Metro station cold load prediction method and system based on similar day and hourly correction

Abstract

This invention proposes a method and system for predicting the cooling load of subway stations based on similar days and hourly corrections, belonging to the field of cooling load prediction technology. Existing technologies often suffer from a situation where the supply of cooling sources exceeds demand, resulting in measured temperatures inside buildings being far lower than design temperatures, leading to significant energy waste. Furthermore, a precise and rapid prediction method is lacking. This invention establishes a historical database of the target subway station and performs feature labeling based on the data. Feature labels include historical cooling load influencing variables and historical cooling load. The feature labels are categorized, and the operating conditions most closely related to the prediction time are identified. The cooling load of the historical cooling load influencing variables is considered as the base value for the cooling load at the prediction time. Simultaneously, the historical cooling load at that time is introduced, and the sum of the base value and the correction term is the predicted cooling load value. This application guarantees prediction accuracy, is unaffected by errors from large datasets, increases the upper limit of accuracy, and achieves a more precise supply-demand balance.

Description

Technical Field

[0001] This invention relates to the field of cooling load forecasting technology, and in particular to a method and system for forecasting the cooling load of subway stations based on similar days and hourly corrections. Background Technology

[0002] Ventilation and air conditioning (VAC) accounts for approximately 40% of a building's energy consumption, making it a major energy consumer and playing a crucial role in achieving dual-carbon goals. In buildings such as subway stations, VAC equipment is specifically responsible for providing the necessary cooling to maintain a suitable temperature inside the building. However, in actual operation, the supply of cooling resources far exceeds the demand, resulting in measured temperatures inside the building being significantly lower than the design temperature, leading to substantial energy waste. Energy conservation and emission reduction must be addressed at the source. Matching supply and demand is the first step towards achieving precise regulation of cooling resources, thus giving rise to the need for predictive cooling load demand.

[0003] Currently, research on building cooling load prediction mainly focuses on above-ground public buildings, and the optimization of prediction models primarily concentrates on data quality control, algorithm adjustment, and hyperparameter selection—the data and algorithmic levels. While such adjustments certainly enhance model robustness and accuracy, the "micro-data, algorithmic approach" can obscure the inherent technicality of the problem. With the development of computer and Internet of Things (IoT) technologies, big data has become a new approach for solving existing problems across various industries. However, computers are merely tools for problem-solving, not the optimal solution for all problems.

[0004] In a building cooling load prediction competition in 2023, the method with the best prediction results in the early stages of the competition was to find cooling loads at similar times. This phenomenon is inspiring. At present, focusing only on the merits of the algorithm and the adjustment of hyperparameters cannot achieve a qualitative leap in improving the quality of prediction results.

[0005] Therefore, it is necessary to predict the cooling load of subway stations. Given the limited research in this field and the poor correlation between prediction methods and professional expertise, this application proposes a prediction method based on the concept of similar days for subway station cooling load. Specifically, it proposes a method and system for predicting subway station cooling load based on similar days and hourly corrections. A similar day refers to finding the historically most similar operating condition to the prediction time, based on an understanding of the concept of cooling load in environmental control engineering. The cooling load at that time is considered the baseline for the prediction time, and a correction term for the cooling load at that time is introduced. The sum of the baseline and the correction term is the predicted value. Summary of the Invention

[0006] To address the issue that existing technologies for predicting cooling load primarily rely on large-scale data accumulation in their models, leading to inaccuracies in predicting cooling load in underground public areas such as subway stations, and a significant mismatch between predicted and actual cooling loads at subway stations, this invention first proposes a subway station cooling load prediction method based on similar days and hourly corrections. The method specifically includes the following steps: Step S1: Establish a historical database of the target subway station and perform feature labeling based on the data. Feature labeling includes historical cooling load influencing variables and historical cooling load. Step S2: Extract the influence factors of each feature marker in the historical database and organize each influence factor accordingly; Step S3: Establish an impact factor model based on the impact factors. The impact factor model includes: scaling, constructing a judgment matrix, and calculating the eigenvalues ​​and eigenvectors of the judgment matrix. Step S4: Weight the impact factor model to obtain the feature weights of each feature label; Step S5: Establish a similar day search model for cooling load, standardize the similar day data, calculate the similarity distance between the predicted day and historical data based on feature weights, and obtain the prediction base through the similar day search model; Step S6: Based on the least squares method and the gray box time-by-time correction model for big data, the predicted change value is obtained; Step S7: Based on the prediction base and the prediction change value, the prediction value is obtained, which is the predicted hourly cooling load.

[0007] Furthermore, in step S1, the feature markers include the hourly train traffic log M, weather type data P, and hourly outdoor dry-bulb temperature T. d Outdoor hourly wet-bulb temperature T w D. Weekday and holiday type; C. Hourly cooling load data; Among them, the variables affecting historical cooling load include the hourly train traffic log M, weather type data P, and hourly outdoor dry-bulb temperature T. d Outdoor hourly wet-bulb temperature T w Type D: Weekdays and Holidays; Historical cooling load includes hourly cooling load data C.

[0008] Furthermore, in step S1, the weather type data P includes: sunny P1, cloudy P2, overcast P3, and rainy P4; The workday and holiday type D includes workdays D1, non-workdays D2 excluding holidays, and holidays D3.

[0009] Furthermore, in step S2, the formula for extracting the influence factors of each feature marker is as follows: ; Where r represents the relative coefficient between two linearly related variables. This represents the i-th data in variable one; This represents the i-th data in variable two; This represents the average value of variable one; This represents the average value of variable two; Substituting the feature labels into the above equation yields the set of coefficients for the feature labels.

[0010] Furthermore, in step S3, the scaling is divided as follows: Set a base scale C = {1, 2, 3, 4, 5, 6, 7, 8, 9}, place the correlation coefficient set into the base scale, and perform scaling on the correlation coefficient set to obtain the coefficient set after scaling. A judgment matrix is ​​constructed based on the coefficient set after scaling. After solving the judgment matrix, the set of matrix eigenvalues ​​and eigenvectors are obtained.

[0011] Furthermore, in step S4, the weighting of the impact factor analysis model includes: first, taking the maximum value in the matrix eigenvalue set and taking the corresponding eigenvector, then subtracting the minimum value from each item in the eigenvector and rounding it down, and then normalizing the minimum value to zero to obtain the eigenweight of each eigenvalue label.

[0012] Furthermore, in step S5, the similar day data is standardized: the similar day data and the historical database data are subjected to maximum and minimum value scaling calculations to obtain the various sets of data on the historical cooling load influencing variables; The similarity distances of the feature weights are calculated sequentially with the historical database data to obtain the similar day with the smallest similarity distance. This day is then selected as the most recent similar day, and the historical data of the cold load influence variables on that day are used as the prediction base.

[0013] Furthermore, in step S6, the similar day data is calculated using the least squares method and the big data gray box hourly correction model to obtain the hourly cooling load correction value of the historical cooling load on the most recent similar day, and this hourly cooling load correction value is used as the predicted change value.

[0014] Furthermore, in step S7, the data of the historical cooling load influence variable of the most recent day is used as the prediction base, the hourly cooling load correction value of the most recent similar day is used as the prediction change value, and the prediction base and the prediction change value are added together to obtain the hourly cooling load of the prediction day.

[0015] According to another aspect of the present invention, a subway station cooling load forecasting system based on similar days and hourly correction is also proposed. The system is applicable to any of the above-mentioned subway station cooling load forecasting methods based on similar days and hourly correction, and specifically includes the following modules: Data acquisition module: Used to collect historical data of the target subway station and classify the historical data of the target subway station into historical cooling load influencing variables, including hourly train traffic logarithmics M, weather type data P, and hourly outdoor dry-bulb temperature T. d Outdoor hourly wet-bulb temperature T w , D, weekday and holiday types; and historical cooling load, which includes hourly cooling load data C; Data processing module: Connects to the data acquisition module, used to mark historical data by features, extract the influence factors of each feature mark, establish an influence factor model, and then weight the influence factors to obtain the weight of each feature mark; establish a cooling load similar day search model based on the feature mark weights, standardize the similar day data, calculate the similarity distance between the similar day and historical data based on the feature weights, and obtain the hourly cooling load value of the most recent similar day as the prediction base through the similar day search model; Data Analysis Module: Connected to the data acquisition module and data processing module, it is used to predict changes. It uses the least squares method and a big data gray box hourly correction model to obtain the hourly correction value of historical cooling load on the most recent similar day, which is used as the predicted change value. Cooling load forecasting module: Connected to the data processing module and data analysis module, it is used to forecast the cooling load for the day. Specifically, it obtains the forecast value of the cooling load for the forecast day by calculating the forecast base and the forecast change value.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: Firstly, this invention establishes a historical database of the target subway station and performs feature labeling based on the data. The feature labels include historical cooling load influencing variables and historical cooling load itself. The data in the historical database is then categorized, identifying the historical operating conditions most closely related to the predicted time. The cooling load of similar historical days is used as the prediction base, and the historical cooling load at that time is also incorporated. The sum of the base and the correction term is the predicted cooling load value. This data-driven cooling load prediction algorithm, along with hourly correction values, allows for precise refinement to the prediction value for a specific day, ensuring prediction accuracy and preventing the impact of large-scale data errors. Secondly, the influence factors of each feature marker in the historical database are extracted and organized accordingly. An influence factor model is then established, and the influence factor model is weighted to obtain the feature weights of each feature marker. A similarity day search model for cooling load is established, and the similarity distance between the predicted day and historical data is calculated based on the feature weights. The prediction base is obtained through the similarity day search model. Similar operating days are found, and the cooling load prediction is refined to each time of day. Traditional prediction methods have varying prediction results, with some days showing excellent performance and others extremely poor performance. While the upper limit is high, the lower limit is low, making them insufficient for practical operation and maintenance. The hourly correction term in this invention, based on the base, makes the predicted cooling load value at each time more consistent with the operating conditions of that day, achieving a more accurate supply-demand balance. Thirdly, this invention is based on the least squares method and a gray box hourly correction model for big data to obtain the predicted change value. It also obtains the predicted value based on the prediction base and the predicted change value, thereby improving the accuracy of the prediction and the flexibility of the cooling load prediction. Attached Figure Description

[0017] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 The flowchart shows the method for predicting the cooling load of subway stations based on similar days and hourly corrections. Figure 2 This is a schematic diagram of an embodiment of a subway station cooling load prediction method based on similar days and hourly corrections; Figure 3 This is a structural diagram of a subway station cooling load prediction system based on similar days and hourly corrections. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0020] The specific embodiments of the present invention will be described below.

[0021] To address the issue of inaccurate energy consumption control and prediction in existing technologies for public building areas, leading to significant resource waste, this invention proposes a prediction method and system for subway station cooling load based on the concept of similar days and hourly corrections. A similar day refers to finding the historically most similar operating condition to the predicted time, based on the concept of cooling load in environmental control systems. The cooling load at that time is considered the baseline for the predicted cooling load, and a correction term for that time is introduced, with the sum of the baseline and the correction term equal to the predicted value.

[0022] Example 1 like Figure 1 and Figure 2 As shown, this invention proposes a method for predicting the cooling load of subway stations based on similar days and hourly corrections, specifically including the following steps: Step S1: Establish a historical database of the target subway station based on the subway station operation data, and perform feature labeling based on the data. The feature labeling includes historical cold load influencing variables and historical cold load.

[0023] The goal is to find the day in history that is most similar to the predicted day's operating conditions, use the hourly cooling load of that day as the baseline for the predicted hourly cooling load, and collect other variables that affect the cooling load, including hourly train traffic logs, weather type data, hourly outdoor dry-bulb temperature, hourly outdoor wet-bulb temperature, weekday and holiday types, and hourly cooling load data.

[0024] In this embodiment, since the data collected is based on the operating conditions of similar days, it includes the impact of factors such as the number of trains running in the target subway station, weather conditions, outdoor temperature, and whether it is a weekday or holiday on the cooling load of the target subway station during the operating hours of that day. These factors are used as historical cooling load influence variables. Based on the above data, feature labels are created, and each data feature is labeled as follows: hourly train running count M, weather type data P, and hourly outdoor dry-bulb temperature T. d Outdoor hourly wet-bulb temperature T w D. Weekday and holiday data type; C. Hourly cooling load data. Specifically, the hourly train operation pair number M = {M1, M2, M3, … ,M} is set. n Weather type data P = { P1, P2, P3, …, P} n Outdoor hourly dry-bulb temperature T d = { Td1, Td2, Td3, …, Td n Outdoor hourly wet-bulb temperature T w = { Tw1, Tw2, Tw3, …,Tw n}, the types of weekdays and holidays D = { D1, D2, D3, …, D n}, Hourly cooling load data C = { C1, C2, C3, ..., C n For weather-related data, sunny days are assigned a value of 1, cloudy days a value of 2, overcast days a value of 3, and rainy days a value of 4. For work and holiday-related data, workdays are defined as D1 with a value of 1; non-workdays (excluding holidays) are defined as D2 with a value of 2; and holidays, such as New Year's Day, Spring Festival, Qingming Festival, Labor Day, Dragon Boat Festival, Mid-Autumn Festival, and National Day, are defined as D3 with a value of 3. In the above features, n represents the number of historical data days.

[0025] Specifically, each element is set as a 17-dimensional vector, with the outdoor hourly dry-bulb temperature T as the vector. d Taking the data as an example, the first element Td1 in this feature is (20.2, 21.3, 22.5, 22.6, 22.9, 24.0, 25.1, 26.9, 26.9, 27.7, 2.69, 26.4, 25.2, 24.5, 23.3, 23.1, 22.8). This represents the hourly fluctuation data of the target station's dry-bulb temperature on a certain day from 6:00 to 22:00 in the historical database. The outdoor temperature includes both hourly dry-bulb temperature and hourly wet-bulb temperature. The hourly dry-bulb temperature refers to the actual temperature of the ambient air during normal days, unaffected by moisture evaporation and cooling. The wet-bulb temperature refers to the temperature saturated with water vapor under isobaric adiabatic conditions, representing the extreme temperature reached after moisture evaporation and heat absorption. By controlling multiple feature data in this way, a comprehensive understanding of the daily conditions within the historical data is achieved. This approach, which does not rely entirely on big data calculations, ensures the accuracy of predictions, and the results are not affected by the errors of large datasets.

[0026] Step S2 involves extracting the influence factors of each feature label from the historical database and preprocessing each influence factor. After preparing all data feature labels, the correlation coefficients of each data feature are calculated sequentially, and the absolute values ​​of the calculated correlation coefficients for each feature are compiled. The details are as follows: ; Where r represents the relative coefficient between two linearly related variables. This represents the i-th data in variable one; This represents the i-th data in variable two; This represents the average value of variable one; This represents the average value of variable two. Substituting the various feature labels into the above formula and rearranging, we obtain a set of correlation coefficients: ; Specifically, R16 R represents the absolute value of the correlation coefficient calculated between the hourly logarithmic number of train operations and the hourly data of cooling load; 26 R represents the absolute value of the correlation coefficient calculated between weather type data and hourly cooling load data. 36 R represents the absolute value of the correlation coefficient calculated between hourly outdoor dry-bulb temperature and hourly cooling load data; 46 R represents the absolute value of the correlation coefficient calculated between hourly outdoor wet-bulb temperature and hourly cooling load data; 56 This represents the absolute value of the correlation coefficient between weekdays and holidays and hourly data on cooling load.

[0027] Step S3: Establish an impact factor model based on the preprocessing results of the impact factors. Establishing the impact factor model includes: scaling, constructing a judgment matrix, and calculating the eigenvalues ​​and eigenvectors of the judgment matrix. Based on the processed impact factors, the factor model of the impact factors is established. First, scaling is performed. In this embodiment, the basic scale is set as C = {1, 2, 3, 4, 5, 6, 7, 8, 9}, and the aforementioned set of correlation coefficients R is scaled to the basic scale C. Specifically, the i-th element in set R is set to R... i This indicates that, after scaling, R i The value becomes: round(8×(R) i -R min ) / (R max -R min )+1), where R min The value is 1, R max The value is 9. The `round()` function indicates that the value is rounded to the nearest integer. From this, we obtain the set of correlation coefficients R = {R... 16 , R 26 , R 36 , R 46 , R 56}

[0028] Next, the judgment matrix is ​​constructed by selecting the set of correlation coefficients after scaling: ; Calculate the eigenvalues ​​and eigenvectors of the matrix, based on the definition of a matrix, and ensure that real numbers exist. Non-zero n-dimensional column vectors Satisfy the following formula: ; After solving, we obtain the set of matrix eigenvalues: { 1, 2, 3, 4, 5}, and the corresponding feature vectors: { 1, 2, 3, 4, 5}.

[0029] Step S4 involves establishing the impact factor model and assigning weights to it to obtain the feature weights for each feature label. The largest eigenvalue is then selected. corresponding feature vector Each term in the eigenvector is subtracted from the minimum value and then positive, i.e. After calculation, let To achieve feature vector normalization, i.e. Finally, the weights of each feature are obtained. 1, 2, 3, 4, 5}. Among them, 1 is the calculation weight of the hourly train operation logarithm for matching similar days of cooling load; This refers to the weighting of weather type data when matching days with similar loads. The calculation weight of outdoor hourly dry-bulb temperature for load similarity day matching; The calculation weight of outdoor hourly wet-bulb temperature for load similarity day matching; The calculation weight of weekdays and holidays when matching load-similar days.

[0030] Step S5: Establish a similar day search model for cooling load based on the weights of each feature to obtain similar day data. Standardize the similar day data and calculate the similar distance between the similar day and historical data according to the feature weights. The day with the smallest similar distance is the most recent similar day. Define the hourly cooling load value of the most recent similar day as the prediction base.

[0031] First, similar daily data of the same type and historical databases are scaled based on their maximum and minimum values ​​to achieve standardization: ; Where, x i X is the i-th element corresponding to the standardized data in the historical cooling load influence variables; i X is the i-th element in the dataset corresponding to the historical cooling load influence variables; min X represents the minimum value of the corresponding element in the dataset of historical cooling load influence variables; max This represents the maximum value of the corresponding element in the dataset of historical cold load influence variables.

[0032] Thus, a standardized similar day dataset is obtained: Predict the number of train pairs running hourly each day (m) pre =(m1, m2, m3, … ,m 17 ); Forecast daily weather type data p pre =(p1,p2,p3,…,p 17 ); Predicted daily outdoor hourly dry-bulb temperature t d-pre =(td1, td2, td3, … ,td 17 ); Predicted daily outdoor hourly wet-bulb temperature t w-pre =(tw1, tw2, tw3, … ,tw 17 ); Forecast of workday and holiday types d pre =(d1,d2,d3, … ,d 17 ).

[0033] Next, the similarity distance L between similar days and historical data is calculated sequentially according to the weights, as shown in the following formula:

[0034] In the formula, This represents the similarity distance between a similar day and the nth day in historical data; The time indicates the hour, which is the time period during which the subway station operates, from 6:00 AM to 10:00 PM (i = 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22). Indicates the first At present, the number of vehicles driving in recent days; Historical data on day n The number of trains traveling at any given time; Indicates the first What is the current weather pattern? Historical data on day n Weather type at any given time; Indicates the first At present, the dry-bulb temperature is similar to that of the past few days; Historical data on day n Dry bulb temperature at any given time; Indicates the first At this moment, the wet-bulb temperature is similar to that of the previous day; Historical data on day n The wet-bulb temperature at any given moment; Indicates the first Currently, what are the types of holidays? Historical data on day n The types of holidays at any given time; , , , , This represents the weight of each feature.

[0035] Next, calculate the set of similar distances for all days in the recent and historical data:

[0036] Find the minimum similarity distance The time corresponding to the minimum similarity distance is called the nearest similar day; the hourly cooling load on the nearest similar day. This serves as the baseline for predicting the daily cooling load.

[0037] Step S6: Based on similar daily data, the predicted change value is obtained using the least squares method and a large-scale gray-box hourly correction model. The large-scale gray-box hourly correction model refers to improving forecast accuracy by pre-screening data using a mechanistic model; the specific model is as follows: ; in, This represents the load correction value at time i on the predicted date; This indicates the predicted dry-bulb temperature at time i on the predicted day; This indicates the predicted wet-bulb temperature at time i on the predicted day; This represents the number of vehicle trips at time i on the predicted day; The time of the predicted date is represented by (i = 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22). A represents the amplitude of indoor load change caused by the piston effect of the vehicle. The above historical data are fitted with values ​​related to the weather type using the forgetting least squares method. That is, A is 32.95 for sunny weather, 25.12 for cloudy weather, 15.55 for overcast weather, and 13.08 for rainy weather.

[0038] Step S7: Based on the forecast base and the forecast change value, the forecast value is obtained, which is the predicted hourly cooling load. The predicted daily hourly cooling load value is calculated as follows: ; In the formula, This indicates the predicted load at time i on the predicted day; This represents the load base at time i on the predicted day; This represents the load correction value at time i on the predicted day. This allows us to obtain the accurate predicted cooling load value within the subway station.

[0039] The prediction results in this embodiment are more stable and suitable for practical engineering applications. This avoids the situation where current data-driven prediction models (such as neural networks and support vector machines), while exhibiting high overall average accuracy in testing, show drastic fluctuations in prediction performance at the daily level. Some predictions are extremely accurate, while others have large errors, resulting in a situation of "high upper limit and low lower limit." This instability makes it difficult for maintenance personnel to trust the prediction results and dare not directly use them to guide the real-time scheduling and energy-saving control of the air conditioning system, fearing serious cooling shortages or waste. The model has strong interpretability, breaking through the "black box" limitation and constructing a "grey box" prediction framework. Figure 2 As shown, the prediction base is established by creating a historical database of stations, a database of characteristic influencing factors, an analysis model of characteristic variable influencing factors, weighting of influencing factors, and a selection model for similar days of cooling load. The predicted change value is then obtained by adding the two. Specifically, the amplitude coefficients calculated using the least squares method for different weather conditions are assigned values. The value of A is determined based on the weather type of the prediction day: the first preset value is used for sunny days, the second for cloudy days, the third for overcast days, and the fourth for rainy days. The first preset value is greater than the second, the second is greater than the third, and the third is greater than the fourth. The first preset value is 32.95, the second is 25.12, the third is 15.55, and the fourth is 13.08. This calculation method yields accurate values ​​that also reflect actual weather conditions.

[0040] In a specific embodiment, taking a subway station in Tianjin as an example, this subway station is set as the target subway station, and historical data of the target subway station is collected. The time period is selected from May 1st to September 2nd, a total of 125 days, and the daily working hours of the subway station are set to 6:00 to 22:00, resulting in a total of 2125 sets of data (some data is omitted). Among them, the hourly train operation pairs are set as feature 1, weather type data is set as feature 2, outdoor hourly dry-bulb temperature is set as feature 3, outdoor hourly wet-bulb temperature is set as feature 4, weekday and holiday types are set as feature 5, and the hourly cooling load data is defined as the target layer, as shown in the table below: Table 1. Historical data of a subway station in Tianjin over 125 days.

[0041] This embodiment incorporates 2125 sets of data (omitted in the table). Following the steps in this embodiment, predictions can be made to obtain similar days for the specific day to be predicted, as well as the most recent similar day obtained through further calculation. The data of the most recent similar day is used as the prediction base. Based on the data of the similar days, the predicted change value is obtained through the least squares method and a large-scale gray-box hourly correction model. The prediction base and the predicted change value are then superimposed to obtain the specific cooling load prediction value for the day to be predicted. This method yields more accurate prediction results that conform to actual operating conditions, avoiding resource waste and saving costs.

[0042] Example 2 like Figure 3 As shown, the present invention also proposes a subway station cooling load forecasting system based on similar days and hourly corrections, using the subway station cooling load forecasting method based on similar days and hourly corrections as described in any of Embodiment 1, including the following modules: Data acquisition module: Used to collect historical data of the target subway station and classify the historical data of the target subway station into historical cooling load influencing variables, including hourly train traffic logarithmics M, weather type data P, and hourly outdoor dry-bulb temperature T. d Outdoor hourly wet-bulb temperature T w The data includes: D (weekday and holiday types); and historical cooling load, which includes hourly cooling load data (C).

[0043] Data processing module: Connected to the data acquisition module, this module is used to label historical data by features, extract the influencing factors of each feature label, establish an influencing factor model, and then weight the influencing factors to obtain the weights of each feature label. Based on the feature label weights, a similarity day search model for cooling load is established. The similarity day data is standardized, and the similarity distance between the similar day and historical data is calculated based on the feature weights. The hourly cooling load value of the most recent similar day is obtained through the similarity day search model as the prediction base.

[0044] Data Analysis Module: Connected to the data acquisition module and data processing module, it is used to predict changes. It uses the least squares method and a big data gray box hourly correction model to obtain the hourly correction value of historical cooling load on the most recent similar day, which is then used as the predicted change value.

[0045] Cooling load forecasting module: Connected to the data processing module and data analysis module, it is used to forecast the cooling load for the day. Specifically, it obtains the forecast value of the cooling load for the day by calculating the forecast base and the forecast change value.

[0046] This invention establishes a historical database of the target subway station based on actual subway station data and performs feature labeling based on the data. The feature labels include historical cooling load influencing variables and historical cooling load. The data in the historical database is then categorized, and the operating conditions most closely related to the prediction time are identified. The cooling load of the historical cooling load influencing variables is considered as the base value for the cooling load at the prediction time. Simultaneously, the historical cooling load at that time is introduced, and the sum of the base value and the correction term is the predicted cooling load value. This method, based on definite data and adjusted according to actual conditions, ensures the accuracy of the predicted cooling load value. Finding similar operating days allows for the refinement of the cooling load prediction to each hour of each day. Traditional forecasting methods often produce excellent results on some days and poor results on others, with a high upper limit but a low lower limit, making them unsuitable for practical operation and maintenance. This invention's hourly correction term, based on the base value, makes the predicted cooling load value at each hour more consistent with the operating conditions of that day, achieving a more accurate supply-demand balance. Based on the least squares method and a gray-box hourly correction model using large data sets, the predicted change value is obtained, improving the accuracy and flexibility of the prediction.

[0047] Example 3 An electronic device, comprising: Processor and memory; The processor executes the steps of the subway station cooling load prediction method based on similar days and hourly correction, as described in any of Embodiment 1, by calling programs or instructions stored in memory.

[0048] Example 4 A computer-readable storage medium includes computer program instructions that cause a computer to perform the steps of a subway station cooling load prediction method based on similar days and hourly corrections as described in any of Embodiment 1.

[0049] Computer-readable storage media may take the form of any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may, for example, include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0050] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting the cooling load of subway stations based on similar days and hourly corrections, characterized in that, Includes the following steps: Step S1: Establish a historical database of the target subway station based on the subway station data, and perform feature marking based on the data. The feature marking includes historical cooling load influencing variables and historical cooling load. Step S2: Extract the influence factors of each feature marker in the historical database, and preprocess each influence factor; Step S3: Establish an impact factor model based on the impact factor preprocessing results. The steps for establishing the impact factor model include: scaling, constructing a judgment matrix, and calculating the eigenvalues ​​and eigenvectors of the judgment matrix. Step S4: While establishing the impact factor model, the impact factor model is also weighted to obtain the feature weights of each feature label. Step S5: Establish a similar day search model for cooling load based on the weights of each feature to obtain similar day data. Standardize the similar day data and calculate the similar distance between the similar day and the historical data based on the weights of each feature. The day with the smallest similar distance is the most recent similar day. Define the hourly cooling load value of the most recent similar day as the prediction base. Step S6: Based on the similar day data, the predicted change value is obtained by using the least squares method and the big data gray box hourly correction model; Step S7: Obtain the predicted value based on the predicted base and the predicted change value. The predicted value is the predicted hourly cooling load.

2. The method for predicting the cooling load of subway stations based on similar days and hourly corrections according to claim 1, characterized in that, In step S1, the feature markers include the hourly train traffic log M, weather type data P, and hourly outdoor dry-bulb temperature T. d Outdoor hourly wet-bulb temperature T w D. Weekday and holiday type; C. Hourly cooling load data; The historical cooling load influencing variables include the hourly train traffic log M, the weather type data P, and the hourly outdoor dry-bulb temperature T. d The outdoor hourly wet-bulb temperature T w The workday and holiday type D; The historical cooling load includes the hourly cooling load data C.

3. The method for predicting the cooling load of subway stations based on similar days and hourly corrections according to claim 2, characterized in that, In step S1, the weather type data P includes: sunny P1, cloudy P2, overcast P3, and rainy P4; The workday and holiday type D includes workday D1, non-workdays D2 excluding holidays, and holidays D3.

4. The method for predicting the cooling load of subway stations based on similar days and hourly corrections according to claim 3, characterized in that, In step S2, the formula for extracting the influence factors of each feature marker is as follows: ； Where r represents the relative coefficient between two linearly related variables. This represents the i-th data in variable one; This represents the i-th data in variable two; This represents the average value of variable one; This represents the average value of variable two; Substituting the feature labels into the above formula yields the coefficient set of the feature labels.

5. The method for predicting the cooling load of subway stations based on similar days and hourly corrections according to claim 4, characterized in that, In step S3, the scaling is divided as follows: Set a basic scale C = {1, 2, 3, 4, 5, 6, 7, 8, 9}, place the coefficient set into the basic scale, and perform scaling on the coefficient set to obtain the scaled coefficient set; The judgment matrix is ​​constructed based on the coefficient set after scaling. After solving the judgment matrix, the set of matrix eigenvalues ​​and eigenvectors are obtained.

6. The method for predicting the cooling load of subway stations based on similar days and hourly corrections according to claim 5, characterized in that, In step S4, the weighting of the influence factor analysis model includes: first, taking the maximum value in the set of matrix eigenvalues ​​and taking the corresponding eigenvector as equal; then, subtracting the minimum value from each item in the eigenvector and rounding it down; then, normalizing the minimum value to zero to obtain the eigenweight of each eigenlabel.

7. The method for predicting the cooling load of subway stations based on similar days and hourly corrections according to claim 1, characterized in that, In step S5, the similar day data is standardized: the similar day data and historical database data are subjected to maximum and minimum value scaling calculation to obtain the various sets of data of the historical cooling load influence variables; The similarity distances of the feature weights are calculated sequentially with the historical database data to obtain the similar day with the smallest similarity distance. This day is then selected as the most recent similar day, and the historical cold load influence variable data on the most recent similar day is used as the prediction base.

8. The method for predicting the cooling load of subway stations based on similar days and hourly corrections according to claim 1, characterized in that, In step S6, the similar day data is calculated using the least squares method and the big data gray box hourly correction model to obtain the hourly cooling load correction value of the historical cooling load on the most recent similar day, and the hourly cooling load correction value is used as the predicted change value.

9. The method for predicting the cooling load of subway stations based on similar days and hourly corrections according to claim 1, characterized in that, In step S7, the data of the historical cooling load influence variable of the most similar day is used as the prediction base, the hourly cooling load correction value of the most recent similar day is used as the prediction change value, and the prediction base and the prediction change value are added together to obtain the hourly cooling load of the prediction day.

10. A subway station cooling load forecasting system based on similar days and hourly corrections, the system being applicable to the subway station cooling load forecasting method based on similar days and hourly corrections as described in any one of claims 1 to 9, characterized in that, Specifically, it includes the following modules: Data acquisition module: Used to collect historical data of the target subway station and classify the historical data of the target subway station into historical cooling load influencing variables, including hourly train traffic logarithmics M, weather type data P, and hourly outdoor dry-bulb temperature T. d Outdoor hourly wet-bulb temperature T w , D, weekday and holiday types; and historical cooling load, which includes hourly cooling load data C; Data processing module: connected to the data acquisition module, used to mark the historical data by features, extract the influence factors of each feature mark, establish an influence factor model, and then weight the influence factors to obtain the weight of each feature mark; A similar day search model for cooling load is established based on feature label weights. The similar day data is standardized, and the similarity distance between the similar day and historical data is calculated based on the feature weights. The hourly cooling load value of the most recent similar day is obtained through the similar day search model as the prediction base. Data analysis module: connected to the data acquisition module and the data processing module, used to predict the change value, and to obtain the hourly cooling load correction value of the historical cooling load on the most recent similar day using the least squares method and the big data gray box hourly correction model, as the predicted change value; Cooling load forecasting module: Connected to the data processing module and the data analysis module, it is used to forecast the cooling load for the day. Specifically, it obtains the forecast value of the cooling load for the forecast day by calculating the forecast base and the forecast change value.

Images