Subway station energy consumption working condition fine division method and system

Abstract

The invention discloses a metro station energy consumption working condition fine division method and system, and relates to the technical field of metro energy consumption. According to the method, a cross-correlation function and a fused fuzzy time cognitive map model are used for dynamically analyzing time delay correlation among multi-dimensional data, a data scanning window is adjusted according to subway operation characteristics, data time migration is corrected, and data synchronism and reliability are improved; a sliding window method and a dynamic window width adjusting strategy are adopted, so that the adaptability of the model to different operation time periods is enhanced, and the accuracy and robustness of data correction are ensured; through Savitzky-Golay filtering, jump point detection and a dynamic trend model, transverse errors of multi-dimensional data on a time axis are eliminated, key jump points are aligned, transverse error correction is refined, and high-consistency fine correction data are provided for subsequent working condition division.

Description

Technical Field

[0001] The present invention relates to the technical field of subway energy consumption, and in particular to a method and system for finely dividing the energy consumption conditions of a subway station. Background Art

[0002] With the rapid development of subway construction, energy consumption in subway stations has become a major concern. Existing methods for classifying subway station operating conditions based on operational data typically assume temporal data synchronization, meaning that the time coordinates of the sampling data for different variables are consistent. However, in actual subway station operations, the response characteristics of different equipment to passenger flow vary significantly. For example, elevators can respond quickly to passenger flow, while air conditioners require a certain amount of passenger flow to respond. Furthermore, when passenger flow triggers changes in air flow and pressure parameters, there is a delay in the changes in variables such as temperature, CO2 concentration, and air conditioning cooling load. Therefore, even if the sampling data for each variable has the same time coordinates, the information lags, causing uncorrected operating condition classification to deviate from reality and fail to meet the refined energy conservation needs of subway stations. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and system for fine-grained division of energy consumption conditions in subway stations, so as to improve the above-mentioned technical problems.

[0004] In order to achieve the above-mentioned object of the invention, the embodiment of the present invention provides the following technical solutions:

[0005] A method for finely dividing energy consumption conditions of subway stations is provided, which includes:

[0006] Obtaining multi-dimensional hourly energy consumption data of a subway station; the multi-dimensional hourly energy consumption data includes hourly carbon dioxide sensor data, hourly equipment energy consumption data, hourly passenger flow data, and hourly temperature data;

[0007] The multi-dimensional hourly energy consumption data is corrected using the cross-correlation function and the fusion fuzzy time cognitive graph model to obtain the multi-dimensional hourly energy consumption correction data.

[0008] Perform lateral error correction on the multi-dimensional hourly energy consumption correction data to obtain multi-dimensional hourly energy consumption precision calibration data;

[0009] The TICC algorithm is used to divide the multi-dimensional hourly energy consumption precision calibration data to obtain the working condition division results.

[0010] A system for fine-grained classification of energy consumption conditions in subway stations is provided, which includes:

[0011] Multi-dimensional hourly energy consumption collection module, used to obtain multi-dimensional hourly energy consumption data of a subway station;

[0012] A data correction module is used to correct the multi-dimensional hourly energy consumption data using a cross-correlation function and a fusion fuzzy time cognitive graph model to obtain multi-dimensional hourly energy consumption correction data;

[0013] The data precision calibration module is used to perform lateral error correction on the multi-dimensional hourly energy consumption correction data to obtain multi-dimensional hourly energy consumption precision calibration data;

[0014] The working condition division module is used to divide the multi-dimensional hourly energy consumption precision calibration data using the TICC algorithm to obtain the working condition division results.

[0015] The beneficial effects of the present invention are:

[0016] This invention uses cross-correlation functions and a fused fuzzy time cognitive graph model to dynamically analyze the time delay correlation between multidimensional data. It also adjusts the data scanning window according to the characteristics of subway operations, corrects data time offsets, and improves data synchronization and reliability. A sliding window method and a dynamic window width adjustment strategy are used to enhance the model's adaptability to different operating periods, ensuring the accuracy and robustness of data correction. Through Savitzky-Golay filtering, transition point detection, and a dynamic trend model, lateral errors in multidimensional data on the time axis are eliminated, key transition points are aligned, and lateral error correction is refined, providing highly consistent, precisely calibrated data for subsequent operating condition division. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 A flow chart of a method in an embodiment of the present invention;

[0019] Figure 2 Flowchart of an improved wave search algorithm in an embodiment of the present invention;

[0020] Figure 3 2 is a system structure diagram in an embodiment of the present invention. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present invention.

[0022] See also Figure 1 This embodiment provides a method for finely dividing the energy consumption conditions of a subway station, which includes:

[0023] S1. Obtain multi-dimensional hourly energy consumption data for a subway station. The multi-dimensional hourly energy consumption data includes hourly CO2 sensor data, hourly equipment energy consumption data, hourly passenger flow data, and hourly temperature data. All data points in each category are time series. The hourly equipment energy consumption data includes hourly energy consumption data for chillers, fans, lighting, elevators, and VLSI units. Each hourly data point has a corresponding timestamp. For example, hourly energy consumption data includes energy consumption data and its timestamp.

[0024] S2. Use the cross-correlation function (CCF) and the fused fuzzy temporal cognitive graph model (FTCM) to correct the multi-dimensional hourly energy consumption data to obtain multi-dimensional hourly energy consumption correction data.

[0025] The S2 includes:

[0026] S2-1. Calculate the time delay between any two data types in multi-dimensional hourly energy consumption data. Taking hourly CO2 sensor data and hourly device energy consumption data as examples, when calculating the time delay between any hourly CO2 sensor data point and any hourly device energy consumption data point, the time delay refers to the time offset of the CO2 sensor data relative to the hourly device energy consumption data, and can be positive or negative.

[0027] S2-2. Use the cross-correlation function to calculate the correlation coefficient of any two data types in the multi-dimensional hourly energy consumption data at different time delays.

[0028] Therefore, the formula corresponding to S2-2 is:

[0029]

[0030] Among them, a and b represent any two different data types in the multi-dimensional hourly energy consumption data, k represents the delay, and CCF ab (k) represents the correlation coefficient between data type a and data type b at time delay k, a i Represents the i-th data point in data type a, b i+k represents the i+kth data point in data type b, μ a 、μ b Respectively represent the mean values corresponding to data type a and data type b, σ a , σ b They represent the standard deviations corresponding to data types a and b, respectively, n represents the data length, and ∑(·) represents the summation function.

[0031] S2-3. Sort the correlation coefficients of any two data types under different delays from large to small, select the maximum value of the corresponding correlation coefficient as the correlation result (correlation degree) corresponding to the two data types, and use the delay corresponding to each correlation result as the delay estimation value.

[0032] S2-4. Based on the operational characteristics of the subway station, the correlation results, and their estimated delays, determine the data scanning range and dynamically adjust the window width. Based on the operational characteristics of the subway station, set the scanning range to one week, and initially set the window width to 24 hours to capture daily patterns. Adjust the window width based on the correlation results and their estimated delays, setting the window to at least include the estimated delay value corresponding to the correlation result. If the correlation exceeds the correlation threshold, reduce the window width to increase sensitivity to delay changes; otherwise, increase the window width. The window width reduction and increase values are set based on actual needs.

[0033] S2-5. Based on the data scanning range, the sliding window method is used to calculate the first time delay between the multi-dimensional hourly energy consumption data in the window, which is conducive to extracting the relationship between the first time delay and correlation results of the multi-dimensional hourly energy consumption data over time and analyzing the dynamic changes in the status of various equipment in the subway station.

[0034] S2-6. Based on the first time delay and its correlation coefficient, the multi-dimensional hourly energy consumption data is corrected using a fused fuzzy temporal cognitive map model (FTCM) to obtain multi-dimensional hourly energy consumption correction data.

[0035] The S2-6 includes:

[0036] S2-6-1. Perform trend segmentation on multi-dimensional hourly energy consumption data to obtain trend units. The process is as follows:

[0037] Obtain the changing trend of multi-dimensional hourly energy consumption data; the changing trend includes rising, falling and stable. Taking the hourly energy consumption data corresponding to the fan as an example, the hourly energy consumption data is defined as X = [x1, x2, ..., x t ,…,x T ], according to the energy consumption value at each moment, calculate the energy consumption difference at adjacent moments; based on each energy consumption difference, divide the energy consumption hourly data; if the energy consumption difference is the energy consumption difference x in a certain continuous time period t -x t-1 If the energy consumption difference values in a certain continuous time period are all negative, the change trend of the continuous time period is downward; in other cases, the change trend of the continuous time period is stable. Among them, T represents the total duration of the hourly energy consumption data corresponding to the fan, x1, x2, x t-1 、x t 、x T They represent the energy consumption values corresponding to time 1, time 2, time t-1, time t and the last time respectively.

[0038] Based on the changing trend, the starting point of the changing trend is used as the segmentation point, and the multi-dimensional hourly energy consumption data is segmented using the piecewise linear approximation method to obtain the trend units corresponding to the carbon dioxide sensor hourly data, equipment energy consumption hourly data, passenger flow hourly data, and temperature hourly data. Taking the hourly energy consumption data corresponding to the fan as an example, the expression of each trend unit is (t ij,0 ,p ij ,y ij,0 ), p ij , t ij,0 、y ij,0 They are respectively the trend slope and starting time (reflecting the rate of change) of the jth time period in the i-th energy consumption hourly data and its variable value (vertical intercept / energy consumption value).

[0039] S2-6-2. Determine the linear relationship corresponding to each trend unit; based on each linear relationship, calculate the predicted value of each trend unit. Taking the hourly energy consumption data corresponding to the fan as an example, the formula for the linear relationship corresponding to the trend unit is:

[0040] y ij,n (t ij,n )=p ij (t ij,n -t ij,0 )+y ij,0 ;

[0041] Among them, t ij,n represents the nth moment in the jth time period of the i-th energy consumption hourly data, y ij,n (t ij,n ) represents the time tij,n The corresponding variable value.

[0042] Taking the hourly energy consumption data corresponding to the fan as an example, the formula corresponding to the predicted value is:

[0043]

[0044] Where, Δt ij,n Indicates time t ij,n and the starting time t ij,0 The difference between , s represents the number of sampling points, represents the predicted value corresponding to the nth moment in the jth time period of the i-th hourly energy consumption data. Since different variables have different time delays, for example, the time delay between CO2 and the energy consumption data corresponding to the fan is 3, and the time delay between fan energy consumption and passenger flow is 2. Based on fan energy consumption, it is necessary to calculate the difference between the CO2 value at the current moment and the three previous moments, as well as the difference between the passenger flow value at the current moment and the two previous moments. Thus, the moments corresponding to the time delays between different variables are used as sampling points, and the number of sampling points, s, is obtained. In this embodiment, the unit of delay is hour (h).

[0045] S2-6-3. Calculate the error between each predicted value and the multi-dimensional hourly energy consumption data at the same time, i.e., at time t ij,n The variable value under is subtracted from the predicted value, and the difference is the corresponding error.

[0046] S2-6-4. Based on the first time delay, its correlation coefficient, and each error, determine the nodes, directed edges, edge weights, and fuzzy relationships, and construct a fusion fuzzy time cognitive graph model, including:

[0047] Analyze the temporal distribution and magnitude of each error to identify the causal relationship between variables (multi-dimensional hourly energy consumption data). For example, if the error of variable A and the error of variable B change synchronously over time, it indicates a causal relationship between A and B.

[0048] Based on the causal relationship, the corresponding first delay is selected as the delay information between each variable. Nodes and directed edges are determined, and initial values for edge weights are set to create a directed time graph. Each node represents a variable (such as elevator energy consumption or CO2 concentration). Directed edges are established based on the causal relationship and delay information. The direction of the directed edge represents the causal relationship, and its weight W is calculated using the correlation coefficient and error. The corresponding formula is: W = CCF × e-λ|k|. λ represents the delay attenuation factor, e represents the error, CCF represents the correlation coefficient, and |·| represents the absolute value.

[0049] Fuzzy relationships are established on a directed time graph to obtain a fused fuzzy temporal cognitive graph model. Trend units corresponding to the directed time graph are mapped, and the corresponding fuzzified trend strength is determined based on the change trend (trend slope) in each trend unit. Fuzzy logic mapping is performed on each fuzzified trend strength to obtain the corresponding dynamic response pattern, which is used as a causal relationship rule. The causal relationship rule is superimposed on the directed time graph to obtain a fused fuzzy temporal cognitive graph model.

[0050] When new multi-dimensional hourly energy consumption data is input into the fused fuzzy time cognitive graph model, the trend unit needs to be recalculated and the delay attenuation factor λ needs to be updated. The edge weights need to be optimized through error feedback to achieve dynamic adjustment.

[0051] The process of establishing fuzzy relationships is as follows:

[0052] 1. Map each trend unit into a fuzzy set and determine the fuzzy trend strength; the fuzzy trend strength includes [strong rise, weak rise, stable] or [strong decline, weak decline, stable]. Taking [strong rise, weak rise, stable] as an example, when the p of any trend unit is ij When the p of any trend unit is greater than 3, the fuzzy trend strength is defined as "strong rise"; ij In the range of (0.5, 3], the fuzzy trend strength is defined as "weakly rising"; when the p of any trend unit ij If it is in the range of (0, 0.5], the fuzzification trend strength is defined as “stable”.

[0053] 2. Define causal relationship rules based on fuzzy trend strength. Fuzzy trend strength (such as "strong increase" and "weak decrease") is mapped to a dynamic response pattern to the target variable (such as "medium intensity delayed increase") through fuzzy logic. The dynamic response pattern is used as a causal relationship rule and stored in the rule base. The dynamic response pattern can supplement the static association of directed graph edge weights and enhance the ability to express nonlinear relationships. For example, Rule 1 (if CO2 concentration "strongly increases", elevator energy consumption "delayed increase") can achieve adaptive correction of causal strength by dynamically adjusting the edge weight (such as W = W0*1.5).

[0054] S2-6-5. Input the multi-dimensional hourly energy consumption data into the fusion fuzzy time cognitive graph model, and output the multi-dimensional hourly energy consumption correction data.

[0055] The improved wave search algorithm (IWSA) is used to train the fusion fuzzy time cognitive graph model; Figure 2 As shown, the IWSVA algorithm includes the following processes:

[0056] T1. Set optimization parameters and contribution weights; optimization parameters include population size, maximum number of iterations, exchange period, similarity threshold, step size adjustment factor, shrinkage factor, initial step size, and weight smoothing factor. The step size adjustment factor is greater than 1, and the shrinkage factor range is (0.1). The contribution weight is the contribution of each fractal method in the current search phase, specifically reflected in the proportion of new solutions generated by each elementary low-discrepancy sequence in the next iteration. The updated weight is used to guide the allocation of subsequent search resources, ensuring that the best-performing fractal methods receive more resources, thereby exploring the search space more efficiently.

[0057] T2. Generate corresponding initial solutions using N groups of elementary low-discrepancy sequences (fractal method), obtain N groups of initial solutions, and generate corresponding candidate pools; in this embodiment, N is 4, and the elementary low-discrepancy sequences use Sobol elementary low-discrepancy sequences, Latinhypercube elementary low-discrepancy sequences, Halton elementary low-discrepancy sequences, and Hammersley elementary low-discrepancy sequences. The initial solutions in each group include the optimal memory factor, coordination coefficient, and association weight matrix of the fusion fuzzy temporal cognitive graph model. These initial solutions are evenly distributed in the high-dimensional space, improving the global exploration capability.

[0058] T3. Calculate the fitness of each initial solution by fusing the fuzzy temporal cognitive graph model, generate new solutions, and initially update each candidate pool.

[0059] The T3 includes:

[0060] T3-1. Assign the initial solutions generated by each elementary low-discrepancy sequence to the corresponding candidate pool.

[0061] T3-2. Calculate the fitness of each initial solution in each candidate pool by fusing the fuzzy temporal cognitive graph model.

[0062] T3-3. Randomly select an elementary low-discrepancy sequence.

[0063] T3-4. Select a parent solution from the candidate pool of the elementary low-discrepancy sequence; generate a new solution through the fractal strategy of the elementary low-discrepancy sequence based on the fitness and contribution weight of the initial solution.

[0064] T3-5. Perform wave propagation perturbation via the initial step size.

[0065] T3-6. Calculate the fitness of the new solution; retain the one with the highest fitness and store it in the corresponding candidate pool, completing the initial update of the candidate pool.

[0066] T3-7. Repeat T3-3 to T3-7 until all elementary low-discrepancy sequences are selected.

[0067] T4. Update contribution weights; select the solution with the highest fitness in each candidate pool and randomly insert it into other candidate pools, and update each candidate pool twice.

[0068] T5. Merge the solutions in each secondary updated candidate pool to obtain a global solution set; calculate the Euclidean distance between the solutions in the global solution set; merge or eliminate the solutions in the global solution set according to the Euclidean distances to obtain an updated global solution set.

[0069] According to each Euclidean distance, the solutions of the global solution set are merged or eliminated to obtain the updated global solution set:

[0070] If the Euclidean distance between two solutions is not less than the similarity threshold, they are considered similar solutions and merged. If the Euclidean distance is less than the similarity threshold, the solution with higher fitness is retained and the solution with lower fitness is eliminated. In addition, if the number of solutions after merging and eliminating is less than the number of initial solutions, new solutions are randomly generated to supplement them.

[0071] T6. Divide the updated global solution set equally into each of the second-updated candidate pools, calculate the fitness improvement rate of each second-updated candidate pool, and update the initial step size based on each fitness improvement rate. The fitness improvement rate is the difference between the fitness of the second-updated candidate pool and the fitness of the candidate pool before the update, divided by the fitness of the candidate pool before the update.

[0072] T7. Select the solution with the highest fitness in the updated global solution set and update the global optimal solution.

[0073] T8. Repeat T3 to T7 until the maximum number of iterations is reached or the prediction error is minimized, obtain the global optimal solution and apply it to the fusion fuzzy temporal cognitive graph model.

[0074] The improved wave search algorithm complements four fractal methods, avoids the blind spots of traditional random sampling, improves the global search capability, and automatically allocates resources according to the search stage, balancing exploration and development, thereby improving the optimization efficiency of FTCM.

[0075] S3. Perform lateral error correction on the multi-dimensional hourly energy consumption correction data to obtain multi-dimensional hourly energy consumption precision-calibrated data. Because lateral errors exist between the multi-dimensional hourly energy consumption data and the multi-dimensional hourly energy consumption correction data, lateral error correction must be performed on the multi-dimensional hourly energy consumption correction data to improve the accuracy of the operating condition classification. The lateral error refers to the difference between the multi-dimensional hourly energy consumption data and the multi-dimensional hourly energy consumption correction data on the lateral time coordinate axis.

[0076] Thus, the S3 includes:

[0077] S3-1. Based on the correlation coefficient and slope in S2, segment the multi-dimensional hourly energy consumption correction data to obtain hourly energy consumption correction segmented data.

[0078] S3-2. Calculate the mutual information of the hourly energy consumption correction segmented data and filter it to obtain hourly energy consumption correction feature data.

[0079] The S3-2 includes:

[0080] S3-2-1. Set the number of bins, discretize each hourly energy consumption correction segment data, and obtain hourly energy consumption correction discrete data. N1 is the data size of the hourly energy consumption correction segment data.

[0081] S3-2-2. Calculate the mutual information of each discrete data point in each hourly energy consumption corrected discrete data point. Use the energy consumption-related data point in each hourly energy consumption corrected discrete data point as the target variable, and the remaining data point as the input features. For example, the input features are the data corresponding to temperature and passenger flow in the hourly energy consumption corrected discrete data point, and the target variable is the data corresponding to the energy consumption of each device in the hourly energy consumption corrected discrete data point.

[0082] Mutual information I(X d ; Y d ) is:

[0083]

[0084] Where p(·) represents the marginal probability distribution function, log(·) represents the logarithmic function with a constant as the base, and X d 、Y d Represent the input features and target variables respectively, x d 、y d Represent the discrete data corresponding to the input features and target variables respectively.

[0085] Mutual information can measure the statistical dependence between two variables, which is conducive to the selection of key feature variables.

[0086] S3-2-3: Eliminate the hourly energy consumption correction discrete data that is less than the mutual information threshold to obtain hourly energy consumption correction feature data. In this embodiment, the mutual information threshold is 0.3.

[0087] S3-3. Perform Savitzky-Golay filtering and transition point detection on the hourly energy consumption correction characteristic data to obtain a set of transition point positions.

[0088] The S3-3 includes:

[0089] S3-3-1. Perform Savitzky-Golay filtering on the hourly energy consumption correction characteristic data to obtain hourly energy consumption correction filtered data to avoid the influence of noise on the data.

[0090] S3-3-2. Traverse the hourly energy consumption correction filter data and determine the transition point based on the transition point rule.

[0091] The transition point rule is that the value of the middle data point of three consecutive data points in the hourly energy consumption correction filter data is the maximum or minimum value, then the middle data point is used as the transition point, and the corresponding expression is:

[0092]

[0093] Among them, P i 、P i+1 、P i+2 Indicates the i-th data, i+1-th data, and i+2-th data in the hourly energy consumption correction filter data, Y t Indicates the transition threshold.

[0094] S3-3-3. Determine the position of each transition point and obtain a set of transition point positions.

[0095] S3-4. Divide the trend segments of the hourly energy consumption correction characteristic data based on the set of transition point positions and establish a corresponding dynamic trend model.

[0096] The S3-4 includes:

[0097] S3-4-1. Based on the set of transition point locations, divide the trend segment boundaries. Regions with consistent slope directions (positive / negative) within consecutive windows and amplitude changes that do not exceed the amplitude threshold are classified as the same trend segment.

[0098] S3-4-2. Based on the trend segment boundaries, the hourly energy consumption correction characteristic data is divided to obtain several trend segments.

[0099] S3-4-3. Establish the initial dynamic trend model of each trend segment. Taking a trend segment as an example, the corresponding initial dynamic trend model formula is: (t jump,start ,t jump,end ,p jump ,τ jump ). Among them, t jump,start , t jump,end Respectively represent the starting time of the trend segment, p jump Represents the average rate of change of the trend segment, τ jump Indicates the duration of the trend segment.

[0100] Average rate of change p jump The corresponding formula is:

[0101]

[0102] Among them, N jump Indicates the length of the trend segment, P(t jump,i+1 ), P(t jump,i ) represent the i-th moment t of the trend segment respectively. jump,i , the i+1th moment t jump,i+1 The value of the corresponding data.

[0103] S3-4-4. Use the same method from S2 to S3-4-2 to obtain the historical trend segments corresponding to the multi-dimensional hourly energy consumption historical data.

[0104] S3-4-5. Match each historical trend segment with each trend segment to obtain a matching result; update each initial dynamic trend model based on the matching result to obtain a corresponding dynamic trend model.

[0105] The process of S3-4-5 is:

[0106] Calculate the average rate of change difference and amplitude change difference between historical trend segments and trend segments. Consider historical trend segments with average rate of change difference and amplitude change difference within the threshold range as similar trend segments. Set the threshold range based on actual conditions.

[0107] Calculate the average duration of each trend segment and the corresponding similar trend segment; based on each average duration, adjust the duration τ jump Apply the adjusted duration to each initial dynamic trend model to obtain the corresponding dynamic trend model, i.e. (t jump,start ,t jump,end ,p jump ,τ current ).

[0108] Taking any initial dynamic trend model as an example, the adjusted duration τ current The corresponding formula is:

[0109]

[0110] Among them, τ hist It represents the average rate of change of the same trend segment of the initial dynamic trend model, p current 、p hist They represent the average change rate of the trend segment of the initial dynamic trend model and the same trend segment, ΔP hist , ΔP current They respectively represent the same type trend segments and the amplitude changes of trend segments of the initial dynamic trend model.

[0111] S3-5. Perform lateral error correction on the multi-dimensional hourly energy consumption correction data through a dynamic trend model to obtain multi-dimensional hourly energy consumption precision-calibrated data.

[0112] The S3-5 includes:

[0113] S3-5-1. Calculate the state transition time difference set of the multi-dimensional hourly energy consumption correction data and the multi-dimensional hourly energy consumption data; the formula corresponding to the state transition time difference is:

[0114]

[0115] Where, Δt i It represents the state transition time difference between the i-th multi-dimensional hourly energy consumption correction data and the corresponding multi-dimensional hourly energy consumption data, t jump,pre,i Indicates the time corresponding to the jump position of the i-th multi-dimensional hourly energy consumption correction data, t jump,real,i Indicates the time corresponding to the jump position of the i-th multi-dimensional hourly energy consumption data. i If Δt is greater than 0, the relationship between the multi-dimensional hourly energy consumption correction data and the multi-dimensional hourly energy consumption data is a prediction lag; when Δt i If it is less than 0, the relationship between the multi-dimensional hourly energy consumption correction data and the multi-dimensional hourly energy consumption data is prediction ahead.

[0116] S3-5-2. Fit the probability distribution of the state transition time difference set using the kernel density estimation method, and calculate the corresponding deviation mean and confidence interval.

[0117] S3-5-3. Shift the multidimensional hourly energy consumption data along the time axis by the deviation mean value so that the transition points of the multidimensional hourly energy consumption correction data align with the transition points of the multidimensional hourly energy consumption data, thereby obtaining updated multidimensional hourly energy consumption data. If data breakpoints occur after the shift, cubic spline interpolation is used to fill in the missing values.

[0118] S3-5-4. Based on the state transition time difference set and confidence interval, update the parameters of the dynamic trend model, that is, update the adjusted duration τ current .

[0119] When Δt i If it is greater than 0, τ needs to be shortened current , so that τ current Closer to the measured jump point; when Δt i If it is less than 0, τ needs to be extended current , to match the measured data.

[0120] τ current The update formula is:

[0121] τ new =ατ current+(1-α)τ avg ;

[0122] Among them, τ new represents the updated duration; α represents the weight coefficient, which is determined by the confidence interval of Δt or the distribution characteristics of the kernel density estimate.

[0123] S3-5-5. Based on the parameters of the updated dynamic trend model, the updated multi-dimensional hourly energy consumption data is corrected to obtain the multi-dimensional hourly energy consumption precision-calibrated data.

[0124] S4. Use the TICC algorithm (inverse covariance clustering algorithm) to divide the multi-dimensional hourly energy consumption precision calibration data to obtain the working condition division results.

[0125] The S4 includes:

[0126] S4-1. Set a time window and divide the multi-dimensional hourly energy consumption data into time windows to obtain the corresponding hourly energy consumption window data. The time window length is W. Divide the multi-dimensional hourly energy consumption data into T1-W+1 overlapping windows. Set the sliding step size of adjacent windows to 1 hour to ensure temporal consistency and avoid missing gradual changes in transition periods. T1 is the sequence length of the multi-dimensional hourly energy consumption data.

[0127] S4-2. Set TICC parameters. TICC parameters include the number of clusters K and the sparsity parameter λ. s And the maximum number of iterations N TICC λ s The value is 0.1, N TICC The value is 500 and the elbow method is used to determine the number of clusters K, that is, calculate the within-cluster sum of squares (WCSS) and find the critical point (elbow) where the downward trend of WCSS changes from steep to gentle. The value corresponding to this critical point is used as the number of clusters K.

[0128] Calculate the intra-cluster sum of squares WCSS(k TICC )The corresponding formula is:

[0129]

[0130] Among them, k TICC Indicates the number of clusters, C i represents the hourly energy consumption window data set of the i-th cluster, μ i represents the centroid of the ith cluster, ||·|| represents the norm, and x TICC Indicates hourly energy consumption window data.

[0131] S4-3. Calculate the inverse covariance matrix of each hourly energy consumption window data. First, calculate the covariance matrix of each hourly energy consumption window data; invert each covariance matrix to obtain the inverse covariance matrix. The formula corresponding to the covariance matrix is:

[0132]

[0133] Among them, M1 represents the number of samples in the time window, that is, the number of samples of the hourly energy consumption window data, x mi 、x mj Respectively represent the values of the i-th dimension and the j-th dimension of the m-th hourly energy consumption window data, Respectively represent the mean of the i-th dimension and the j-th dimension, σ ij Represents the covariance matrix corresponding to the i-th dimension and the j-th dimension of the hourly energy consumption window data.

[0134] The diagonal elements and off-diagonal elements in each covariance matrix reflect the variability and correlation of the covariance matrix, respectively.

[0135] S4-4, with the goal of minimizing the inverse covariance matrix, use the TICC algorithm to iterate the inverse covariance matrix, alternating between E-step cluster assignment and M-step inverse covariance matrix update until convergence or the maximum number of iterations N is reached TICC , and get the clustering results.

[0136] In the E step, for each time window, the cost of belonging to the cluster is calculated, and the optimal cluster sequence is found using dynamic programming (Viterbi algorithm) to minimize the total cost and the smoothing term.

[0137] The M step updates the inverse covariance matrix. For each cluster, the covariance matrices of all windows belonging to the cluster are collected and the mean is calculated to represent the global characteristics of the cluster. The matrix is input into the Graphical LASSO algorithm, and the weak correlation is eliminated through L1 regularization to generate a sparse inverse covariance matrix.

[0138] S4-5. Calculate objective indicators of the clustering results; based on these objective indicators, evaluate the clustering results using the Analytic Hierarchy Process (AHP) and the Entropy Weight Method (EWM) to obtain evaluation results. Objective indicators include the Silhouette Coefficient, the Calinski-Harabasz Index, and the Davies-Bouldin Coefficient.

[0139] The S4-5 includes:

[0140] S4-5-1. Use silhouette coefficient, Calinski-Harabasz index and Davies-Bouldin coefficient to perform multi-dimensional evaluation on the clustering results to obtain multi-dimensional evaluation results.

[0141] The entropy weighting method was used to assign weights to the three evaluation indicators, and the multi-dimensional evaluation results were calculated based on the weights. The silhouette coefficient quantifies the compactness and separation of clusters by calculating the average distance between a sample point and other data points in the same cluster and the minimum average distance to the nearest cluster, with a value range of [-1, 1]. The Calinski-Harabasz index reflects the balance between inter-cluster separation and intra-cluster compactness by calculating the ratio of inter-cluster dispersion to intra-cluster dispersion. The Davies-Bouldin index measures inter-cluster differentiation based on the ratio of the average intra-cluster distance to the cluster center distance. Smaller values indicate better clustering results.

[0142] S4-5-2. Using the analytic hierarchy process and based on the operators' professional knowledge of energy consumption conditions, a judgment matrix is constructed and the subjective weights of the three indicators are calculated to obtain the subjective evaluation results.

[0143] S4-5-3. Based on the multi-dimensional evaluation results and the subjective evaluation results, a weighted average is calculated to obtain an evaluation result. In this embodiment, the subjective weight is 0.4 and the objective weight is 0.6.

[0144] S4-6. Determine the working condition classification result based on the evaluation results. Select the cluster result corresponding to the evaluation result with the highest comprehensive coefficient as the working condition classification result.

[0145] In summary, the present invention uses the cross-correlation function (CCF) and the fused fuzzy time cognitive map model (FTCM) to dynamically analyze the time delay correlation between multi-dimensional data (such as carbon dioxide concentration, equipment energy consumption, passenger flow, temperature), and adjusts the data scanning window according to the characteristics of subway operation, corrects the data time offset, and improves data synchronization and reliability; adopts the sliding window method and the dynamic adjustment window width strategy to optimize the scanning range according to the data fluctuation and correlation results, enhances the adaptability of the model to different operating periods, and ensures the accuracy and robustness of data correction; through Savitzky-Golay filtering, jump point detection and dynamic A dynamic trend model is used to eliminate lateral errors in multi-dimensional data on the time axis, align key transition points, and provide highly consistent precision-calibrated data for subsequent working condition division. The TICC algorithm (inverse covariance clustering) is used to perform multi-dimensional time series clustering on the corrected data. The clustering results are optimized by combining the analytic hierarchy process and the entropy weight method to accurately identify typical energy consumption conditions in subway stations and improve energy utilization efficiency. The wave search algorithm is improved. Through multi-fractal strategies and adaptive resource allocation, the training process of the fusion fuzzy temporal cognitive graph model is accelerated, balancing global exploration and local development capabilities to ensure the model's rapid convergence and optimal solution acquisition in complex high-dimensional data.

[0146] like Figure 3 As shown, a system for finely dividing energy consumption conditions of subway stations includes:

[0147] Multi-dimensional hourly energy consumption collection module, used to obtain multi-dimensional hourly energy consumption data of a subway station;

[0148] A data correction module is used to correct the multi-dimensional hourly energy consumption data using a cross-correlation function and a fusion fuzzy time cognitive graph model to obtain multi-dimensional hourly energy consumption correction data;

[0149] The data precision calibration module is used to perform lateral error correction on the multi-dimensional hourly energy consumption correction data to obtain multi-dimensional hourly energy consumption precision calibration data;

[0150] The working condition division module is used to divide the multi-dimensional hourly energy consumption precision calibration data using the TICC algorithm to obtain the working condition division results.

[0151] In summary, this system designs modules corresponding to data acquisition, correction, fine calibration, and working condition division, supports flexible expansion to different subway stations, adapts to diverse operating environments and equipment configurations, and provides a universal solution for intelligent energy-saving management of urban rail transit.

[0152] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A method for finely dividing energy consumption conditions of subway stations, characterized in that: include: Obtaining multi-dimensional hourly energy consumption data of a subway station; the multi-dimensional hourly energy consumption data includes hourly carbon dioxide sensor data, hourly equipment energy consumption data, hourly passenger flow data, and hourly temperature data; The multi-dimensional hourly energy consumption data is corrected using the cross-correlation function and the fusion fuzzy time cognitive graph model to obtain the multi-dimensional hourly energy consumption correction data. Perform lateral error correction on the multi-dimensional hourly energy consumption correction data to obtain multi-dimensional hourly energy consumption precision calibration data; The TICC algorithm is used to divide the multi-dimensional hourly energy consumption precision calibration data to obtain the working condition division results.

2. The method for finely dividing the energy consumption conditions of a subway station according to claim 1 is characterized in that: The data correction process includes: Calculate the time delay between any two data types in multi-dimensional hourly energy consumption data; The correlation coefficient of any two data types in the multi-dimensional hourly energy consumption data at different time delays is calculated using the cross-correlation function; Sort the correlation coefficients of any two data types at different delays from large to small, select the maximum value of the corresponding correlation coefficient as the correlation result corresponding to the two data types, and use the delay corresponding to each correlation result as the delay estimate; Based on the operational characteristics of the subway station, the correlation results and their estimated delays, the data scanning range is determined and the window width is dynamically adjusted. Based on the data scanning range, the sliding window method is used to calculate the first time delay between the multi-dimensional hourly energy consumption data in the window; Based on the first time delay and its correlation coefficient, the multi-dimensional hourly energy consumption data is corrected using a fusion fuzzy time cognitive graph model to obtain multi-dimensional hourly energy consumption correction data.

3. The method for finely dividing the energy consumption conditions of a subway station according to claim 2 is characterized in that: The method of performing data correction on the multi-dimensional hourly energy consumption data based on the first time delay and its correlation coefficient by using a fusion fuzzy time cognitive graph model to obtain multi-dimensional hourly energy consumption correction data includes: Perform trend segmentation on multi-dimensional hourly energy consumption data to obtain trend units; Determine the linear relationship corresponding to each trend unit; calculate the predicted value of each trend unit based on each linear relationship; Calculate the error between each predicted value and the multi-dimensional hourly energy consumption data at the same time; Based on the first time delay and its correlation coefficient and each error, nodes, directed edges, edge weights and fuzzy relationships are determined to construct a fusion fuzzy time cognitive graph model; The multi-dimensional hourly energy consumption data is input into the fusion fuzzy time cognitive graph model, and the multi-dimensional hourly energy consumption correction data is output.

4. The method for finely dividing the energy consumption conditions of a subway station according to claim 2 is characterized in that: The improved wave search algorithm is used to train the fusion fuzzy time cognitive graph model. The following processes are included: T1. Set optimization parameters and contribution weights; T2. Generate corresponding initial solutions using N groups of elementary low-discrepancy sequences, obtain N groups of initial solutions, and generate corresponding candidate pools; T3. Calculate the fitness of each initial solution by integrating the fuzzy temporal cognitive graph model, generate new solutions, and initially update each candidate pool; T4. Update contribution weights; select the solution with the highest fitness in each candidate pool and randomly insert it into other candidate pools, and update each candidate pool twice; T5. Merge the solutions in each secondarily updated candidate pool to obtain a global solution set and update it to obtain an updated global solution set; T6. Divide the updated global solution set equally into each secondary updated candidate pool, calculate the fitness improvement rate of each secondary updated candidate pool; based on each fitness improvement rate, update the initial step size; T7. Select the solution with the highest fitness in the updated global solution set and update the global optimal solution; T8. Repeat T3 to T7 until the maximum number of iterations is reached or the prediction error is minimized, obtain the global optimal solution and apply it to the fusion fuzzy temporal cognitive graph model.

5. The method for finely dividing the energy consumption conditions of a subway station according to claim 1 is characterized in that: The process of lateral error correction is: Segmenting the multi-dimensional hourly energy consumption correction data to obtain hourly energy consumption correction segmented data; Calculate the mutual information of the hourly energy consumption correction segmented data and filter it to obtain hourly energy consumption correction feature data; Perform Savitzky-Golay filtering and transition point detection on the hourly energy consumption correction feature data to obtain a set of transition point positions; Divide the trend segments of hourly energy consumption correction characteristic data based on the jump point position set and establish a corresponding dynamic trend model; The multi-dimensional hourly energy consumption correction data is corrected for horizontal errors through a dynamic trend model to obtain multi-dimensional hourly energy consumption precision-calibrated data.

6. The method for finely dividing the energy consumption conditions of a subway station according to claim 5 is characterized in that: Dividing the trend segments of hourly energy consumption correction characteristic data based on the jump point position set and establishing the corresponding dynamic trend model includes: Divide the trend segment boundaries based on the set of jump point positions; Based on the trend segment boundaries, the hourly energy consumption correction characteristic data is divided to obtain several trend segments; Establish the initial dynamic trend model of each trend segment; Obtain historical trend segments corresponding to multi-dimensional hourly energy consumption historical data; Match each historical trend segment with each trend segment to obtain a matching result; update each initial dynamic trend model based on the matching result to obtain a corresponding dynamic trend model.

7. The method for finely dividing the energy consumption conditions of a subway station according to claim 6 is characterized in that: The dynamic trend model is used to perform horizontal error correction on the multi-dimensional hourly energy consumption correction data, and the multi-dimensional hourly energy consumption precision correction data obtained includes: Calculating a state transition time difference set of multi-dimensional hourly energy consumption correction data and multi-dimensional hourly energy consumption data; The probability distribution of the state transition time difference set is fitted by the kernel density estimation method, and the corresponding deviation mean and confidence interval are calculated; Shifting the multi-dimensional hourly energy consumption data along the time axis by a deviation mean value so that the transition point of the multi-dimensional hourly energy consumption correction data is aligned with the transition point of the multi-dimensional hourly energy consumption data, thereby obtaining updated multi-dimensional hourly energy consumption data; Update the parameters of the dynamic trend model based on the state transition time difference set and confidence interval; Based on the updated parameters of the dynamic trend model, the updated multi-dimensional hourly energy consumption data is corrected to obtain the multi-dimensional hourly energy consumption precision-calibrated data.

8. The method for finely dividing the energy consumption conditions of a subway station according to claim 5 is characterized in that: The TICC algorithm is used to divide the multi-dimensional hourly energy consumption precision calibration data to obtain the working condition division results, including: Set the time window and divide the multi-dimensional hourly energy consumption precision calibration data into time windows to obtain the corresponding hourly energy consumption window data; Set TICC parameters; TICC parameters include the number of clusters, sparsity parameters, and maximum number of iterations; Calculate the inverse covariance matrix of each hourly energy consumption window data; With the goal of minimizing the inverse covariance matrix, the TICC algorithm is used to iterate the inverse covariance matrix to obtain the clustering results; Calculate the objective indicators of clustering results; based on the objective indicators, evaluate the clustering results through the hierarchical analysis method and entropy weight method to obtain the evaluation results; Based on the evaluation results, the working condition division results are determined.

9. A system for finely dividing energy consumption conditions of a subway station, used to implement the method for finely dividing energy consumption conditions of a subway station according to any one of claims 1 to 8, characterized in that: include: Multi-dimensional hourly energy consumption collection module, used to obtain multi-dimensional hourly energy consumption data of a subway station; A data correction module is used to correct the multi-dimensional hourly energy consumption data using a cross-correlation function and a fusion fuzzy time cognitive graph model to obtain multi-dimensional hourly energy consumption correction data; The data precision calibration module is used to perform lateral error correction on the multi-dimensional hourly energy consumption correction data to obtain multi-dimensional hourly energy consumption precision calibration data; The working condition division module is used to divide the multi-dimensional hourly energy consumption precision calibration data using the TICC algorithm to obtain the working condition division results.

Images