Building energy consumption management system and method based on big data

Abstract

The invention relates to the technical field of intelligent buildings, in particular to a building energy consumption management system and method based on big data. The method specifically comprises the steps of collecting multi-source energy consumption data of a building; carrying out local data cleaning by adopting a sliding window, a robust mean value and an absolute median difference statistical method, eliminating abnormal value interference in combination with dynamic threshold evaluation and a self-adaptive substitution mechanism, and generating a multi-dimensional energy consumption feature; multi-dimensional energy state anchor points are generated through finite field coding to realize data stability verification; a self-attention mechanism and residual time sequence modeling are fused to construct an energy consumption model, and a historical clustering center is introduced to carry out dynamic prediction correction, so that the generalization ability of the model is improved; and on the basis of the normalized deviation of a predicted value and an actual observed value, an abnormal level is judged by adopting a sliding window standard deviation and a multi-level threshold value, and a continuous triggering mechanism linkage alarm strategy is combined to realize energy consumption abnormal detection and hierarchical response. According to the invention, a closed-loop management process of data acquisition, processing, prediction, optimization and feedback is formed.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent buildings, and specifically relates to a building energy consumption management system and method based on big data. Background Art

[0002] With the acceleration of the urbanization process and the popularization of the concept of green buildings, building energy consumption management has become an important link in achieving the goal of energy conservation and carbon reduction; traditional building energy consumption monitoring systems mostly rely on single-sensor data and adopt an offline batch processing method, which has problems such as high data processing latency, insufficient anomaly detection sensitivity, and weak generalization ability of prediction models, and it is difficult to adapt to complex and changeable building energy consumption scenarios.

[0003] The Chinese invention patent with the publication number CN118211721B discloses an artificial intelligence-based building energy consumption prediction method, which uses temperature sensors and humidity sensors and inputs energy bills through raw data collection sites, and connects to an energy supply management system and a meteorological site to obtain building energy consumption-related data, which helps to improve the accuracy and comprehensiveness of the building energy consumption prediction model, detects and processes building energy consumption-related data, associates the data with timestamps and extracts the average time characteristics of the timestamp data, constructs a linear regression model and inputs the average time characteristics of temperature data, humidity data, and building characteristic data, constructs a decision tree model and inputs the characteristic data of the building, the historical energy consumption data of the building, and the climate data of the area where the building is located, creates a visualization interface and displays the final prediction results, performance indicators, and actual energy consumption data after integrating the two models in real time, helping users intuitively understand the building energy consumption situation.

[0004] In the prior art, the energy consumption analysis method based on cloud centralized computing can process a large amount of data, but due to the limitations of network transmission bandwidth and cloud computing power, the real-time performance and edge response efficiency are relatively low; and the conventional threshold warning mechanism is easily interfered by instantaneous fluctuations and has a high false alarm rate, making it difficult to support accurate energy consumption diagnosis and dynamic optimization; in addition, building energy consumption is affected by the coupling of multiple factors such as regional environment, equipment operation status, and user behavior, and traditional time series prediction models often ignore the multi-dimensional feature correlation, resulting in limited prediction accuracy and difficulty in meeting the requirements of refined management; in this context, there is an urgent need for an energy consumption management system that integrates edge intelligence, big data analysis, and a dynamic correction mechanism to achieve closed-loop management of real-time data collection, efficient preprocessing, accurate prediction, and intelligent decision-making, providing reliable technical support for building energy conservation optimization. Summary of the Invention

[0005] The object of the present invention is to address the problems in the background art and propose a building energy consumption management system and method based on big data.

[0006] The technical solution of the present invention: A building energy consumption management method based on big data includes the following specific implementation steps:

[0007] S1. Collect the energy consumption data in the building;

[0008] S2. Construct a window sequence for the energy consumption data through a sliding window, calculate the robust mean and absolute median difference to eliminate the interference of extreme values, evaluate data anomalies based on the normalized deviation and perform adaptive substitution using an exponentially decaying weight, dynamically update the statistics and window parameters to maintain the adaptive ability, generate multi-dimensional energy consumption features using z-score normalization, and construct multi-dimensional energy state anchors by combining finite field coding and a preset hash function;

[0009] S3. Generate finite field coded data through the multi-dimensional energy state anchors and multi-dimensional energy consumption features and solve the linear equation to achieve data stability verification, construct an energy consumption prediction model, based on the time series feature matrix, use multi-scale residual convolution and self-attention mechanism to extract local trends and long-term dependence relationships, after outputting the basic prediction value through a multi-layer perceptron, dynamically weighted fusion is combined with the historical clustering center to correct the prediction deviation and generate the energy consumption prediction value;

[0010] S4. Generate a standardized error index by calculating the absolute error between the predicted value and the actual energy consumption value and performing normalization processing based on the standard deviation of the sliding window, divide the anomaly levels, based on the continuous anomaly trigger determination mechanism within the sliding window, determine the anomaly persistence according to the effective trigger rate threshold, and trigger a hierarchical alarm strategy according to the anomaly levels;

[0011] S5. Visualize and display the energy consumption data, energy consumption prediction value, anomaly level, anomaly persistence, and alarm strategy.

[0012] Preferably, the adaptive substitution process is as follows:

[0013] S21. Based on the energy consumption data S = {s1, s2,..., s i ,…, s t}, construct a window sequence of length w in a sliding window manner: W t = {s t-w+1 , s t-w+2 ,…, s t};

[0014] wherein, s i represents the original energy consumption value collected at the i-th moment; w represents the window length; W t represents the continuous observation data within the current window;

[0015] S22. Calculate the robust statistical mean and local fluctuation amplitude:

[0016] σ t = median(|s′ i - μ t |); s′i ∈W′ t ;

[0017] wherein, W t ' represents the data after removing the upper and lower 10% extreme values of W t ; s′ i represents the i-th data value in W t '; μ t represents the robust mean; σ t represents the absolute median deviation based on μ t ; median() represents the median function;

[0018] S23. Evaluate the deviation degree of the current value s t from the robust mean:

[0019] Calculate the discriminant function:

[0020] wherein, δ t represents the normalized deviation; ε represents a small constant to prevent division by zero; θ t represents the dynamic threshold; k represents the empirical coefficient; Var(W t ) represents the variance within the current window; accept indicates that the current data s t has no abnormality; filter indicates that the current data s t is determined to be abnormal;

[0021] S24. If the current data s t is determined to be abnormal, i.e., filter, then perform adaptive substitution:

[0022]

[0023] wherein, exp() represents the natural exponential function; represents the substitution value; w i represents the decay weight assigned to the time distance; α represents the decay speed control factor.;

[0024] Preferably, the construction process of the multi-dimensional energy state anchor is as follows:

[0025] S31. Encode the multi-dimensional energy consumption feature x i to obtain the encoded data Dc = H(data) ∈ GF(p) n ;

[0026] wherein, data is the multi-dimensional energy consumption feature x iThe binary-encoded data; H() is a pre-set hash function with an output length of n×log2p; p is a pre-defined large prime number satisfying p≡3mod 4; n is the set number of variables; GF(p) n represents an n-dimensional vector space composed of elements {0, 1, 2, …, p−1} in the finite field GF(p);

[0027] S32. Define the linear equation S·EⅠ = Dc, and solve to obtain the first-order encoding tuple EⅠ = S -1 ·Dc = {z1, z2, …, z i , …, z n};

[0028] where S is a pre-set invertible n×n matrix over the finite field GF(p); z i is any solution of the equation S·EⅠ = Dc;

[0029] S33. Calculate the second-order encoding tuple EⅡ = {y1, y2, …, y i , …, y n};

[0030] where y i is any element in the second-order encoding tuple EⅡ; c i is an element in the pre-defined constant vector c = [c1, c2, …, c i , …, c n , and c ∈ GF(p) n ;

[0031] S34. Define the linear equation B·A = EⅡ, and solve to obtain the multi-dimensional energy state anchor A = B -1 ·EⅡ;

[0032] where B is a pre-set invertible n×n matrix over the finite field GF(p).

[0033] Preferably, the verification process for realizing data stability verification is as follows:

[0034] S41. Calculate the decoding tuple D = B·A = {d1, d2, …, d i , …, d n};

[0035] where d i is any element in the decoding tuple D;

[0036] S42. Calculate the decoding factor fd:

[0037] S43. Calculate the verification factor fc = S·fd;

[0038] S43. Calculate the secondary encoded data Dc′ = H(data′);

[0039] where data′ is the binary encoded data of the received multi-dimensional energy consumption feature x i ;

[0040] S44. If fc = Dc′, the verification passes, that is, the received multi-dimensional energy consumption feature x i has stability and the data positioning has reliability; otherwise, an alarm is immediately given.

[0041] Preferably, the model architecture of the energy consumption prediction model includes:

[0042] Input layer: Construct a time series feature matrix, and splice the multi-dimensional feature vectors of N consecutive time steps within the sliding time window into a two-dimensional input matrix X t :

[0043] where x t represents the multi-dimensional energy consumption feature vector extracted at the t-th moment; T represents the transpose operation of the matrix; represents a real number matrix with N rows and d columns; d represents the feature dimension at each moment;

[0044] Time series feature residual modeling layer: Introduce multi-scale convolution and residual connection, use one-dimensional convolution to perform multi-scale processing on the input time series, and combine the features of different convolution scales with the residual network: H1 = ReLU(Conv k'＝3 (X t )) + Conv k'＝5 (X t );

[0045] where H1 represents the feature representation after convolution and activation; d′ represents the output dimension after convolution; Conv k' represents the one-dimensional convolution operation; k′ represents the size of the convolution kernel; ReLU() represents the ReLU activation function;

[0046] Key dependence extraction layer: Introduce the self-attention mechanism and output the feature representation H2:

[0047]

[0048] Q = H1W Q ; K = H1W K ; V = H1W V ;

[0049] where Attention(Q, K, V) represents the self-attention operation; Q, K, V represent the query, key, and value matrices obtained through linear transformation; WQ , W K , W V represents the weight matrix; d k represents the dimension of the query vector; softmax() represents the softmax function;

[0050] Prediction layer: Use a multi-layer perceptron for regression to initially predict the energy consumption value at the next moment

[0051] Among them, represents the feature representation at time step t obtained from the self-attention mechanism;

[0052] Dynamic prediction correction layer: Combine the historical clustering centers to dynamically weight and fuse to correct the prediction deviation and generate the energy consumption prediction value.

[0053] Preferably, the generation process of combining the historical clustering centers to dynamically weight and fuse to correct the prediction deviation and generate the energy consumption prediction value is as follows:

[0054] S61. Construct a historical feature clustering dictionary: Through clustering analysis of the energy consumption characteristics of a large number of historical buildings, form a set of typical energy consumption pattern vectors, that is, clustering centers:

[0055] Among them, C j represents the center vector of the j-th type of energy consumption pattern; N j represents the number of samples in the j-th class; represents the sample belonging to the j-th class; M represents the total number of energy consumption pattern categories;

[0056] S62. Calculate the similarity between the current feature and the historical features: Match the feature vector after modeling at the current time step with each historical pattern:

[0057] Among them, j * represents the closest historical feature center; ||·||2 represents the Euclidean distance;

[0058] S63. Dynamic prediction fusion:

[0059] Among them, represents the basic prediction value; represents the average energy consumption value corresponding to the historical clustering center j * ; λ represents the weighting coefficient, λ ∈ [0, 1]; represents the final prediction value after fusion and correction.

[0060] Preferably, the process of abnormal level division is as follows:

[0061] S71. Dynamic deviation calculation and normalization processing: At each time step t+1, record the actual energy consumption value and compare it with the predicted value to obtain the absolute error ε t+1 :

[0062] S72. Introduce moving standard deviation normalization processing:

[0063] where e t+1 represents the normalized error index; σ [t-W,t] represents the standard deviation of the actual energy consumption in the past W time steps;

[0064] S73. Based on the normalized error index e t+1 , define the anomaly discrimination as a multi-level rule:

[0065]

[0066] where θ1 and θ1 are the dynamic thresholds set using the historical 95% and 99% quantiles respectively; Level(t+1) is the anomaly level at time t+1.

[0067] Preferably, the continuous anomaly trigger determination mechanism based on the sliding window is:

[0068] Define the anomaly level sequence in the past M time steps as {Level(t+1-M+1),...,Level(t+1)}. If the following conditions are met, the anomaly is considered valid:

[0069]

[0070] where I() represents the indicator function; γ represents the effective trigger rate threshold, γ∈[0,1]; M represents the sliding window width.

[0071] The technical solution of the present invention: A building energy consumption management system based on big data, which is used to execute the above-mentioned building energy consumption management method based on big data, including:

[0072] A multi-source energy consumption data acquisition module, which is used to build a distributed sensor network inside the building and collect building energy consumption data;

[0073] An edge data processing module, which is used to perform preliminary data filtering locally;

[0074] An intelligent analysis and prediction module, which is used to build an energy consumption prediction model using a big data processing engine and output energy consumption prediction values;

[0075] An intelligent optimization decision-making module, which is used to formulate an alarm strategy based on the energy consumption prediction value

[0076] The visualization display and alarm module is used to visually display data and give alarms in a timely manner.

[0077] Compared with the prior art, the above technical solution of the present invention has the following beneficial technical effects:

[0078] The present invention designs a building energy consumption management system and method based on big data. Through a distributed sensor network and an edge computing module, it realizes the real-time collection and local preprocessing of multi-source energy consumption data at the minute level. It uses a sliding window robust mean, absolute median difference statistics, and an adaptive substitution mechanism to eliminate noise and outliers, significantly improving data quality and processing efficiency. It integrates a self-attention mechanism and a residual time series modeling technique, combines finite field coding to generate multi-dimensional energy state anchor points for data stability verification, and dynamically corrects predicted values through historical clustering centers, effectively enhancing the model's generalization ability and prediction accuracy for complex scenarios. Based on a normalized deviation and a multi-level threshold determination mechanism, combined with sliding window statistics and a persistent triggering rule, it realizes a high-sensitivity abnormal grading alarm and a linkage control strategy, significantly reducing the false alarm rate. The present invention forms a closed-loop management process through multi-dimensional feature integration and visual feedback, providing reliable technical support for building energy consumption dynamic monitoring, abnormal diagnosis, and energy-saving optimization, and overall improving the real-time, robust, and intelligent level of energy consumption management. Description of the Drawings

[0079] Figure 1 It is a system architecture diagram of a building energy consumption management system based on big data proposed by the present invention;

[0080] Figure 2 It is a method flow diagram of a building energy consumption management method based on big data proposed by the present invention. Detailed Embodiments

[0081] Example 1, as Figure 1 shown, a building energy consumption management system based on big data proposed by the present invention includes: a multi-source energy consumption data collection module, an edge data processing module, an intelligent analysis and prediction module, an intelligent optimization decision module, and a visualization display and alarm module.

[0082] The multi-source energy consumption data collection module constructs a distributed sensor network inside the building to collect building energy consumption data;

[0083] The edge data processing module performs preliminary data filtering locally (including but not limited to removing noise, duplicate values, and obvious outliers);

[0084] The intelligent analysis and prediction module uses a big data processing engine to build an energy consumption prediction model and outputs energy consumption prediction values;

[0085] The intelligent optimization decision-making module formulates an alarm strategy based on the predicted energy consumption value;

[0086] The visualization display and alarm module visually displays the data and gives an alarm in a timely manner.

[0087] Example 2, as Figure 2 shown, a building energy consumption management method based on big data proposed by the present invention is applied to a building energy consumption management system proposed in Example 1, and its specific implementation steps are as follows:

[0088] S1. The multi-source energy consumption data acquisition module uses a variety of sensor nodes (including but not limited to electricity, water meters, gas meters, heating and cooling air-conditioning load sensors) deployed inside the building to collect real-time energy consumption data in different areas of the building at a minute-level frequency, and transmits the energy consumption data to the edge data processing module;

[0089] S2. The edge data processing module combines weighted window screening, robust statistical detection and edge intelligent judgment mechanisms to achieve high-efficiency, low-power, intelligent local preprocessing, filter the energy consumption data, and provide an accurate and structured original data stream for data processing. The specific implementation process is as follows:

[0090] S21. Construct a window sequence of length w for the energy consumption data stream S = {s1, s2,..., s i ,…, s t} in a sliding window manner: W t = {s t-w+1 , s t-w+2 ,…, s t};

[0091] where s i represents the original energy consumption value collected at the i-th moment; w represents the window length; W t represents the continuous observation data within the current window;

[0092] S22. Calculate the robust statistical mean and local fluctuation amplitude, and introduce the trimmed mean and median absolute deviation (MAD) as the reference metrics to avoid the interference of extreme values:

[0093]

[0094] σ t = median(|s′ i - μ t |);

[0095] s′ i ∈ W t ′;

[0096] where Wt ′ represents the data after removing the upper and lower 10% extreme values (in this embodiment, the largest and smallest 5% are removed); s′ t represents the i-th data value in W′; μ i represents the robust mean (a mean that is less sensitive to outliers); σ t represents the median absolute deviation (MAD) based on μ t for measuring the amplitude of fluctuations; median() represents the median function; t represents the data after removing the upper and lower 10% extreme values (in this embodiment, the largest and smallest 5% are removed); s′ t represents the i-th data value in W′; μ

[0097] S23. Evaluate the degree of deviation of the current value s t from the robust mean:

[0098] Introduce a discrimination function:

[0099] θ t = k·log(1 + Var(W t ));

[0100] where δ t represents the normalized deviation; ε represents a small constant to prevent division by zero; θ t represents the dynamic threshold; k represents an empirical coefficient; Var(W t ) represents the variance within the current window; accept indicates that the current data s t has no outliers; filter indicates that the current data s t is determined to be an outlier;

[0101] S24. If the current data s t is determined to be an outlier (i.e., filter), then perform adaptive substitution:

[0102]

[0103] w i = exp(-α·i);

[0104] where exp() represents the natural exponential function; represents the substitution value; w i represents the decay weight assigned to the time distance; α represents the decay rate control factor;

[0105] S25. Update the sliding window to W t+1 and dynamically refresh the statistical quantities robust mean μ t+1 , median absolute deviation σ t+1 , dynamic threshold θ t+1 , to ensure adaptability under different time periods and regional conditions, and continuously execute in a rolling manner;

[0106] S26. Perform a data normalization procedure using the z - score normalization method and generate a data frame in a unified format. Based on this, construct the multi - dimensional energy consumption feature x i = [t i , z i , T i , H i , U i , O i ;

[0107] Among them, t i represents the timestamp; z i represents the spatial location information (floor / area code); T i , H i represent the ambient temperature and humidity respectively; U i represents the regional energy consumption of area z i ; O i represents the user activity status (including but not limited to the number of devices turned on, the number of people);

[0108] S27. Generate multi - dimensional energy state anchor points for the multi - dimensional energy consumption feature x i . The generation process is as follows:

[0109] S2701. Encode the multi - dimensional energy consumption feature x i to obtain the encoded data Dc = H(data) ∈ GF(p) n ;

[0110] Among them, data is the binary encoded data of the multi - dimensional energy consumption feature x i ; H() is a preset hash function with an output length of n×log2p; p is a predefined large prime number satisfying p≡3 mod 4; n is the set number of variables; GF(p) n represents an n - dimensional vector space composed of elements {0, 1, 2, …, p - 1} in the finite field (Galois Field) GF(p);

[0111] S2702. Define the linear equation S·EⅠ = Dc and solve it to obtain the first - order encoded tuple EⅠ = S -1 ·Dc = {z1, z2, …, z i , …, z n};

[0112] Among them, S is a preset invertible n×n matrix in the finite field GF(p); z i is any solution of the equation S·EⅠ = Dc;

[0113] S2703. Calculate the second - order encoded tuple EⅡ = {y1, y2, …, yi , …, y n};

[0114]

[0115] Among them, y i is any element in the second-order coding tuple EⅡ; c i is an element in the predefined constant vector c = [c1, c2, …, c i , …, c n , c ∈ GF(p) n ;

[0116] S2704. Define the linear equation B·A = EⅡ, and solve to obtain the multi-dimensional energy state anchor point A = B -1 ·EⅡ;

[0117] Among them, B is a preset invertible n×n matrix over the finite field GF(p);

[0118] S28. Transmit {the multi-dimensional energy state anchor point A, the multi-dimensional energy consumption feature x i} to the intelligent analysis and prediction module;

[0119] S3. The intelligent analysis and prediction module realizes the efficient processing of data and the scheduling of model tasks based on the big data engine (including but not limited to Spark or Flink). Specifically, after aggregating the data feature vectors x i from the edge devices, under the control of the big data engine, schedule and execute the constructed energy consumption prediction model (fusing the self-attention mechanism and the residual time series modeling) to achieve high-precision prediction and anomaly warning of large-scale building energy consumption;

[0120] To ensure the stability of the multi-dimensional energy consumption feature and the reliability of data positioning, it is necessary to first perform a verification operation on the multi-dimensional energy consumption feature x i , based on the received multi-dimensional energy state anchor point A and the multi-dimensional energy consumption feature x i . The verification process is as follows:

[0121] Calculate the decoding tuple D = B·A = {d1, d2, …, d i , …, d n};

[0122] Among them, d i is any element in the decoding tuple D;

[0123] Calculate the decoding factor fd:

[0124] Calculate the verification factor fc = S·fd;

[0125] Calculate the secondary encoded data Dc′ = H(data′);

[0126] Among them, data′ is the multi-dimensional energy consumption feature x i received in binary encoded data;

[0127] If fc = Dc′, the verification passes, that is, the received multi-dimensional energy consumption feature x i has stability and the data positioning has reliability; otherwise, an alarm is immediately given;

[0128] In addition, the model architecture of the energy consumption prediction model is as follows:

[0129] (1) Input layer: Construct a time series feature matrix, and splice the multi-dimensional feature vectors of N consecutive time steps within the sliding time window into a two-dimensional input matrix X t :

[0130] Among them, x t represents the multi-dimensional energy consumption feature vector extracted at the t-th moment; T represents the transpose operation of the matrix; represents a real number matrix with N rows and d columns, that is, each element in the matrix is a real number. For the building energy consumption prediction model, this matrix represents the multi-dimensional features containing N time steps, and each time step consists of d real number feature values; d represents the feature dimension at each moment;

[0131] (2) Temporal Feature Residual Modeling Layer (Multi-Scale Residual Temporal Encoder, MSRTE): Introduce multi-scale convolution and residual connection, extract local trends and periodic changes, perform multi-scale processing on the input time series using one-dimensional convolution, and combine the features of different convolution scales through the residual network to enhance the non-linear modeling ability of the model: H1 = ReLU(Conv k'＝3 (X t )) + Conv k'＝5 (X t );

[0132] Among them, H1 represents the feature representation after convolution and activation, with a size of d′ represents the output dimension after convolution; Conv k' represents a one-dimensional convolution operation; k′ represents the size of the convolution kernel (in this embodiment, two different sizes of convolution kernels k = 3 and k = 5 are used to extract features of different time scales, and through the addition operation (residual connection), the convolution features of the two scales are combined); ReLU() represents the ReLU activation function;

[0133] (3) Key Dependency Extraction Layer: The self-attention mechanism (Self-Attention) is introduced to improve the model's ability to model long-term dependencies by calculating the dependencies between time steps, helping the model focus on the important parts of the input data and assigning different attention weights to each time step:

[0134]

[0135] Q = H1W Q ;

[0136] K = H1W K ;

[0137] V = H1W V ;

[0138] where Attention(Q, K, V) represents the self-attention operation; Q, K, and V represent the query (Query), key (Key), and value (Value) matrices obtained through linear transformation, which are generated from the feature matrix H1 of the previous step by the trainable weight matrices W Q , W K , W V ; d k represents the dimension of the query vector; softmax() represents the softmax function;

[0139] Accordingly: After passing through the self-attention mechanism, the key dependency extraction layer outputs the feature representation H2, with the size of

[0140] (4) Prediction Layer: Using the feature output after the above processing for energy consumption prediction, a multi-layer perceptron (MLP) is used for regression to preliminarily predict the energy consumption value at the next moment

[0141] where, represents the feature representation at time step t obtained from the self-attention mechanism;

[0142] (5) Dynamic Prediction Correction Layer: After preliminarily completing the prediction of the preliminary energy consumption value at the next moment, although the model has obtained the key dependency feature representation H2 and the preliminary prediction value of the energy consumption at the next moment However, building energy consumption is affected by various complex factors (including but not limited to regional differences, building types, usage habits). Relying solely on current data for modeling may have problems of overfitting or insufficient generalization. Therefore, the historical big data feature library D hist is introduced as an external reference mechanism to dynamically correct the current prediction and enhance the robustness and generalization ability of the model. Specifically:

[0143] Construct a historical feature clustering dictionary: By performing clustering analysis on the energy consumption characteristics of a large number of historical buildings, a set of typical energy consumption pattern vectors (cluster centers) are formed:

[0144] Among them, C j represents the central vector of the j-th type of energy consumption pattern; N j represents the number of samples in the j-th type; represents the sample belonging to the j-th type; M represents the total number of energy consumption pattern categories, that is, the total number of historical feature clusters;

[0145] Calculate the similarity between the current feature and the historical features: Match the feature vector after modeling at the current time step with each historical pattern:

[0146] Among them, j * represents the closest historical feature center; ||·||2 represents the Euclidean distance (L2 norm), which is used to measure the difference degree between two feature vectors;

[0147] Dynamic prediction fusion:

[0148] Among them, represents the basic prediction value; represents the average energy consumption value corresponding to the historical cluster center j * ; λ represents the weighting coefficient, which is used to control the fusion ratio of the basic prediction and the empirical prediction, λ ∈ [0, 1]; represents the final prediction value after fusion correction;

[0149] Accordingly: Transmit the final prediction value to the intelligent optimization decision-making module;

[0150] S4. Based on the dynamic deviation between the energy consumption prediction value and the actual observation value, combined with the historical fluctuation threshold and the adaptive multi-scale control mechanism, the intelligent optimization decision-making module realizes intelligent, interpretable, and high-precision energy consumption anomaly detection and multi-level linkage warning, specifically as follows:

[0151] S41. Dynamic deviation calculation and normalization processing: At each time step t + 1, record the actual energy consumption value and compare it with the prediction value output in step S3 to obtain the absolute error ε t+1 :

[0152] S42. To facilitate the comparability of errors between different time periods, introduce sliding standard deviation normalization processing:

[0153] Among them, et+1 represents the normalized error index, that is, the degree of normalization of the current anomaly intensity relative to historical volatility; σ [t-W,t] represents the standard deviation of actual energy consumption in the past W time steps;

[0154] S43, based on the normalized error index e t+1 , define abnormality discrimination as multi-level rules:

[0155]

[0156] Among them, θ1 and θ2 are dynamic thresholds set using the historical 95% and 99% quantiles respectively; Level(t+1) is the abnormality level at time t+1;

[0157] It should be noted that the percentile represents the value of a certain percentage of a set of data; the 95% percentile means that 95% of the data values are less than or equal to this value; the 99% percentile means that 99% of the data values are less than or equal to this value;

[0158] S44. To prevent false alarms caused by instantaneous spike errors or data acquisition jitter, a continuous anomaly trigger judgment mechanism within a sliding window is introduced: the anomaly level sequence of the past M time steps is defined as {Level(t+1-M+1),...,Level(t+1)}. If the following conditions are met, the anomaly is considered valid:

[0159]

[0160] Where I() represents the indicator function; γ represents the effective trigger rate threshold, γ∈[0,1]. For example, γ=0.6 means that the trigger is triggered only when 60% of the past time steps are abnormal; M represents the sliding window width;

[0161] It should be noted that the indicator function I() is also called the indicative function. It is a function that only outputs 0 or 1 to indicate whether a certain condition is met. It is defined as:

[0162] S45. Define different alarm strategies based on the abnormality level Level (t+1) and its persistence:

[0163] Strategy 1: Level 0; normal building energy consumption; no measures;

[0164] Strategy 2: Level 1; slight abnormality in building energy consumption; send a prompt message to the administrator and record the log;

[0165] Strategy 3: Level 2: Severe abnormal building energy consumption; linked to air conditioning / lighting energy consumption strategy adjustment, administrator SMS push, automatic triggering of remote shutdown / lock control, and initiation of abnormality reporting process;

[0166] S46. Transmit the abnormal level Level(t + 1), its persistence, and the formulated alarm strategy to the visualization display and alarm module;

[0167] S5. The visualization display and alarm module uses the built-in visualization device to display the energy consumption data, predicted values the abnormal level Level(t + 1), its persistence, and the formulated alarm strategy.

[0168] The embodiments of the present invention have been described in detail above in conjunction with the accompanying drawings. However, the present invention is not limited thereto. Within the scope of knowledge possessed by those skilled in the art to which the present invention pertains, various changes can be made without departing from the gist of the present invention.

Claims

1. A building energy consumption management method based on big data, characterized in that It includes the following specific implementation steps: S1. Collect the energy consumption data in the building; S2. Construct a window sequence for the energy consumption data through a sliding window, calculate the robust mean and absolute median difference to eliminate the interference of extreme values, evaluate data anomalies based on the normalized deviation and perform adaptive substitution using an exponentially decaying weight, dynamically update the statistics and window parameters to maintain the adaptive ability, generate multi-dimensional energy consumption features using z-score normalization, and construct multi-dimensional energy state anchors by combining finite field coding and a pre-set hash function; S3. Generate finite field coded data through the multi-dimensional energy state anchors and multi-dimensional energy consumption features and solve the linear equation to achieve data stability verification, construct an energy consumption prediction model, based on the time series feature matrix, use multi-scale residual convolution and self-attention mechanism to extract local trends and long-term dependence relationships, after outputting the basic prediction value through a multi-layer perceptron, dynamically weighted fusion is combined with the historical clustering center to correct the prediction deviation and generate the energy consumption prediction value; S4. Generate a standardized error index by calculating the absolute error between the prediction value and the actual energy consumption value and performing normalization based on the standard deviation of the sliding window, divide the anomaly levels, based on the continuous anomaly trigger determination mechanism within the sliding window, determine the anomaly persistence according to the effective trigger rate threshold, and trigger the hierarchical alarm strategy according to the anomaly levels; S5. Visualize and display the energy consumption data, energy consumption prediction value, anomaly level, anomaly persistence, and alarm strategy.

2. The method for managing building energy consumption based on big data according to claim 1, characterized in that, The adaptive substitution process is as follows: S21. Based on the energy consumption data S = {s1, s2,..., s i ,…, s t}, construct a window sequence of length w in a sliding window manner: W t = {s t-w+1 , s t-w+2 ,…, s t}; Among them, s i represents the original energy consumption value collected at the i-th moment; w represents the window length; W t represents the continuous observation data within the current window; S22. Calculate the robust statistical mean and local fluctuation amplitude: Among them, W t ' represents the data after removing the extreme values of the upper and lower 10% of W t ; s i ' represents the i-th data value in W t '; μ t represents the robust mean; σ t represents the absolute median deviation based on μ t ; median() represents the median function; S23. Evaluate the degree of deviation of the current value s t from the robust mean: Calculate the discriminant function: θ t = k·log(1 + Var(W t )); Among them, δ t represents the normalized deviation; ε represents a small constant to prevent division by zero; θ t represents the dynamic threshold; k represents the empirical coefficient; Var(W t ) represents the variance within the current window; accept indicates that the current data s t has no anomaly; filter indicates that the current data s t is determined to be anomalous; S24. If the current data s t is determined to be an anomaly, i.e., a filter, then perform adaptive replacement: Among them, exp() represents the natural exponential function; represents the substitution value; w i represents the decay weight assigned to the time distance; α represents the decay speed control factor.

3. A building energy consumption management method based on big data according to claim 1, characterized in that, The construction process of the multi-dimensional energy state anchor is as follows: S31. Encode the multi-dimensional energy consumption feature x i to obtain the encoded data Dc = H(data) ∈ GF(p) n ; where data is the multi-dimensional energy consumption feature x i 's binary encoded data; H() is a pre-set hash function with an output length of n×log2p; p is a pre-defined large prime number satisfying p≡3mod 4; n is the set number of variables; GF(p) n represents an n-dimensional vector space composed of elements {0, 1, 2, …, p - 1} on the finite field GF(p); S32. Define the linear equation S·EⅠ = Dc, and solve to obtain the first-order coding tuple EⅠ = S -1 ·Dc = {z1, z2, …, z i , …, z n}; where S is a preset invertible n×n matrix over the finite field GF(p); z i is any solution of the equation S·EⅠ = Dc; S33. Calculate to obtain the second-order coding tuple EⅡ = {y1, y2, …, y i , …, y n}; where y i is any element in the second-order coding tuple EⅡ; c i is an element in the predefined constant vector c = [c1, c2, …, c i , …, c n , c ∈ GF(p) n ; S34. Define the linear equation B·A = EⅡ, and solve to obtain the multi-dimensional energy state anchor point A = B -1 ·EⅡ; Where B is a pre-set invertible n×n matrix over the finite field GF(p).

4. The method for managing building energy consumption based on big data according to claim 3, wherein The verification process for realizing data stability verification is as follows: S41. Calculate the decoded tuple D = B·A = {d1, d2, …, d i , …, d n}; where d i is any element in the decoding tuple D; S42. Calculate the decoding factor fd: S43. Calculate the verification factor fc = S·fd; S43. Calculate the secondary coded data Dc′ = H(data′); where data′ is the binary-encoded data of the received multi-dimensional energy consumption feature x i ; S44. If fc = Dc′, the verification passes, that is, the received multi-dimensional energy consumption feature x i has stability and the data positioning has reliability; otherwise, an alarm is immediately issued.

5. A building energy consumption management method based on big data according to claim 1, characterized in that, The model architecture of the energy consumption prediction model includes: Input layer: Construct a time series feature matrix, and concatenate the multi-dimensional feature vectors of N consecutive time steps within a sliding time window into a two-dimensional input matrix X t : where x t represents the multi-dimensional energy consumption feature vector extracted at the t-th moment; T represents the transpose operation of the matrix; represents a real matrix with N rows and d columns; d represents the feature dimension at each moment; Temporal Feature Residual Modeling Layer: Introduce multi-scale convolution and residual connection, use one-dimensional convolution to perform multi-scale processing on the input time series, and combine the features of different convolution scales with the residual network: H1 = ReLU(Conv k'＝3 (X t )) + Conv k'＝5 (X t ); Among them, H1 represents the feature representation after convolution and activation; d′ represents the output dimension after convolution; Conv k' represents a one-dimensional convolution operation; k′ represents the size of the convolution kernel; ReLU() represents the ReLU activation function; Key dependence extraction layer: Introduce the self-attention mechanism and output the feature representation H2: Q = H1W Q ; K = H1W K ; V = H1W V ; Among them, Attention(Q, K, V) represents the self-attention operation; Q, K, and V represent the query, key, and value matrices obtained through linear transformation; W Q , W K , W V represent weight matrices; d k represents the dimension of the query vector; softmax() represents the softmax function; Prediction layer: Use a multi-layer perceptron for regression to initially predict the energy consumption value at the next moment Among them, represents the feature representation at time step t obtained from the self-attention mechanism; Dynamic prediction correction layer: Dynamically weighted fusion is combined with the historical clustering center to correct the prediction deviation and generate the energy consumption prediction value.

6. The building energy consumption management method based on big data according to claim 5, characterized in that, The generation process of dynamically weighted fusion combined with the historical clustering center to correct the prediction deviation and generate the energy consumption prediction value is as follows: S61. Construct a historical feature clustering dictionary: By performing clustering analysis on the energy consumption characteristics of a large number of historical buildings, a set of typical energy consumption pattern vectors, i.e., clustering centers, are formed: Among them, C j represents the center vector of the j-th energy consumption pattern; N j represents the number of samples in the j-th class; represents the samples belonging to the j-th class; M represents the total number of energy consumption pattern classes; S62. Calculate the similarity between the current feature and the historical feature: the feature vector after modeling at the current time step Perform similarity matching with each historical pattern: where j * represents the center of the nearest historical feature; ||·||2 represents the Euclidean distance; S63, Dynamic Prediction Fusion: Among them, represents the basic predicted value; represents the historical clustering center j * corresponding average energy consumption value; λ represents the weighting coefficient, λ ∈ [0, 1]; represents the final predicted value after fusion correction.

7. The method for managing building energy consumption based on big data according to claim 6, wherein, The process of dividing the anomaly levels is as follows: S71. Dynamic deviation calculation and normalization processing: At each time step t+1, record the actual energy consumption value and compare it with the predicted value to obtain the absolute error ε t+1 : S72. Introduce sliding standard deviation normalization processing: Among them, e t+1 represents the normalized error index; σ [t-W,t] represents the standard deviation of the actual energy consumption in the past W time steps; S73. Based on the normalized error index e t+1 , define the anomaly discrimination as a multi-level rule: Where θ1 and θ1 are respectively dynamic thresholds set using the historical 95% and 99% quantiles; Level(t + 1) is the anomaly level at time t + 1.

8. The building energy consumption management method based on big data according to claim 7, characterized in that The continuous anomaly trigger determination mechanism based on the sliding window is: Define the anomaly level sequence of the past M time steps as {Level(t + 1 - M + 1),..., Level(t + 1)}. If the following conditions are met, the anomaly is considered valid: Where I() represents the indicator function; γ represents the effective trigger rate threshold, γ ∈ [0, 1]; M represents the sliding window width.

9. A building energy consumption management system based on big data, which is used to execute a building energy consumption management method based on big data according to any one of claims 1 to 8, characterized in that, It includes: Multi-source energy consumption data acquisition module, used to construct a distributed sensor network inside the building and collect building energy consumption data; Edge data processing module, used to perform preliminary data filtering locally; Intelligent analysis and prediction module, used to construct an energy consumption prediction model using a big data processing engine and output the energy consumption prediction value; Intelligent optimization decision-making module, which is used to formulate an alarm strategy based on the predicted energy consumption value; Visualization display and alarm module, which is used to visually display data and give alarms in a timely manner.

Images