Fuel gas central station data processing optimization method based on data analysis

Abstract

The invention relates to the technical field of fuel gas data processing, in particular to a fuel gas central station data processing optimization method based on data analysis, which comprises the following steps: S1, data acquisition and preprocessing: acquiring fuel gas use data in real time; s2, data analysis: analyzing the preprocessed gas use data; s3, gas pipe network fault prediction: identifying weak links and potential fault points in the pipe network; s4, user behavior analysis: providing personalized services and suggestions; s5, optimization processing: optimizing a gas supply strategy in combination with the prediction result of the gas pipe network; and S6, real-time monitoring and user feedback: dynamically adjusting the optimized gas supply strategy according to monitoring data and user feedback information. According to the invention, it is ensured that the system can early warn and process potential problems in time, so that the gas supply strategy is optimized, and the safety and operation efficiency of the system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of gas data processing, and in particular to a gas middle station data processing optimization method based on data analysis. Background Art

[0002] With the acceleration of urbanization and the improvement of living standards, gas has become one of the indispensable energy sources in residents' daily life and industrial production. In order to ensure the stability and safety of gas supply, the data processing and optimization methods of the gas middle platform are particularly important. By collecting, analyzing and predicting gas usage data, the efficiency of the gas supply system can be improved, resource waste can be reduced, and potential safety hazards can be effectively prevented and dealt with.

[0003] The existing SaaS platform has provided 12 core business services including engineering construction management, customer service, billing management, mobile operations, collaborative office, digital reconciliation, etc. These services have effectively improved the management efficiency and business processing capabilities of enterprises and standardized business processes. However, despite certain achievements, the existing system still faces some challenges in data management and use:

[0004] Most existing gas data processing methods rely on simple data collection and basic statistical analysis, lacking in-depth data mining and intelligent analysis. Traditional methods make it difficult to monitor gas usage in real time and cannot promptly detect and predict fault points in the pipeline network. In addition, the lack of user behavior analysis and personalized services makes it impossible for gas supply strategies to effectively meet the needs of different users, resulting in inefficient gas utilization, increased safety hazards, and low user satisfaction.

[0005] The present invention aims to provide a gas middle-station data processing optimization method based on data analysis, dynamically adjust the gas supply strategy, comprehensively improve the safety, reliability and efficiency of the gas supply system, meet the diverse needs of users, and improve user satisfaction. Summary of the invention

[0006] Based on the above objectives, the present invention provides a gas middle-station data processing optimization method based on data analysis.

[0007] The data processing optimization method of the gas middle station based on data analysis includes the following steps:

[0008] S1, data collection and preprocessing: collect gas usage data in real time through the sensor network, including gas pressure, flow and temperature, and preprocess the collected gas usage data;

[0009] S2, data analysis: analysis of pre-processed gas usage data, including trend analysis, anomaly detection, and demand forecasting;

[0010] S3, gas pipeline network fault prediction: Fault prediction of the gas pipeline network, identification of weak links and potential fault points in the pipeline network, including:

[0011] S31, feature extraction: extract key features from gas pipeline network data, including pressure change rate, flow fluctuation and temperature gradient;

[0012] S32, model construction: construct a gas pipeline network fault prediction model based on the extracted key features to predict gas pipeline network faults;

[0013] S33, real-time prediction: inputting real-time gas pipeline network data into the gas pipeline network fault prediction model to perform fault prediction and identify weak links and potential fault points in the pipeline network;

[0014] S34, result verification: verify the accuracy of the prediction results by comparing with the actual fault records and inspection results, and continuously optimize and adjust the gas pipeline network fault prediction model;

[0015] S4, user behavior analysis: Analyze users' gas usage behavior through big data and provide personalized services and suggestions, including smart gas usage suggestions and energy-saving solution push;

[0016] S5, optimization processing: Based on the analysis results of gas usage data and user behavior, combined with the prediction results of the gas pipeline network, the gas supply strategy is optimized, including adjusting the supply pressure and optimizing the supply route;

[0017] S6, real-time monitoring and user feedback: establish a real-time monitoring mechanism to monitor gas usage, collect user feedback information, and dynamically adjust the optimized gas supply strategy based on monitoring data and user feedback information.

[0018] Optionally, the data collection and preprocessing in S1 includes:

[0019] S11, data collection: real-time collection of gas usage data, including:

[0020] Air pressure data collection: Use pressure sensors to monitor the air pressure in the pipeline in real time;

[0021] Flow data collection: Use flow meter sensor to measure the flow of gas;

[0022] Temperature data collection: Use temperature sensors to monitor the temperature changes of gas;

[0023] S12, data preprocessing: preprocessing the collected gas usage data, including data cleaning and data format conversion.

[0024] Optionally, the data analysis in S2 includes:

[0025] S21, Trend analysis: Use linear regression model to perform trend analysis on the preprocessed gas usage data, expressed as:

[0026] y = β0 + β1x;

[0027] Among them, y is the predicted gas usage, x is time, β0 and β1 are regression coefficients;

[0028] S22, anomaly detection: The seasonal autoregressive integrated moving average model (SARIMA) is used to detect anomalies in the preprocessed gas usage data, which is expressed as:

[0029]

[0030] Among them, y t is the current data point, μ is a constant term, is the autoregressive coefficient, θ1 is the moving average coefficient, ∈ t is the error term, Φ1 and Θ1 are seasonal autoregressive and moving average coefficients, and s is the length of the season;

[0031] S23, demand forecasting: The grey forecasting model (GM) is used to forecast the demand of the pre-processed gas usage data, including:

[0032] S231, data accumulation generation: the original data sequence x = {x(1), x(2), ..., x(n)} is accumulated and generated once to obtain a new data sequence X = {X(1), X(2), ..., X(n)}, which is expressed as:

[0033]

[0034] S232, establish differential equation: establish a grey differential equation according to the data sequence generated by accumulation, which is expressed as:

[0035]

[0036] Among them, a and b are parameters to be determined;

[0037] S233, solving parameters: using the least square method to solve the unknown parameters a and b, expressed as:

[0038] Define the matrix B and vector Y:

[0039]

[0040] Use the least squares method to solve for the unknown parameters:

[0041]

[0042] Among them, (BT B) 1 It is B T The inverse matrix of B, B T is the transpose of matrix B;

[0043] S234, prediction value generation: Based on the parameters obtained, a prediction model is established and the prediction value of the original data is solved. It is expressed as:

[0044]

[0045] in, is the predicted value of the cumulative generated sequence at the k+1th moment, X(1) is the value of the cumulative generated sequence at the initial moment, e is the base of the natural logarithm, k is the time index, is the predicted value of the original data sequence at the k+1th moment, It is the predicted value of the cumulatively generated sequence at the kth moment.

[0046] Optionally, the feature extraction in S31 includes:

[0047] S311, Pressure change rate: Calculate the rate of change of pressure over time, expressed as:

[0048]

[0049] Where ΔP is the pressure change rate, P(t) is the pressure at the current moment, P(t-1) is the pressure at the previous moment, and Δt is the time interval;

[0050] S312, flow fluctuation: Calculate the fluctuation amplitude of flow over time, expressed as:

[0051]

[0052] Among them, ΔF is the flow fluctuation, F i is the flow value at the i-th moment, is the average value of flow, n is the number of observations;

[0053] S313, temperature gradient: calculate the temperature variation at different locations in the pipe network, expressed as:

[0054]

[0055] in, is the temperature gradient, T(x2) and T(x1) are the temperatures at the network locations x2 and x1, respectively, and x2-x1 is the distance between the two locations.

[0056] Optionally, the gas pipeline network fault prediction model in S32 adopts a graph neural network (GNN) model, and the graph neural network (GNN) model includes:

[0057] S321, graph representation and feature extraction: The gas pipeline network is represented as a graph G = (V, E), where nodes V represent gas equipment and sensors, and edges E represent pipeline connections. Key features are extracted from the gas pipeline network data as node attributes. The feature vector h of each node v is v Including pressure change rate ΔP, flow fluctuation ΔF and temperature gradient

[0058] S322, node feature enhancement: introduce time feature t and position feature p as additional dimensions of node features, expressed as:

[0059]

[0060] S323, Message passing and update mechanism: Optimize the message passing function, consider the characteristics of the edge uv The weighted sum of the characteristics of the pipeline (such as pipeline length, pipeline diameter) and the node characteristics is expressed as:

[0061] Messaging:

[0062] in, is the message of node v at level k, is the set of neighbor nodes of node v, α uv is the attention weight between node u and node v, W e and W h is the weight matrix, b h is bias;

[0063]

[0064] Node Update:

[0065] in, is the feature representation of node v at layer k, W m is the message weight matrix, b m For message bias;

[0066] S324, multi-layer stacking and global information fusion: introduce the global pooling layer to fuse the information of all nodes together for global feature extraction, expressed as:

[0067]

[0068] Among them, h G It is a global feature representation, and Pooling is average pooling;

[0069] S325, fault prediction output: Use the fully connected layer to perform fault prediction and predict whether the node has a fault, expressed as:

[0070]

[0071] in, is the fault prediction value of node v, σ is the Sigmoid activation function, W o is the output layer weight matrix, b o is the bias for the output layer.

[0072] Optionally, the real-time prediction in S33 includes:

[0073] S331, real-time data input: the real-time collected gas pipeline network data is used as the node feature vector h v Input into the gas network fault prediction model;

[0074] S332, forward propagation: Through the gas pipeline fault prediction model, the node feature vector h v Perform forward propagation calculations, pass through each layer of the graph neural network in turn, and update the node feature representation Until the Kth layer, the final node feature representation is obtained

[0075] S333, Fault Prediction: Using the fully connected layer and activation function, the final node feature representation Fault prediction is performed, which is expressed as:

[0076]

[0077] in, is the fault prediction value of node v, σ is the Sigmoid activation function, W o is the output layer weight matrix, b o is the output layer bias;

[0078] S334, Identification of weak links and potential failure points: Based on the failure prediction value Identify weak links and potential failure points in the gas pipeline network. When the set failure threshold is exceeded, node v is marked as a potential failure point.

[0079] Optionally, the result verification in S34 includes:

[0080] S341, prediction result collection: collect the fault prediction results of the gas pipeline network fault prediction model on the gas pipeline network data

[0081] S342, actual fault record comparison: the predicted results The actual fault record v Compare and calculate the accuracy indicators of the predictions, including accuracy, precision, recall, and F1 score;

[0082] S343, inspection result comparison: compare the prediction results with the inspection results to evaluate the effect of the model in practical application;

[0083] S344, model optimization and adjustment: Based on the verification results, the model parameters are optimized and adjusted using the back propagation algorithm and gradient descent method, expressed as:

[0084]

[0085] Among them, θ is the model parameter, η is the learning rate, and L(θ) is the loss function.

[0086] Optionally, the user behavior analysis in S4 includes:

[0087] S41, data collection: The user's gas usage data is collected in real time through the smart gas meter and sensor network to form a user gas usage behavior data set D, which is expressed as:

[0088] D={(u i ,t j ,x ij )};

[0089] Among them, u i For user i,t j is time j,x ij is the gas usage of user i at time j;

[0090] S42, behavior pattern recognition: using K-means clustering algorithm to identify the user's gas usage behavior pattern;

[0091] S43, Personalized service and suggestion generation: Based on the identified gas usage behavior patterns, including intelligent gas usage suggestions and energy-saving solution push, personalized services and suggestions are generated, including:

[0092] S431, intelligent gas usage suggestions: Based on the user's behavior pattern, the Apriori algorithm is used to generate intelligent gas usage suggestions;

[0093] S432, energy-saving solution push: Based on the user's behavior pattern and usage data, regression analysis is used to predict the user's gas demand and provide energy-saving solutions.

[0094] Optionally, the optimization process in S5 includes:

[0095] S51, supply pressure adjustment: Based on the gas usage data and user behavior analysis results, combined with the prediction results of the gas pipeline network, the gas supply pressure is adjusted using an optimization algorithm, which is expressed as:

[0096]

[0097] Among them, P opt is the optimized supply pressure, P base is the basic supply pressure, α is the adjustment coefficient, w i is the weight of user i, d i is the actual demand of user i, is the predicted demand of user i;

[0098] S52, supply route optimization: Based on the fault prediction results of the gas pipeline network, the flow distribution algorithm is used to optimize the gas supply route, which is expressed as:

[0099] min∑ (i,j)∈E c ij f ij ;

[0100] Constraints:

[0101]

[0102] Among them, c ij represents the transmission cost of edge (i, j), f ij represents the flow on edge (i,j), d i represents the demand of node i, and E represents the edge set;

[0103] S53, comprehensive optimization: Combine the results of pressure adjustment and route optimization, comprehensively consider the cost and benefit, and use linear programming to formulate the optimal gas supply strategy, which is expressed as:

[0104] Objective function:

[0105] Constraints:

[0106] Flow conservation constraints:

[0107] Capacity constraints:

[0108] Supply pressure constraint: P min ≤P opt ≤P max ;

[0109] Among them, u ij is the capacity of edge (i,j), P opt is the optimized supply pressure, P baseis the basic supply pressure, α is the adjustment coefficient, w ij is the weight of user i at position j, d ij is the actual demand of user i at location j, is the predicted demand of user i at location j, β is the weight coefficient of supply pressure adjustment, P min and P max are the minimum and maximum limits for supply pressure.

[0110] Optionally, the real-time monitoring and user feedback in S6 includes:

[0111] S61, Real-time monitoring mechanism established: Through the sensor network and smart gas meters, real-time gas usage data including pressure, flow and temperature are collected and transmitted to the monitoring center;

[0112] S62, user feedback information collection: collect user feedback information on gas supply through user feedback platforms (such as mobile applications or customer service systems), including gas use experience, safety hazards and suggestions;

[0113] S63, Data Fusion and Analysis: Real-time monitoring data D t User feedback information f Fusion, using data analysis algorithms to evaluate the effectiveness of the current gas supply strategy and identify potential problems, expressed as:

[0114] S eval =α∑ t D t +β∑ i U f ;

[0115] Among them, S eval is the evaluation score of the supply strategy, α and β are weight coefficients;

[0116] S64, dynamic adjustment: according to the evaluation score S eval Based on the potential problems identified, the gas supply strategy is dynamically adjusted using an adaptive optimization algorithm, which is expressed as:

[0117]

[0118] Among them, θ is the parameter of the gas supply strategy, η is the learning rate, L(θ,S eval ) is the loss function based on the evaluation score.

[0119] Beneficial effects of the present invention:

[0120] The present invention collects gas usage data in real time through a sensor network and performs preprocessing, thereby ensuring the accuracy, synchronization and consistency of the data, and uses linear regression, SARIMA model and grey prediction model algorithms to perform data analysis, effectively perform trend analysis, anomaly detection and demand forecasting, thereby ensuring the accuracy and reliability of data processing, providing a solid data foundation for subsequent fault prediction and optimization processing, and ultimately improving the safety and efficiency of the gas supply system.

[0121] The present invention improves the ability to capture the dynamic characteristics of the gas pipeline network through feature extraction, node feature enhancement, message transmission and update mechanism, multi-layer stacking and global information fusion in the graph neural network model, combined with pressure change rate, flow fluctuation, temperature gradient, time characteristics and location characteristics, combines real-time prediction with result verification, continuously optimizes and adjusts model parameters, improves the accuracy and real-time performance of fault prediction, ensures that the system can timely warn and handle potential problems, thereby optimizing the gas supply strategy and improving the safety and operational efficiency of the system.

[0122] The present invention analyzes the user's gas usage behavior through user behavior analysis and utilizes big data to provide personalized services and suggestions, including intelligent gas usage suggestions and energy-saving solution push, optimizes the analysis results of gas usage data and user behavior, and optimizes the gas supply strategy in combination with the prediction results of the gas pipeline network, including adjusting the supply pressure and optimizing the supply route, combining the real-time monitoring mechanism with user feedback, and dynamically adjusting the gas supply strategy through data fusion and analysis, thereby improving the reliability and safety of gas supply, reducing operating costs, effectively responding to demand fluctuations and potential failure points, and enhancing the overall benefits of the system and user satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0123] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0124] Figure 1 A schematic diagram of the optimization method flow of an embodiment of the present invention;

[0125] Figure 2 Schematic diagram of gas pipeline network fault prediction according to an embodiment of the present invention. DETAILED DESCRIPTION

[0126] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0127] It should be noted that the references to "one embodiment", "an embodiment", "an exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0128] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0129] like Figure 1-Figure 2 As shown, the data processing optimization method of the gas middle station based on data analysis includes the following steps:

[0130] S1, data collection and preprocessing: collect gas usage data in real time through the sensor network, including gas pressure, flow and temperature, and preprocess the collected gas usage data;

[0131] S2, data analysis: analysis of pre-processed gas usage data, including trend analysis, anomaly detection, and demand forecasting;

[0132] S3, gas pipeline network fault prediction: predict gas pipeline network faults, identify weak links and potential fault points in the pipeline network, and improve prediction accuracy, including:

[0133] S31, feature extraction: extract key features from gas pipeline network data, including pressure change rate, flow fluctuation and temperature gradient;

[0134] S32, model construction: construct a gas pipeline network fault prediction model based on the extracted key features to predict gas pipeline network faults;

[0135] S33, real-time prediction: inputting real-time gas pipeline network data into the gas pipeline network fault prediction model to perform fault prediction and identify weak links and potential fault points in the pipeline network;

[0136] S34, result verification: verify the accuracy of the prediction results by comparing with the actual fault records and inspection results, and continuously optimize and adjust the gas pipeline network fault prediction model;

[0137] S4, User Behavior Analysis: Through big data analysis of users’ gas usage behavior, we provide personalized services and suggestions, including smart gas usage suggestions and energy-saving solution push, to improve user satisfaction and service quality, while optimizing the overall gas supply strategy;

[0138] S5, optimization processing: Based on the analysis results of gas usage data and user behavior, combined with the prediction results of the gas pipeline network, the gas supply strategy is optimized, including adjusting the supply pressure and optimizing the supply route;

[0139] S6, real-time monitoring and user feedback: establish a real-time monitoring mechanism to monitor gas usage, collect user feedback information, and dynamically adjust the optimized gas supply strategy based on monitoring data and user feedback information;

[0140] Through the above steps, not only the accuracy and real-time performance of data processing are improved, but also the fault prediction and response capabilities are enhanced, thereby improving user satisfaction and the overall efficiency and safety of gas supply.

[0141] Data collection and preprocessing in S1 include:

[0142] S11, data collection: real-time collection of gas usage data, including:

[0143] Air pressure data collection: Use pressure sensors to monitor the air pressure in the pipeline in real time;

[0144] Flow data collection: Use flow meter sensor to measure the flow of gas;

[0145] Temperature data collection: Use temperature sensors to monitor the temperature changes of gas;

[0146] S12, data preprocessing: preprocessing the collected gas usage data, including data cleaning and data format conversion;

[0147] Data cleaning uses data cleaning algorithms to remove outliers and noise data, including:

[0148] Outlier detection: Use the 3σ principle to detect outliers, that is, for each data point x in the data set i , if |x i-μ|>3σ, then the point is considered an outlier, where μ is the mean of the data set and σ is the standard deviation;

[0149] Data smoothing: The sliding average method is used to smooth the data, expressed as:

[0150]

[0151] Among them, S i is the smoothed data, x j is the jth data point, k is the size of the sliding window;

[0152] Data format conversion: Convert the data formats collected by different sensors into a standard format to facilitate subsequent analysis and processing, including:

[0153] Timestamp alignment: unify the timestamps of different sensor data to ensure data synchronization. Suppose the timestamp of sensor A is t A , the timestamp of sensor B is t B , then the converted timestamp t′ is:

[0154]

[0155] Unit unification: Convert the units of different sensor data into a unified unit, unify the air pressure unit into Pascal (Pa), and the flow unit into cubic meters per hour (m 3 / h), the temperature unit is Celsius (℃);

[0156] Through the above content, the accuracy, synchronization and consistency of the data are ensured, and the data quality is improved, thus providing a solid data foundation for fault prediction and optimization processing, and ultimately improving the safety and efficiency of the gas supply system.

[0157] The data analysis in S2 includes:

[0158] S21, Trend analysis: Use linear regression model to perform trend analysis on the preprocessed gas usage data, expressed as:

[0159] y = β0 + β1x;

[0160] Among them, y is the predicted gas usage, x is time, β0 and β1 are regression coefficients;

[0161] The least squares method is used to estimate the regression coefficients β0 and β1, including:

[0162] Objective function: Assuming a linear regression model, the goal of the least squares method is to minimize the sum of squared errors, expressed as:

[0163]

[0164] Among them, y i is the actual value of the ith observation, x i is the independent variable value of the i-th observation, β0 and β1 are regression coefficients, and n is the total number of observations;

[0165] Solve the regression coefficient: By taking the partial derivatives of S with respect to β0 and β1 respectively and setting them to zero, it is expressed as:

[0166]

[0167] in, is the mean value of x, is the mean value of y;

[0168] Calculate the average: Calculate the average of x and y, expressed as:

[0169]

[0170] S22, anomaly detection: The seasonal autoregressive integrated moving average model (SARIMA) is used to detect anomalies in the preprocessed gas usage data, which is expressed as:

[0171]

[0172] Among them, y t is the current data point, μ is a constant term, is the autoregressive coefficient, θ1 is the moving average coefficient, ∈ t is the error term, Φ1 and Θ1 are seasonal autoregressive and moving average coefficients, and s is the length of the season;

[0173] S23, demand forecasting: The grey forecasting model (GM) is used to forecast the demand of the pre-processed gas usage data, including:

[0174] S231, data accumulation generation: the original data sequence x = {x(1), x(2), ..., x(n)} is accumulated and generated once to obtain a new data sequence X = {X(1), X(2), ..., X(n)}, which is expressed as:

[0175]

[0176] S232, establish differential equation: establish a grey differential equation according to the data sequence generated by accumulation, which is expressed as:

[0177]

[0178] Among them, a and b are parameters to be determined;

[0179] S233, solving parameters: using the least square method to solve the unknown parameters a and b, expressed as:

[0180] Define the matrix B and vector Y:

[0181]

[0182] Use the least squares method to solve for the unknown parameters:

[0183]

[0184] Among them, (B T B) 1 It is B T The inverse matrix of B, B T is the transpose of matrix B;

[0185] S234, prediction value generation: Based on the parameters obtained, a prediction model is established and the prediction value of the original data is solved. It is expressed as:

[0186]

[0187] in, is the predicted value of the cumulative generated sequence at the k+1th moment, X(1) is the value of the cumulative generated sequence at the initial moment, e is the base of the natural logarithm, k is the time index, is the predicted value of the original data sequence at the k+1th moment, It is the predicted value of the cumulatively generated sequence at the kth moment;

[0188] Through the above content, the accuracy and reliability of data analysis are ensured, thus providing solid data support and scientific basis for optimizing gas supply strategies and improving system safety and efficiency.

[0189] Feature extraction in S31 includes:

[0190] S311, Pressure change rate: Calculate the rate of change of pressure over time, expressed as:

[0191]

[0192] Where ΔP is the pressure change rate, P(t) is the pressure at the current moment, P(t-1) is the pressure at the previous moment, and Δt is the time interval;

[0193] S312, flow fluctuation: Calculate the fluctuation amplitude of flow over time, expressed as:

[0194]

[0195] Among them, ΔF is the flow fluctuation, F iis the flow value at the i-th moment, is the average value of flow, n is the number of observations;

[0196] S313, temperature gradient: calculate the temperature variation at different locations in the pipe network, expressed as:

[0197]

[0198] in, is the temperature gradient, T(x2) and T(x1) are the temperatures at the network locations x2 and x1, respectively, and x2-x1 is the distance between the two locations;

[0199] Through the above content, the prediction accuracy and reliability of the model can be improved, helping to timely identify weak links and potential failure points in the pipeline network, thereby optimizing the gas supply strategy and improving the safety and efficiency of the system.

[0200] The gas pipeline fault prediction model in S32 adopts the graph neural network (GNN) model, which includes:

[0201] S321, graph representation and feature extraction: The gas pipeline network is represented as a graph G = (V, E), where nodes V represent gas equipment and sensors, and edges E represent pipeline connections. Key features are extracted from the gas pipeline network data as node attributes. The feature vector h of each node v is v Including pressure change rate ΔP, flow fluctuation ΔF and temperature gradient

[0202] S322, node feature enhancement: In order to better capture the dynamic characteristics of the node, the time feature t and the position feature p are introduced as additional dimensions of the node feature, which can be expressed as:

[0203]

[0204] S323, Message passing and update mechanism: Optimize the message passing function, consider the characteristics of the edge uv The weighted sum of the characteristics of the pipeline (such as pipeline length, pipeline diameter) and the node characteristics is expressed as:

[0205] Messaging:

[0206] in, is the message of node v at level k, is the set of neighbor nodes of node v, α uv is the attention weight between node u and node v, W e and W h is the weight matrix, b h is bias;

[0207]

[0208] Node Update:

[0209] in, is the feature representation of node v at layer k, W m is the message weight matrix, b m For message bias;

[0210] S324, multi-layer stacking and global information fusion: In order to capture global information, a global pooling layer is introduced to fuse the information of all nodes together for global feature extraction, which is expressed as:

[0211]

[0212] Among them, h G It is a global feature representation, and Pooling is average pooling;

[0213] Average pooling is expressed as:

[0214]

[0215] Among them, h G is the global feature representation, V is the node set, is the feature representation of node v at the Kth layer, |V| is the size of the node set, that is, the number of nodes;

[0216] S325, fault prediction output: Use the fully connected layer to perform fault prediction and predict whether the node has a fault, expressed as:

[0217]

[0218] in, is the fault prediction value of node v, σ is the Sigmoid activation function, W o is the output layer weight matrix, b o is the output layer bias;

[0219] Through the above content, combined with the pressure change rate, flow fluctuation, temperature gradient, time characteristics and location characteristics, the ability to capture the dynamic characteristics of the gas pipeline network is improved. The message transmission and node update mechanism more accurately reflect the dependency between nodes, and enhance the extraction of important information through the attention mechanism. The global average pooling further integrates the global information to ensure the comprehensiveness and accuracy of fault prediction. The overall model improves the reliability and real-time performance of gas pipeline network fault prediction, which helps to identify weak links and potential fault points in the pipeline network in advance, thereby optimizing the gas supply strategy and improving the safety and operational efficiency of the system.

[0220] Real-time predictions in S33 include:

[0221] S331, real-time data input: the real-time collected gas pipeline network data is used as the node feature vector h v Input into the gas network fault prediction model;

[0222] S332, forward propagation: Through the gas pipeline fault prediction model, the node feature vector h v Perform forward propagation calculations, pass through each layer of the graph neural network in turn, and update the node feature representation Until the Kth layer, the final node feature representation is obtained

[0223] S333, Fault Prediction: Using the fully connected layer and activation function, the final node feature representation Fault prediction is performed, which is expressed as:

[0224]

[0225] in, is the fault prediction value of node v, σ is the Sigmoid activation function, W o is the output layer weight matrix, b o is the output layer bias;

[0226] S334, Identification of weak links and potential failure points: Based on the failure prediction value Identify weak links and potential failure points in the gas pipeline network. When the set failure threshold is exceeded, node v is marked as a potential failure point;

[0227] Through the above content, the weak links and potential failure points in the gas pipeline network are identified according to the prediction results, ensuring that the system can timely warn and deal with potential problems, thereby improving the safety and reliability of gas supply.

[0228] The fault threshold is set through historical data analysis and is expressed as:

[0229] Collect historical data: Collect a large amount of historical data, including node feature data in normal and fault states. v and the corresponding fault label y v ;

[0230] Calculate the distribution of fault prediction values: Use the gas pipeline network fault prediction model to predict historical data and obtain the fault prediction value Distribution of

[0231] Determine the threshold: Based on the distribution of the failure prediction value, the 95% percentile is selected as the failure threshold, expressed as:

[0232]

[0233] The result verification in S34 includes:

[0234] S341, prediction result collection: collect the fault prediction results of the gas pipeline network fault prediction model on the gas pipeline network data

[0235] S342, actual fault record comparison: the predicted results The actual fault record v Compare and calculate the accuracy indicators of the prediction, including accuracy, precision, recall and F1 score, including:

[0236]

[0237] Among them, TP is the true positive example, that is, the number of instances correctly predicted as faulty, TN is the true negative example, that is, the number of instances correctly predicted as non-faulty, FP is the false positive example, that is, the number of instances incorrectly predicted as faulty, and FN is the false negative example, that is, the number of instances incorrectly predicted as non-faulty;

[0238] S343, inspection result comparison: compare the prediction results with the inspection results, evaluate the effect of the model in practical application, and verify the accuracy of the model;

[0239] S344, model optimization and adjustment: Based on the verification results, the model parameters are optimized and adjusted using the back propagation algorithm and gradient descent method, expressed as:

[0240]

[0241] Among them, θ is the model parameter, η is the learning rate, and L(θ) is the loss function;

[0242] The loss function L(θ) uses the binary cross entropy loss function, expressed as:

[0243]

[0244] Through the above content, the prediction accuracy and reliability of the model are continuously improved, ensuring that the gas supply system can accurately and timely identify and handle potential faults, ensuring the safety and stable operation of the system.

[0245] User behavior analysis in S4 includes:

[0246] S41, data collection: The user's gas usage data is collected in real time through the smart gas meter and sensor network to form a user gas usage behavior data set D, which is expressed as:

[0247] D={(ui ,t j ,x ij )};

[0248] Among them, u i For user i,t j is time j,x ij is the gas usage of user i at time j;

[0249] S42, behavior pattern recognition: K-means clustering algorithm is used to identify the user's gas usage behavior pattern, which is expressed as:

[0250]

[0251] Among them, C i is the i-th cluster, μ i is the centroid of the ith cluster, x is the data point, and k is the number of clusters;

[0252] S43, Personalized service and suggestion generation: Based on the identified gas usage behavior patterns, including intelligent gas usage suggestions and energy-saving solution push, personalized services and suggestions are generated, including:

[0253] S431, intelligent gas usage suggestion: Based on the user's behavior pattern, the Apriori algorithm is used to generate intelligent gas usage suggestions, which are expressed as:

[0254]

[0255] Where A and B are two different behaviors or events, Count(A) is the number of times event A occurs, Count(A∪B) is the number of times events A and B occur simultaneously, and N is the total number of records;

[0256] S432, energy-saving solution push: Based on the user's behavior pattern and usage data, regression analysis is used to predict the user's gas demand and provide an energy-saving solution, which is expressed as:

[0257] y=β0+β1x1+β2x2+…+β n x n ;

[0258] Among them, y is the predicted gas demand, x1, x2,…, x n are the characteristics that affect gas demand, β0,β1,…,β n is the regression coefficient;

[0259] Through the above content, not only the user satisfaction and service quality are improved, but also the overall gas supply strategy can be optimized, the gas utilization efficiency can be improved, the goal of energy conservation and emission reduction can be achieved, and the economy and environmental protection of the system can be enhanced.

[0260] Smart gas usage recommendations include:

[0261] Gas usage time optimization: users are advised to use gas during off-peak hours to reduce the pressure during peak gas usage and possible price fluctuations, and provide users with daily, weekly and monthly analysis of the best gas usage time periods;

[0262] Equipment Usage Suggestions: Prompt users to use gas equipment properly, such as regularly cleaning gas stoves and checking pipeline connections to ensure that equipment operates efficiently. It also recommends users to replace energy-efficient gas equipment and provides corresponding product recommendations.

[0263] Safety reminder: Based on the user's gas usage behavior, it will remind you of possible safety hazards, such as not turning off the gas equipment for a long time, abnormal gas usage, etc. It will also provide suggestions for regular safety inspections and emergency contact information when necessary;

[0264] Gas usage mode suggestions: Based on the user's gas usage history data, more reasonable gas usage modes are suggested, such as reducing the frequency of using gas water heaters, setting the temperature appropriately, and providing a comparative analysis of gas usage behaviors of similar users to help users adjust their gas usage habits;

[0265] Energy-saving tips: Provide gas energy-saving tips, such as using pot lids properly, controlling flame size, etc., and suggest users to save gas by properly arranging cooking time and methods.

[0266] Energy-saving solutions include:

[0267] Gas consumption forecast: Based on the user's historical gas consumption data and current climate conditions, the gas consumption in the future is predicted to help users plan their gas consumption reasonably;

[0268] Energy-saving equipment recommendation: Based on the user's gas usage habits and needs, we recommend more efficient and energy-saving gas equipment and appliances, such as high-efficiency gas stoves, smart gas water heaters, etc., and provide performance comparison and cost analysis of energy-saving equipment to help users make wise choices;

[0269] Energy-saving behavior suggestions: Suggest users to take specific energy-saving behaviors, such as reducing unnecessary gas use, optimizing cooking methods, etc., and provide users with customized energy-saving plans and steps;

[0270] Gas usage optimization suggestions: Through data analysis, we propose optimization solutions for users' gas usage patterns, such as reducing gas usage during peak hours, centralizing the use of gas equipment, etc., and provide energy-saving gas usage strategies in different seasons and time periods;

[0271] Feedback and adjustment: Based on users’ actual gas usage and feedback, we continuously adjust and optimize energy-saving plans, and provide regular energy-saving reports to help users understand energy-saving effects and room for further improvement.

[0272] The optimizations in S5 include:

[0273] S51, supply pressure adjustment: Based on the gas usage data and user behavior analysis results, combined with the prediction results of the gas pipeline network, the gas supply pressure is adjusted using an optimization algorithm, which is expressed as:

[0274]

[0275] Among them, P opt is the optimized supply pressure, P base is the basic supply pressure, α is the adjustment coefficient, w i is the weight of user i, d i is the actual demand of user i, is the predicted demand of user i;

[0276] S52, supply route optimization: Based on the fault prediction results of the gas pipeline network, the flow distribution algorithm is used to optimize the gas supply route to ensure the efficiency and safety of gas transportation, which is expressed as:

[0277] min∑ (i,j)∈E c ij f ij ;

[0278] Constraints:

[0279]

[0280] Among them, c ij represents the transmission cost of edge (i, j), f ij represents the flow on edge (i,j), d i represents the demand of node i, and E represents the edge set;

[0281] S53, comprehensive optimization: Combine the results of pressure adjustment and route optimization, comprehensively consider the cost and benefit, and use linear programming to formulate the optimal gas supply strategy, which is expressed as:

[0282] Objective function:

[0283] Constraints:

[0284] Flow conservation constraints:

[0285] Capacity constraints:

[0286] Supply pressure constraint: P min ≤P opt ≤P max ;

[0287] Among them, u ijis the capacity of edge (i,j), P opt is the optimized supply pressure, P base is the basic supply pressure, α is the adjustment coefficient, w ij is the weight of user i at position j, d ij is the actual demand of user i at location j, is the predicted demand of user i at location j, β is the weight coefficient of supply pressure adjustment, P min and P max are the minimum and maximum limits for supply pressure;

[0288] Through the above, the reliability and safety of gas supply are improved, operating costs are reduced, and demand fluctuations and potential failure points are effectively dealt with, thereby improving the overall efficiency of the system and user satisfaction.

[0289] Real-time monitoring and user feedback in S6 include:

[0290] S61, Real-time monitoring mechanism established: Through the sensor network and smart gas meters, real-time gas usage data including pressure, flow and temperature are collected and transmitted to the monitoring center;

[0291] S62, user feedback information collection: collect user feedback information on gas supply through user feedback platforms (such as mobile applications or customer service systems), including gas use experience, safety hazards and suggestions;

[0292] S63, Data Fusion and Analysis: Real-time monitoring data D t User feedback information f Fusion, using data analysis algorithms to evaluate the effectiveness of the current gas supply strategy and identify potential problems, expressed as:

[0293] S eval =α∑ t D t +β∑ i U f ;

[0294] Among them, S eval is the evaluation score of the supply strategy, α and β are weight coefficients;

[0295] S64, dynamic adjustment: according to the evaluation score S eval Based on the potential problems identified, the adaptive optimization algorithm is used to dynamically adjust the gas supply strategy to ensure the safety and efficiency of the supply, which is expressed as:

[0296]

[0297] Among them, θ is the parameter of the gas supply strategy, η is the learning rate, L(θ,Seval ) is the loss function based on the evaluation score;

[0298] Through the above content, not only the real-time monitoring capability of gas supply is improved, but also user participation is enhanced, and user needs and feedback can be responded to in a timely manner, thereby optimizing the overall performance of the gas supply system and user satisfaction.

[0299] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0300] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A gas middle station data processing optimization method based on data analysis, characterized in that: The following steps are involved: S1, data collection and preprocessing: collect gas usage data in real time through the sensor network, including gas pressure, flow rate and temperature, and preprocess the collected gas usage data; S2, data analysis: analysis of pre-processed gas usage data, including trend analysis, anomaly detection, and demand forecasting; S3, gas pipeline network fault prediction: Fault prediction of the gas pipeline network, identification of weak links and potential fault points in the pipeline network, including: S31, feature extraction: extract key features from gas pipeline network data, including pressure change rate, flow fluctuation and temperature gradient; S32, model construction: construct a gas pipeline network fault prediction model based on the extracted key features to predict gas pipeline network faults; S33, real-time prediction: inputting real-time gas pipeline network data into the gas pipeline network fault prediction model to perform fault prediction and identify weak links and potential fault points in the pipeline network; S34, result verification: verify the accuracy of the prediction results by comparing with the actual fault records and inspection results, and continuously optimize and adjust the gas pipeline network fault prediction model; S4, user behavior analysis: Analyze users' gas usage behavior through big data and provide personalized services and suggestions, including smart gas usage suggestions and energy-saving solution push; S5, optimization processing: Based on the analysis results of gas usage data and user behavior, combined with the prediction results of the gas pipeline network, the gas supply strategy is optimized, including adjusting the supply pressure and optimizing the supply route; S6, real-time monitoring and user feedback: establish a real-time monitoring mechanism to monitor gas usage, collect user feedback information, and dynamically adjust the optimized gas supply strategy based on monitoring data and user feedback information.

2. The data processing optimization method for gas middle station based on data analysis according to claim 1 is characterized in that: The data collection and preprocessing in S1 include: S11, data collection: real-time collection of gas usage data, including: Air pressure data collection: Use pressure sensors to monitor the air pressure in the pipeline in real time; Flow data collection: Use flow meter sensor to measure the flow of gas; Temperature data collection: Use temperature sensors to monitor the temperature changes of gas; S12, data preprocessing: preprocessing the collected gas usage data, including data cleaning and data format conversion.

3. The data processing optimization method for gas middle station based on data analysis according to claim 1 is characterized in that: The data analysis in S2 includes: S21, trend analysis: Use linear regression model to perform trend analysis on the pre-processed gas usage data, expressed as: y = β0 + β1x; Among them, y is the predicted gas usage, x is time, β0 and β1 are regression coefficients; S22, anomaly detection: The seasonal autoregressive integrated moving average model is used to detect anomalies in the preprocessed gas usage data, expressed as: Among them, y t is the current data point, μ is a constant term, is the autoregressive coefficient, θ1 is the moving average coefficient, ∈ t is the error term, Φ1 and Θ1 are seasonal autoregressive and moving average coefficients, and s is the length of the season; S23, demand forecasting: Use the grey forecasting model to forecast the demand of pre-processed gas usage data, including: S231, data accumulation generation: the original data sequence x = {x(1), x(2), ..., x(n)} is accumulated and generated once to obtain a new data sequence X = {X(1), X(2), ..., X(n)}, which is expressed as: S232, establish differential equation: establish a grey differential equation based on the data sequence generated by accumulation, expressed as: Among them, a and b are parameters to be determined; S233, solving parameters: using the least square method to solve the unknown parameters a and b, expressed as: Define the matrix B and vector Y: Use the least squares method to solve for the unknown parameters: Among them, (B T B) 1 It is B T The inverse matrix of B, B T is the transpose of matrix B; S234, prediction value generation: Based on the parameters obtained, a prediction model is established and the prediction value of the original data is solved. It is expressed as: in, is the predicted value of the cumulative generated sequence at the k+1th moment, X(1) is the value of the cumulative generated sequence at the initial moment, e is the base of the natural logarithm, k is the time index, is the predicted value of the original data sequence at the k+1th moment, It is the predicted value of the cumulatively generated sequence at the kth moment.

4. The data processing optimization method for gas middle station based on data analysis according to claim 1 is characterized in that: The feature extraction in S31 includes: S311, Pressure change rate: Calculate the rate of change of pressure over time, expressed as: Where ΔP is the pressure change rate, P(t) is the pressure at the current moment, P(t-1) is the pressure at the previous moment, and Δt is the time interval; S312, flow fluctuation: Calculate the fluctuation amplitude of flow over time, expressed as: Among them, ΔF is the flow fluctuation, F i is the flow value at the i-th moment, is the average value of flow, n is the number of observations; S313, temperature gradient: calculate the temperature variation at different locations in the pipe network, expressed as: in, is the temperature gradient, T(x2) and T(x1) are the temperatures at the network locations x2 and x1, respectively, and x2-x1 is the distance between the two locations.

5. The data processing optimization method for gas middle station based on data analysis according to claim 4 is characterized in that: The gas pipeline network fault prediction model in S32 adopts a graph neural network model, and the graph neural network model includes: S321, graph representation and feature extraction: The gas pipeline network is represented as a graph G = (V, E), where nodes V represent gas equipment and sensors, and edges E represent pipeline connections. Key features are extracted from the gas pipeline network data as node attributes. The feature vector h of each node v is v Including pressure change rate ΔP, flow fluctuation ΔF and temperature gradient S322, node feature enhancement: introduce time feature t and position feature p as additional dimensions of node features, expressed as: S323, Message passing and update mechanism: Optimize the message passing function, consider the characteristics of the edge e uv The weighted sum of the node features is expressed as: Messaging: in, is the message of node v at level k, is the set of neighbor nodes of node v, α uv is the attention weight between node u and node v, W e and W h is the weight matrix, b h is bias; Node Update: in, is the feature representation of node v at layer k, W m is the message weight matrix, b m For message bias; S324, multi-layer stacking and global information fusion: introduce the global pooling layer to fuse the information of all nodes together for global feature extraction, expressed as: Among them, h G It is a global feature representation, and Pooling is average pooling; S325, fault prediction output: Use the fully connected layer to perform fault prediction and predict whether the node has a fault, expressed as: in, is the fault prediction value of node v, σ is the Sigmoid activation function, W o is the output layer weight matrix, b o is the bias for the output layer.

6. The data processing optimization method for gas middle station based on data analysis according to claim 5 is characterized in that: The real-time prediction in S33 includes: S331, real-time data input: the real-time collected gas pipeline network data is used as the node feature vector h v Input into the gas network fault prediction model; S332, forward propagation: Through the gas pipeline fault prediction model, the node feature vector h v Perform forward propagation calculations, pass through each layer of the graph neural network in turn, and update the node feature representation Until the Kth layer, the final node feature representation is obtained S333, Fault Prediction: Using the fully connected layer and activation function, the final node feature representation Fault prediction is performed, which is expressed as: in, is the fault prediction value of node v, σ is the Sigmoid activation function, W o is the output layer weight matrix, b o is the output layer bias; S334, Identification of weak links and potential failure points: Based on the failure prediction value Identify weak links and potential failure points in the gas pipeline network. When the set failure threshold is exceeded, node v is marked as a potential failure point.

7. The data processing optimization method for gas middle station based on data analysis according to claim 6 is characterized in that: The result verification in S34 includes: S341, prediction result collection: collect the fault prediction results of the gas pipeline network fault prediction model on the gas pipeline network data S342, actual fault record comparison: the predicted results The actual fault record v Compare and calculate the accuracy indicators of the predictions, including accuracy, precision, recall, and F1 score; S343, inspection result comparison: compare the prediction results with the inspection results to evaluate the effect of the model in practical application; S344, model optimization and adjustment: Based on the verification results, the model parameters are optimized and adjusted using the back propagation algorithm and gradient descent method, expressed as: Among them, θ is the model parameter, η is the learning rate, and L(θ) is the loss function.

8. The data processing optimization method for gas middle station based on data analysis according to claim 1 is characterized in that: The user behavior analysis in S4 includes: S41, data collection: The user's gas usage data is collected in real time through the smart gas meter and sensor network to form a user gas usage behavior data set D, which is expressed as: D={(u i ,t j ,x ij )}; Among them, u i For user i,t j is time j,x ij is the gas usage of user i at time j; S42, behavior pattern recognition: using K-means clustering algorithm to identify the user's gas usage behavior pattern; S43, Personalized service and suggestion generation: Based on the identified gas usage behavior patterns, including intelligent gas usage suggestions and energy-saving solution push, personalized services and suggestions are generated, including: S431, intelligent gas usage suggestions: Based on the user's behavior pattern, the Apriori algorithm is used to generate intelligent gas usage suggestions; S432, energy-saving solution push: Based on the user's behavior pattern and usage data, regression analysis is used to predict the user's gas demand and provide energy-saving solutions.

9. The method for optimizing gas station data processing based on data analysis according to claim 1 is characterized in that: The optimization process in S5 includes: S51, supply pressure adjustment: Based on the gas usage data and user behavior analysis results, combined with the prediction results of the gas pipeline network, the gas supply pressure is adjusted using an optimization algorithm, which is expressed as: Among them, P opt is the optimized supply pressure, P base is the basic supply pressure, α is the adjustment coefficient, w i is the weight of user i, d i is the actual demand of user i, is the predicted demand of user i; S52, supply route optimization: Based on the fault prediction results of the gas pipeline network, the flow distribution algorithm is used to optimize the gas supply route, which is expressed as: min∑ (i,j)∈E c ij f ij ; Constraints: Among them, c ij represents the transmission cost of edge (i, j), f ij represents the flow on edge (i,j), d i represents the demand of node i, and E represents the edge set; S53, comprehensive optimization: Combine the results of pressure adjustment and route optimization, comprehensively consider the cost and benefit, and use linear programming to formulate the optimal gas supply strategy, which is expressed as: Objective function: Constraints: Flow conservation constraints: Capacity constraints: Supply pressure constraint: P min ≤P opt ≤P max ; Among them, u ij is the capacity of edge (i,j), P opt is the optimized supply pressure, P base is the basic supply pressure, α is the adjustment coefficient, w ij is the weight of user i at position j, d ij is the actual demand of user i at location j, is the predicted demand of user i at location j, β is the weight coefficient of supply pressure adjustment, P min and P max are the minimum and maximum limits for supply pressure.

10. The data processing optimization method for gas middle station based on data analysis according to claim 1 is characterized in that: The real-time monitoring and user feedback in S6 include: S61, Real-time monitoring mechanism established: Through the sensor network and smart gas meters, real-time gas usage data including pressure, flow and temperature are collected and transmitted to the monitoring center; S62, user feedback information collection: collect user feedback information on gas supply through the user feedback platform, including gas use experience, safety hazards and suggestions; S63, Data Fusion and Analysis: Real-time monitoring data D t User feedback information f Fusion, using data analysis algorithms to evaluate the effectiveness of the current gas supply strategy and identify potential problems, expressed as: S eval =α∑ t D t +β∑ i U f ; Among them, S eval is the evaluation score of the supply strategy, α and β are weight coefficients; S64, dynamic adjustment: according to the evaluation score S eval Based on the potential problems identified, the gas supply strategy is dynamically adjusted using an adaptive optimization algorithm, which is expressed as: Among them, θ is the parameter of the gas supply strategy, η is the learning rate, L(θ,S eval ) is the loss function based on the evaluation score.

Images