The method comprises: Obtain the historical building power load value and additional characteristic data of the target building object at each unit time in the historical time period; Compare the historical building power load value of each unit time with the preset threshold value, and determine the work characteristic value of each unit time for characterizing the work intensity type of the employees based on the comparison result; The additional characteristic data and the working characteristic value of each unit time in the historical time period are used as model input, and the historical building power load value of each unit time in the historical time period is used as model output, and the preset model is trained to obtain a prediction model for predicting the building power load; Wherein, when the additional characteristic data includes air conditioning cooling and heating loads, the method further includes: Acquire the building characteristic data of the target building object at each unit time in the historical time period; wherein the building characteristic data includes at least one of the following: building structure data, enclosure structure material data, heating and air conditioning mode and system distribution data, indoor personnel activity law data and building lighting condition data; The building characteristic data of each unit time in the historical time period and the preset indoor temperature value are input into the energy consumption calculation platform, so as to calculate the air conditioning cooling and heating load of the target building object in each unit time in the historical time period through the energy consumption calculation platform.

2. The method for predicting building power load according to claim 1, characterized in that: The additional feature data also includes: weather forecast data and holiday schedule data.

3. The method for predicting building power load according to claim 1, characterized in that: The preset threshold is determined as follows: The K-means clustering algorithm is used to cluster the power load values ​​of the historical buildings at each unit time in the historical time period into two subclasses; Determine the size value of the center point of each subclass respectively, and determine the first critical point and the second critical point based on the size values ​​of the two center points; wherein the first critical point is smaller than the second critical point; The first critical point and the second critical point are set as the preset threshold value.

4. The method for predicting building power load according to claim 3, characterized in that: The historical building power load value of each unit time is compared with a preset threshold value, and a work characteristic value for characterizing the work intensity type of employees of each unit time is determined based on the comparison result, including: Traverse the power load values ​​of historical buildings at each unit time in turn. If the power load value of the historical building currently traversed is less than the first critical point, set the first characteristic value for characterizing weak working intensity for the currently traversed unit time. If the power load value of the historical building currently traversed is greater than the first critical point and less than the second critical point, set the second characteristic value for characterizing medium working intensity for the currently traversed unit time. If the power load value of the historical building currently traversed is greater than the second critical point, set the third characteristic value for characterizing high working intensity for the currently traversed unit time. The first eigenvalue is smaller than the second eigenvalue, and the second eigenvalue is smaller than the third eigenvalue.

5. The method for predicting building power load according to claim 1, characterized in that: The preset model is the TCN-GRU model.

6. The method for predicting building power load according to claim 2, characterized in that: After the preset model is trained to obtain a prediction model for predicting building power load, the method further includes: The additional characteristic data and the working characteristic value of the target building object at the time to be predicted are input into the prediction model, and the output of the prediction model is used as the predicted building power load value.

7. The method for predicting building power load according to claim 6, characterized in that: When the additional characteristic data includes weather forecast data, the additional characteristic data and the working characteristic value of the target building object at the time to be predicted are input into the prediction model, and the output of the prediction model is used as the predicted building power load value, including: Based on the weather forecast data for the time to be predicted, a weather range value is formed based on a preset weather forecast error range; Selecting a plurality of Monte Carlo sample values ​​within the weather range using a Monte Carlo algorithm; Substituting each Monte Carlo sample value for the weather forecast data in the additional feature data in turn and inputting the Monte Carlo sample value into the prediction model together with the working feature value at the time to be predicted, so as to obtain the corresponding predicted building power load value; A predicted building power load interval is formed based on all the predicted building power load values ​​obtained.

8. A device for predicting building power load, characterized in that: The device comprises: An acquisition module, used to acquire the historical building power load value and additional characteristic data of the target building object at each unit time in the historical time period; A determination module, used to compare the historical building power load value of each unit time with a preset threshold value, and determine a work characteristic value for each unit time for characterizing the work intensity type of the employee based on the comparison result; A training module, used to take the additional characteristic data and the working characteristic value of each unit time in the historical time period as model input, and take the historical building power load value of each unit time in the historical time period as model output, train the preset model, and obtain a prediction model for predicting the building power load; Wherein, when the additional characteristic data includes air conditioning cooling and heating loads, the device further includes: A calculation module is used to obtain the building characteristic data of the target building object at each unit time in the historical time period; wherein the building characteristic data includes at least one of the following: building structure data, envelope material data, heating and air-conditioning mode and system distribution data, indoor personnel activity law data and building lighting condition data; the building characteristic data of each unit time in the historical time period and the preset indoor temperature value are input into the energy consumption calculation platform, so as to calculate the air-conditioning cooling and heating load of the target building object at each unit time in the historical time period through the energy consumption calculation platform.

9. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method according to any one of claims 1 to 7 when executing the computer program.

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