A method and system for energy monitoring based on big data

Abstract

The present invention discloses a method and system for energy monitoring based on big data, comprising the following steps: S1, constructing an original data set; S2, preprocessing the original data set; S3, constructing a dynamic energy balance coordination module, optimizing energy distribution through a contextual multi-armed tiger algorithm driven by big data; S4, establishing an adaptive scene generation module, using historical data to train a generative learning model, and outputting a scene control scheme; S5, building a nonlinear energy mapping and transmission optimization module, optimizing the energy transmission path and adjusting the energy flow; S6, developing a multi-dimensional environment coupling perception module, and constructing an association model; S7, constructing an event-triggered fine-tuning response module, and identifying emergencies using an anomaly detection algorithm based on big data classification and clustering; S8, constructing an energy dynamic scene interaction and visualization module, and generating a real-time three-dimensional visualization map. The present invention utilizes big data analysis and generative learning technology, etc., to achieve dynamic monitoring and control of energy systems.

Description

Technical Field

[0001] The present invention relates to the field of energy management and monitoring technology, and in particular to a big data-based energy monitoring method and system. Background Art

[0002] With the continuous growth of energy demand and the continuous optimization of energy structure, the complexity and dynamics of energy systems are increasing. Traditional energy monitoring and management technologies can no longer meet the refined requirements of modern energy systems. Under the current technical background, energy systems usually involve the coordination of multiple energy forms, such as electricity, heat and gas. Their distributed, multi-node characteristics and dynamic changes in environmental conditions further increase the difficulty of system management. Traditional energy monitoring methods usually rely on data collection and static analysis technology at fixed monitoring points, which makes it difficult to respond to dynamic changes and emergencies in complex energy systems in real time and accurately control them.

[0003] Existing energy management technologies mostly focus on data analysis and optimization control of a single energy type, and rarely involve the coordination and coupling relationship between multiple energy forms. For the energy allocation problem in distributed energy systems, traditional methods are usually based on linear optimization or static planning. These methods cannot adapt to the needs of rapid load changes, dynamic fluctuations in energy transmission losses, and real-time changes in environmental conditions in energy systems. At the same time, existing technologies usually only consider environmental factors in simple linear correlation analysis, ignoring the complex nonlinear relationship between environmental variables and energy use, such as the nonlinear effects of temperature, humidity, light intensity, wind speed and other factors on energy consumption, which limits the system's control efficiency and prediction capabilities.

[0004] In terms of event response, traditional energy management systems often lack effective detection and response mechanisms for emergencies. Existing anomaly detection technologies are mostly based on fixed thresholds or static rules, which are difficult to adapt to the complexity and diversity of emergencies in energy systems, such as sudden load fluctuations caused by equipment failures, extreme weather, or supply and demand imbalances. In addition, these technologies usually only passively handle abnormal events that have occurred, lacking active anomaly detection capabilities and dynamic response mechanisms, resulting in slow recovery of energy systems in abnormal states and low operating efficiency.

[0005] In terms of energy control strategy generation, existing technologies are usually based on a single analysis of historical data and lack the ability to intelligently predict and control future scenarios. Although some methods attempt to introduce simulation technology, existing simulation technologies mostly focus on static scene generation under fixed conditions, and it is difficult to generate diverse scenarios for complex dynamic environments and multiple load conditions. The generated control strategy is difficult to cover a variety of operating conditions such as normal, abnormal, and high load, resulting in insufficient operational stability and adaptability of the energy system.

[0006] In addition, in terms of data analysis and decision-making algorithms, the application of big data technology in existing methods mainly focuses on simple data statistics and linear regression models, and fails to fully utilize the advanced algorithms in big data technology to achieve accurate optimization of energy distribution and dynamic prediction of complex scenarios. At the same time, the research on the nonlinear characteristics of energy flow between nodes and the collaborative relationship of multiple nodes is relatively limited. It is difficult for existing technologies to accurately reflect the complex relationship between node characteristics, path efficiency and transmission loss through energy mapping functions, which further restricts the transmission optimization and resource allocation efficiency of energy systems.

[0007] Existing technologies also have significant deficiencies in data visualization and interactive functions. Traditional energy monitoring platforms mostly use two-dimensional plane graphics to display energy flows and node status, lacking the ability to intuitively display energy transmission paths, node loads, and control strategies. For decision makers of complex energy systems, this single static display format is difficult to provide sufficient reference information, thus limiting the efficiency of evaluating the effects of energy system control and adjusting strategies.

[0008] Therefore, how to provide an energy monitoring method and system based on big data is an urgent problem that technical personnel in this field need to solve. Summary of the invention

[0009] One purpose of the present invention is to propose a method and system for energy monitoring based on big data. The present invention makes full use of big data analysis technology, generative learning models and nonlinear optimization algorithms to systematically realize dynamic monitoring, intelligent prediction, regulation optimization and emergency response of energy systems. Through the deep fusion of multimodal data and the contextual multi-armed bandit algorithm driven by big data, the present invention can dynamically balance energy supply and demand, optimize transmission paths and adjust energy flows, comprehensively improve the real-time performance, regulation accuracy and system stability of energy management, and significantly enhance the energy system's ability to adapt to complex environmental conditions and abnormal situations. This method is widely applicable to distributed energy networks and multi-energy collaborative systems, and has extremely high practical value.

[0010] A method for energy monitoring based on big data according to an embodiment of the present invention comprises the following steps:

[0011] S1, deploy a multimodal sensor network to collect energy flow data and environmental data and construct the original data set;

[0012] S2. Preprocessing the original data set, including removing outliers, aligning time and space dimensions, integrating multimodal data, and optimizing the data storage structure using distributed storage technology to obtain a basic data set;

[0013] S3. Building a dynamic energy balance collaborative module based on the basic data set, optimizing energy allocation in real time through a contextual multi-armed bandit algorithm driven by big data, and generating an energy allocation plan;

[0014] S4. Establish an adaptive scenario generation module based on the energy allocation scheme, use historical data to train a generative learning model, construct typical scenarios under normal, abnormal and high-load conditions, and output scenario control schemes;

[0015] S5. Based on the scenario control scheme, a nonlinear energy mapping and transmission optimization module is built, an energy mapping function is constructed, the energy transmission path is optimized, and the energy flow is adjusted;

[0016] S6. Develop a multi-dimensional environmental coupling perception module to integrate environmental data and energy data, and use big data analysis technology to build a correlation model between environmental factors and energy use;

[0017] S7, build an event-triggered fine-tuning response module, combine real-time monitoring data and historical abnormal event data, and use anomaly detection algorithms based on big data classification and clustering to identify emergencies;

[0018] S8. Trigger a dynamic response mechanism according to the emergency event, and build an energy dynamic scene interaction and visualization module to generate a real-time three-dimensional visualization picture to display the energy flow path, node load and adjustment effect.

[0019] Optionally, the S3 specifically includes:

[0020] S31, construct a dynamic energy balance collaborative module based on the basic data set, and model the energy system as a multi-node network, where each node represents a key point of energy generation, storage or consumption; each edge E ij Represents node N i and node N j The energy flow relationship between them, the edge weight reflects the efficiency and loss of energy flow, and the dynamic feature matrix of nodes and edges is generated by combining historical data;

[0021] S32, extracting context information, wherein the context information includes node features, environmental features, and energy flow relationships, and constructing a context feature vector C t ={X t ,E t ,Env t}, where X t represents the node features at time step t, E t Indicates the environmental characteristics at time step t, Env t Represents the energy flow relationship at time step t;

[0022] S33. Define the context-based reward function R(a,t) to evaluate the distribution effect of action a at time step t:

[0023]

[0024] Among them, η(t) represents energy utilization efficiency, S(t) represents node load balance, U(t) represents environmental adaptability, L(t) represents energy transmission loss, E comp (t) represents the complexity cost of the allocation scheme, exp represents the exponential function, κ represents the adjustment coefficient, and ω represents the complexity weight parameter;

[0025] S34, based on the contextual multi-armed bandit algorithm driven by big data, construct the action space A = {a1, a2, ..., a m}, each action a i Represents an energy allocation strategy, selects the optimal action a * :

[0026]

[0027] in, represents the historical average reward of action a, γ represents the decay factor, λ represents the context exploration factor, and N(a) represents the number of selections of action a;

[0028] S35, based on the selected optimal action a * Adjust the node allocation strategy in a multi-node network, including reallocating energy flow paths between nodes and optimizing energy priorities of high-load nodes;

[0029] S36, monitor the adjusted node status and overall system performance, feed back the real-time node status to the basic data set, and update the context feature vector C t , through re-iterative optimization based on the contextual multi-armed bandit algorithm driven by big data, an energy allocation plan is finally generated.

[0030] Optionally, the S4 specifically includes:

[0031] S41, constructing an adaptive scenario generation module based on the energy allocation scheme, extracting time series features, spatial features and allocation weights between nodes in the energy allocation scheme, and combining energy usage features, load distribution features and environmental data in historical data to form a multidimensional data set;

[0032] S42, processing the multidimensional data set, dividing the data in the multidimensional data set into three categories: normal conditions, abnormal conditions, and high load conditions; each category specifically includes a node energy consumption trend dimension, a network load state dimension, and an environmental dynamic change dimension, extracting characteristic data of energy distribution, and generating a training sample set;

[0033] S43, introducing the actual energy allocation strategy under historical high-load conditions and the energy flow records under abnormal conditions, and using the historical data to train a generative learning model, the generative learning model includes an encoder and a decoder structure, the encoder compresses the historical data features into a latent feature space, retains key information and removes redundancy; the decoder reconstructs the target scene data from the latent feature space, so that the generated scene is consistent with the actual situation;

[0034] S44, optimizing the generative learning model in combination with the conditional information, and generating typical scenario data matching normal, abnormal, and high-load conditions based on different conditional constraints;

[0035] S45. Design scenario control solutions using typical scenario data generated by generative learning models; for normal scenarios, output energy usage strategies under stable operating conditions; for abnormal scenarios, provide energy redistribution suggestions; for high-load scenarios, generate load balancing strategies to avoid imbalance caused by overload of a single node;

[0036] S46. Output the generated scenario control plan, wherein the scenario control plan specifically includes a control strategy, an energy allocation ratio between nodes, and an expected load state description.

[0037] Optionally, the S5 specifically includes:

[0038] S51. Construct a nonlinear energy mapping and transmission optimization module based on the scenario control scheme, extract the generation, consumption and transmission characteristics of nodes in the energy system, including node load status, transmission path efficiency and coupling strength between nodes, and generate a dynamic relationship data set;

[0039] S52. Based on the dynamic relationship data set, a nonlinear energy mapping function is defined to describe the efficiency and loss relationship of energy in the transmission process, wherein the input is a node feature matrix and a transmission path feature matrix, and the input is a transmission efficiency matrix and a loss matrix between nodes, which is used to optimize the energy transmission path and adjust the energy flow;

[0040] S53. Based on the historical energy usage data and the transmission path performance data, the energy mapping function is modeled to establish the relationship between node characteristics, path characteristics and transmission efficiency:

[0041]

[0042] Among them, E ij represents the energy transmission efficiency from node i to node j, L ij represents the energy consumption from node i to node j, N i represents the characteristic information of node i, N j represents the characteristic information of node j, θ1, θ2 and θ3 represent exponential parameters, SE Represents the global parameters of the scene control scheme, P ij represents the transmission path characteristics between node i and node j, α E represents the scaling factor of transmission efficiency, β E represents the attenuation coefficient of the effect of node and path characteristics on efficiency, γ E Represents the weight coefficient of the global parameter on efficiency, δ L represents the influence coefficient of node characteristics on loss, λ L represents the adjustment coefficient of path characteristics on loss, η L represents the modulation factor of the path characteristic, exp represents the exponential function;

[0043] S54, optimizing the energy transmission path and adjusting the energy flow based on the energy mapping function, calculating the priority of each path, giving priority to a path combination with high transmission efficiency and low loss, and generating an optimal path table;

[0044] S55, dynamically updating the optimal path table, combining the real-time status of the node and the environmental conditions, capturing the energy fluctuation characteristics of the node and the path through a sliding window method, and dynamically adjusting the priority of the transmission path;

[0045] S56. Adjust the transmission scheme in the energy system based on the updated optimal path table. The transmission scheme specifically includes reallocating the energy flow direction between nodes, adjusting the transmission rate, and optimizing the node load distribution, and finally optimizing the energy transmission path and adjusting the energy flow.

[0046] Optionally, the S6 specifically includes:

[0047] S61. Based on the processed environmental data and energy data, a fusion data structure is constructed using a multi-dimensional environmental coupling perception module, environmental feature dimensions and energy use dimensions are defined, and a preliminary correlation data set is generated;

[0048] S62, performing data mining on the preliminary correlation data set, extracting key influencing factors of environmental data on energy use, constructing a multivariate feature vector and performing dimensionality reduction and feature engineering processing, retaining variables that have a significant impact on energy use, and removing redundant data;

[0049] S63. Based on historical data, a hierarchical modeling method is used to construct a correlation model between environmental factors and energy use, wherein the hierarchical modeling method includes an environmental factor layer, a dynamic relationship layer, and an output layer; the environmental factor layer is used to extract the single factor contribution of environmental characteristics to energy use; the dynamic relationship layer is used to capture the interaction between environmental factors and the dynamic impact on energy use; and the output layer is used to output the predicted value of energy use;

[0050] S64, using big data analysis technology to train the association model, by gradually optimizing the association model parameters, and adjusting the association model in combination with the dynamic change weights of environmental variables and the energy distribution ratio of historical data;

[0051] S65, real-time analysis of current environmental data, combined with the predictive capabilities of the correlation model, to generate energy usage trend data for future time periods, forming energy demand trends based on dynamic changes in environmental variables;

[0052] S66. Feedback the predicted energy demand trend to the energy system, form control strategy recommendations based on environmental data, and output optimized energy allocation plans.

[0053] Optionally, the S7 specifically includes:

[0054] S71. Construct an event-triggered fine-tuning response module to generate an abnormal event dataset D based on historical abnormal event data and real-time monitoring data. event ={d1,d2,…,d n}, each data point d i Including timestamp, anomaly type, node status and energy usage characteristics, and completing the time series and removing noise for the abnormal event data set;

[0055] S72, extracting event feature vector F event ={f1,f2,…,f m}, where each eigenvector f j Including node load change trends, energy flow status, environmental change parameters and abnormal duration, the features are mapped to a unified feature space through a standardized method to generate input data sets for classification and clustering;

[0056] S73. Use big data classification algorithms to perform preliminary detection of abnormal events, and use supervised learning models combined with context information to build dynamic classification functions:

[0057]

[0058] Among them, C(d i ) represents the data point d i Whether it belongs to the abnormal category, F event (t) represents the event feature vector at time t, m represents the dimension of the event feature vector, represents the feature weight parameter, W j represents the feature weight matrix, b j represents the feature bias, σ represents the activation function, ξ represents the historical anomaly influence weight, exp represents the exponential function, κ represents the regularization parameter, ∥F event (t)∥ represents the norm of the event feature vector, θ crepresents the time decay parameter, R(d i ,τ) represents the impact factor of abnormal events at time τ;

[0059] S74. Combine the big data clustering algorithm to perform detailed analysis on the detection results, use the density clustering algorithm to cluster the classified abnormal data points, aggregate the data points with similar event features, and output the abnormal event cluster G. event ={g1,g2,…,g k}, each cluster g k Indicates a class of similar abnormal events;

[0060] S75. Based on the comprehensive results of classification and clustering, calculate the event anomaly score and output the emergency event set.

[0061] A big data-based energy monitoring system according to an embodiment of the present invention includes the following modules:

[0062] Data collection and preprocessing module, used for data collection and preprocessing to generate basic data sets;

[0063] Dynamic energy balance collaboration module, used to optimize energy allocation through contextual multi-armed bandit algorithm and generate energy allocation scheme;

[0064] Adaptive scenario generation module, which uses generative learning models to build normal, abnormal, and high-load scenarios and output scenario control solutions;

[0065] Nonlinear energy mapping and transmission optimization module, which is used to construct energy mapping functions, optimize energy transmission paths and adjust energy flows;

[0066] Multi-dimensional environmental coupling perception module, which is used to integrate environmental data and energy data, build a correlation model between environmental factors and energy use, and predict energy demand trends;

[0067] An event-triggered fine-tuning response module is used to combine real-time monitoring data and historical abnormal event data, use classification and clustering algorithms to identify emergencies and trigger responses;

[0068] The energy dynamic scene interaction and visualization module is used to provide real-time three-dimensional visualization images to display energy flow paths, node loads and regulation effects.

[0069] The beneficial effects of the present invention are:

[0070] First, the present invention constructs a basic data set that can reflect the real-time status and environmental conditions of the energy system through the data acquisition and preprocessing module and the dynamic energy balance collaboration module, and realizes the dynamic energy balance of the multi-node distributed energy system using the contextual multi-armed bandit algorithm. The contextual multi-armed bandit algorithm combines node characteristics, environmental characteristics and energy flow relationships to generate the optimal allocation plan based on real-time supply and demand fluctuations, which not only improves energy utilization efficiency, but also effectively reduces the risk of transmission loss and node load imbalance. Compared with traditional static optimization methods, it has higher real-time performance and flexibility.

[0071] Secondly, the introduction of the adaptive scenario generation module and the nonlinear energy mapping and transmission optimization module gives the present invention significant advantages in scenario prediction and energy transmission optimization. The adaptive scenario generation module uses a generative learning model to construct typical scenarios under normal, abnormal and high-load conditions, and outputs refined control schemes based on these scenarios, thereby providing a scientific basis for the forward-looking control of energy systems. The nonlinear energy mapping and transmission optimization module reveals the complex nonlinear relationship between node characteristics, path efficiency and transmission loss by constructing an energy mapping function, and dynamically adjusts the energy transmission path, further improving the energy transmission efficiency and the overall coordination of system operation.

[0072] In addition, the multi-dimensional environment coupling perception module and the event-triggered fine-tuning response module significantly enhance the system's ability to adapt to environmental changes and emergencies. The multi-dimensional environment coupling perception module integrates environmental data and energy data, builds a correlation model between environmental factors and energy use, and combines big data analysis technology to accurately predict future energy demand trends, enabling the energy system to actively adapt to changes in the external environment and optimize control strategies. The event-triggered fine-tuning response module combines classification and clustering algorithms to identify emergency types in real time and trigger a dynamic response mechanism. It shortens the abnormal processing time by optimizing node status and energy allocation strategies, and improves the robustness and emergency response capabilities of the energy system.

[0073] Finally, the present invention provides a real-time three-dimensional visualization for energy management through the energy dynamic scene interaction and visualization module. Compared with the traditional two-dimensional static display, this module can intuitively display the energy flow path, node load and control strategy effect, providing an interactive operation platform for decision makers, which helps to improve the efficiency and accuracy of control strategy adjustment. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0075] Figure 1This is an overall flow chart of a big data-based energy monitoring method proposed by the present invention;

[0076] Figure 2 This is a structural schematic diagram of a big data-based energy monitoring system proposed in the present invention. DETAILED DESCRIPTION

[0077] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0078] refer to Figure 1 , a big data-based energy monitoring method, comprising the following steps:

[0079] S1, deploy a multimodal sensor network to collect energy flow data and environmental data and construct the original data set;

[0080] S2. Preprocessing the original data set, including removing outliers, aligning time and space dimensions, integrating multimodal data, and optimizing the data storage structure using distributed storage technology to obtain a basic data set;

[0081] S3. Building a dynamic energy balance collaborative module based on the basic data set, optimizing energy allocation in real time through a contextual multi-armed bandit algorithm driven by big data, and generating an energy allocation plan;

[0082] S4. Establish an adaptive scenario generation module based on the energy allocation scheme, use historical data to train a generative learning model, construct typical scenarios under normal, abnormal and high-load conditions, and output scenario control schemes;

[0083] S5. Based on the scenario control scheme, a nonlinear energy mapping and transmission optimization module is built, an energy mapping function is constructed, the energy transmission path is optimized, and the energy flow is adjusted;

[0084] S6. Develop a multi-dimensional environmental coupling perception module to integrate environmental data and energy data, and use big data analysis technology to build a correlation model between environmental factors and energy use;

[0085] S7, build an event-triggered fine-tuning response module, combine real-time monitoring data and historical abnormal event data, and use anomaly detection algorithms based on big data classification and clustering to identify emergencies;

[0086] S8. Trigger a dynamic response mechanism according to the emergency event, and build an energy dynamic scene interaction and visualization module to generate a real-time three-dimensional visualization picture to display the energy flow path, node load and adjustment effect.

[0087] In this implementation, S3 specifically includes:

[0088] S31, construct a dynamic energy balance collaborative module based on the basic data set, and model the energy system as a multi-node network, where each node represents a key point of energy generation, storage or consumption; each edge E ij Represents node N i and node N j The energy flow relationship between them, the edge weight reflects the efficiency and loss of energy flow, and the dynamic feature matrix of nodes and edges is generated by combining historical data;

[0089] S32, extracting context information, wherein the context information includes node features, environmental features, and energy flow relationships, and constructing a context feature vector C t ={X t ,E t ,Env t}, where X t represents the node features at time step t, E t Indicates the environmental characteristics at time step t, Env t Represents the energy flow relationship at time step t;

[0090] S33. Define the context-based reward function R(a,t) to evaluate the distribution effect of action a at time step t:

[0091]

[0092] Among them, η(t) represents energy utilization efficiency, S(t) represents node load balance, U(t) represents environmental adaptability, L(t) represents energy transmission loss, E comp (t) represents the complexity cost of the allocation scheme, exp represents the exponential function, κ represents the adjustment coefficient, and ω represents the complexity weight parameter;

[0093] S34, based on the contextual multi-armed bandit algorithm driven by big data, construct the action space A = {a1, a2, ..., a m}, each action a i Represents an energy allocation strategy, selects the optimal action a * :

[0094]

[0095] in, represents the historical average reward of action a, γ represents the decay factor, λ represents the context exploration factor, and N(a) represents the number of selections of action a;

[0096] S35, based on the selected optimal action a *Adjust the node allocation strategy in a multi-node network, including reallocating energy flow paths between nodes and optimizing energy priorities of high-load nodes;

[0097] S36, monitor the adjusted node status and overall system performance, feed back the real-time node status to the basic data set, and update the context feature vector C t , through re-iterative optimization based on the contextual multi-armed bandit algorithm driven by big data, an energy allocation plan is finally generated.

[0098] In this implementation manner, the S4 specifically includes:

[0099] S41, constructing an adaptive scenario generation module based on the energy allocation scheme, extracting time series features, spatial features and allocation weights between nodes in the energy allocation scheme, and combining energy usage features, load distribution features and environmental data in historical data to form a multidimensional data set;

[0100] S42, processing the multidimensional data set, dividing the data in the multidimensional data set into three categories: normal conditions, abnormal conditions, and high load conditions; each category specifically includes a node energy consumption trend dimension, a network load state dimension, and an environmental dynamic change dimension, extracting characteristic data of energy distribution, and generating a training sample set;

[0101] S43, introducing the actual energy allocation strategy under historical high-load conditions and the energy flow records under abnormal conditions, and using the historical data to train a generative learning model, the generative learning model includes an encoder and a decoder structure, the encoder compresses the historical data features into a latent feature space, retains key information and removes redundancy; the decoder reconstructs the target scene data from the latent feature space, so that the generated scene is consistent with the actual situation;

[0102] S44, optimizing the generative learning model in combination with the conditional information, and generating typical scenario data matching normal, abnormal, and high-load conditions based on different conditional constraints;

[0103] S45. Design scenario control solutions using typical scenario data generated by generative learning models; for normal scenarios, output energy usage strategies under stable operating conditions; for abnormal scenarios, provide energy redistribution suggestions; for high-load scenarios, generate load balancing strategies to avoid imbalance caused by overload of a single node;

[0104] S46. Output the generated scenario control plan, wherein the scenario control plan specifically includes a control strategy, an energy allocation ratio between nodes, and an expected load state description.

[0105] In this implementation manner, S5 specifically includes:

[0106] S51. Construct a nonlinear energy mapping and transmission optimization module based on the scenario control scheme, extract the generation, consumption and transmission characteristics of nodes in the energy system, including node load status, transmission path efficiency and coupling strength between nodes, and generate a dynamic relationship data set;

[0107] S52. Based on the dynamic relationship data set, a nonlinear energy mapping function is defined to describe the efficiency and loss relationship of energy in the transmission process, wherein the input is a node feature matrix and a transmission path feature matrix, and the input is a transmission efficiency matrix and a loss matrix between nodes, which is used to optimize the energy transmission path and adjust the energy flow;

[0108] S53. Based on the historical energy usage data and the transmission path performance data, the energy mapping function is modeled to establish the relationship between node characteristics, path characteristics and transmission efficiency:

[0109]

[0110] Among them, E ij represents the energy transmission efficiency from node i to node j, L ij represents the energy consumption from node i to node j, N i represents the characteristic information of node i, N j represents the characteristic information of node j, θ1, θ2 and θ3 represent exponential parameters, S E Represents the global parameters of the scene control scheme, P ij represents the transmission path characteristics between node i and node j, α E represents the scaling factor of transmission efficiency, β E represents the attenuation coefficient of the effect of node and path characteristics on efficiency, γ E Represents the weight coefficient of the global parameter on efficiency, δ L represents the influence coefficient of node characteristics on loss, λ L represents the adjustment coefficient of path characteristics on loss, η L represents the modulation factor of the path characteristic, exp represents the exponential function;

[0111] S54, optimizing the energy transmission path and adjusting the energy flow based on the energy mapping function, calculating the priority of each path, giving priority to a path combination with high transmission efficiency and low loss, and generating an optimal path table;

[0112] S55, dynamically updating the optimal path table, combining the real-time status of the node and the environmental conditions, capturing the energy fluctuation characteristics of the node and the path through a sliding window method, and dynamically adjusting the priority of the transmission path;

[0113] S56. Adjust the transmission scheme in the energy system based on the updated optimal path table. The transmission scheme specifically includes reallocating the energy flow direction between nodes, adjusting the transmission rate, and optimizing the node load distribution, and finally optimizing the energy transmission path and adjusting the energy flow.

[0114] In this implementation manner, S6 specifically includes:

[0115] S61. Based on the processed environmental data and energy data, a fusion data structure is constructed using a multi-dimensional environmental coupling perception module, environmental feature dimensions and energy use dimensions are defined, and a preliminary correlation data set is generated;

[0116] S62, performing data mining on the preliminary correlation data set, extracting key influencing factors of environmental data on energy use, constructing a multivariate feature vector and performing dimensionality reduction and feature engineering processing, retaining variables that have a significant impact on energy use, and removing redundant data;

[0117] S63. Based on historical data, a hierarchical modeling method is used to construct a correlation model between environmental factors and energy use, wherein the hierarchical modeling method includes an environmental factor layer, a dynamic relationship layer, and an output layer; the environmental factor layer is used to extract the single factor contribution of environmental characteristics to energy use; the dynamic relationship layer is used to capture the interaction between environmental factors and the dynamic impact on energy use; and the output layer is used to output the predicted value of energy use;

[0118] S64, using big data analysis technology to train the association model, by gradually optimizing the association model parameters, and adjusting the association model in combination with the dynamic change weights of environmental variables and the energy distribution ratio of historical data;

[0119] S65, real-time analysis of current environmental data, combined with the predictive capabilities of the correlation model, to generate energy usage trend data for future time periods, forming energy demand trends based on dynamic changes in environmental variables;

[0120] S66. Feedback the predicted energy demand trend to the energy system, form control strategy recommendations based on environmental data, and output optimized energy allocation plans.

[0121] In this implementation manner, the S7 specifically includes:

[0122] S71. Construct an event-triggered fine-tuning response module to generate an abnormal event dataset D based on historical abnormal event data and real-time monitoring data. event ={d1,d2,…,d n}, each data point d i Including timestamp, anomaly type, node status and energy usage characteristics, and completing the time series and removing noise for the abnormal event data set;

[0123] S72, extracting event feature vector F event ={f1,f2,…,f m}, where each eigenvector f j Including node load change trends, energy flow status, environmental change parameters and abnormal duration, the features are mapped to a unified feature space through a standardized method to generate input data sets for classification and clustering;

[0124] S73. Use big data classification algorithms to perform preliminary detection of abnormal events, and use supervised learning models combined with context information to build dynamic classification functions:

[0125]

[0126] Among them, C(d i ) represents the data point d i Whether it belongs to the abnormal category, F event (t) represents the event feature vector at time t, m represents the dimension of the event feature vector, represents the feature weight parameter, W j represents the feature weight matrix, b j represents the feature bias, σ represents the activation function, ξ represents the historical anomaly influence weight, exp represents the exponential function, κ represents the regularization parameter, ∥F event (t)∥ represents the norm of the event feature vector, θ c represents the time decay parameter, R(d i ,τ) represents the impact factor of abnormal events at time τ;

[0127] S74. Combine the big data clustering algorithm to perform detailed analysis on the detection results, use the density clustering algorithm to cluster the classified abnormal data points, aggregate the data points with similar event features, and output the abnormal event cluster G. event ={g1,g2,…,g k}, each cluster g k Indicates a class of similar abnormal events;

[0128] S75. Based on the comprehensive results of classification and clustering, calculate the event anomaly score and output the emergency event set.

[0129] refer to Figure 2 , an energy monitoring system based on big data, including the following modules:

[0130] Data collection and preprocessing module, used for data collection and preprocessing to generate basic data sets;

[0131] Dynamic energy balance collaboration module, used to optimize energy allocation through contextual multi-armed bandit algorithm and generate energy allocation scheme;

[0132] Adaptive scenario generation module, which uses generative learning models to build normal, abnormal, and high-load scenarios and output scenario control solutions;

[0133] Nonlinear energy mapping and transmission optimization module, which is used to construct energy mapping functions, optimize energy transmission paths and adjust energy flows;

[0134] Multi-dimensional environmental coupling perception module, which is used to integrate environmental data and energy data, build a correlation model between environmental factors and energy use, and predict energy demand trends;

[0135] An event-triggered fine-tuning response module is used to combine real-time monitoring data and historical abnormal event data, use classification and clustering algorithms to identify emergencies and trigger responses;

[0136] The energy dynamic scene interaction and visualization module is used to provide real-time three-dimensional visualization images to display energy flow paths, node loads and regulation effects.

[0137] Embodiment 1:

[0138] In order to verify the feasibility of the present invention in implementation, the present invention is applied to a certain industrial park. There are various types of energy in the industrial park, including electricity, thermal energy and gas. There are multiple distributed energy nodes (such as solar power generation devices, gas boilers and energy storage equipment) and a large number of energy-consuming equipment (such as production equipment, lighting systems and office electrical equipment) in the park. Since the energy use demand of each node is obviously dynamic and is affected by environmental changes (such as temperature, humidity and light intensity), it is difficult for traditional energy monitoring methods to grasp the overall energy flow status of the park in real time. In addition, the production load in the industrial park will have peaks and valleys due to order demand, and the probability of sudden failures during equipment operation is high. The accuracy and real-time performance of energy distribution and regulation have become the main management problems.

[0139] To meet this challenge, we introduced an energy monitoring and optimization system based on big data. The system deploys a multimodal sensor network to collect energy data (such as voltage, current, temperature) and environmental data (such as temperature and humidity, light intensity, wind speed) from each node in the park in real time. Through the data acquisition and preprocessing module, the collected raw data is cleaned and integrated to generate a basic data set to provide support for subsequent analysis.

[0140] In the energy management of the park, the dynamic energy balance coordination module analyzes the energy supply and demand relationship of multiple nodes in real time, and quickly generates energy allocation strategies through the contextual multi-armed bandit algorithm. For example, when a production workshop has a surge in electricity demand due to an increase in orders, the system identifies the load change of the node, and combines the load status of other nodes and the remaining capacity of the energy storage node to dynamically adjust energy allocation, giving priority to high-load nodes while reducing the power consumption of low-priority nodes (such as vacant offices). In the test, the module was able to complete energy redistribution within 5 seconds of a load surge, and the power supply voltage of the production workshop was stabilized within the range of 220±2V after adjustment.

[0141] In addition, the adaptive scenario generation module constructs three typical scenarios based on historical data: normal, abnormal, and high-load. For example, under high-temperature weather conditions in summer, the generative learning model simulates the high-load scenario caused by the large-scale use of air-conditioning equipment, and generates a load balancing strategy for the scenario, including giving priority to photovoltaic power generation devices and enabling backup gas-fired power generation equipment to balance the pressure on the power grid. In actual applications, the module starts the backup equipment 2 hours in advance during the predicted high-load period, ensuring the smooth operation of the energy system.

[0142] The nonlinear energy mapping and transmission optimization module further optimizes the energy flow path within the park. In the system test, when the power grid transmission line is congested, the system automatically optimizes the transmission path and switches the load from the high-loss path to the low-loss path by adjusting the energy flow direction. It has been verified that the transmission loss after optimization is reduced by 15%, and the average transmission efficiency between nodes is increased to 92%.

[0143] The multi-dimensional environmental coupling perception module integrates environmental data and energy usage data, and builds a correlation model between environmental factors and energy demand. In the past month, the system has detected that for every 1°C increase in temperature, the power consumption of the industrial cooling system increases by about 3%. Based on this relationship, the system predicts high temperature weather for the next day and adjusts the energy allocation strategy in advance, avoiding power shortages during peak hours in actual operation.

[0144] The event-triggered fine-tuning response module is used to respond to emergencies such as equipment failure or imbalance in energy supply and demand. In a simulation test, the main power supply equipment in a workshop suddenly failed. The system detected the abnormality within 1 second, switched to the backup energy supply through the dynamic response mechanism, and restored power within 3 seconds. After testing, the fluctuation of the power supply voltage of the workshop equipment after the backup plan was implemented was less than 1%.

[0145] Finally, the energy dynamic scene interaction and visualization module provides a three-dimensional real-time display of the energy flow in the park. On the visual interface of the control center, managers can clearly observe the real-time status of each node, the energy flow path, and the effect of system regulation, and further adjust the strategy according to the optimization suggestions provided by the interface. For example, when the load of a node is continuously higher than 90%, the system prompts the user on the interface to adjust the priority of the node to avoid potential risks.

[0146] Table 1 Comparison of average performance data between traditional system and system of the present invention

[0147]

[0148]

[0149] By analyzing the above Table 1, it can be clearly seen that the system of the present invention has significant advantages in data collection, processing speed, energy allocation adjustment capability, emergency response capability and system efficiency. The data collection frequency of the traditional system is only 2 times per minute, while the system of the present invention can achieve real-time data collection 60 times per minute, and the collection frequency has increased by 30 times, providing basic support for high-precision energy monitoring and management. At the same time, the traditional system has an average delay of 873 seconds in data processing, while the system of the present invention only takes 1.03 seconds, which greatly reduces the delay by 99.88%, enabling the system to analyze data in real time and respond quickly.

[0150] In terms of energy distribution adjustment, the traditional system takes an average of 32.67 seconds to complete the adjustment, while the system of the present invention only takes 4.5 seconds, shortening the adjustment time by 86.22%, significantly improving the response speed and dynamic adjustment capability of the energy system under load change conditions. Similarly, in terms of the response time of emergencies, the system of the present invention takes only 3 seconds from abnormal detection to response completion, while the traditional system takes an average of 37.33 seconds, increasing the response speed by 91.96%, reflecting the high efficiency of the present invention in emergency management.

[0151] In addition, in terms of the two core dimensions of transmission efficiency and energy loss rate, the system of the present invention also demonstrates strong optimization capabilities. The average transmission efficiency of the traditional system is 79%, while the system of the present invention is increased to 92%, an increase of 16.46%, effectively reducing the energy loss during transmission. The energy loss rate of the traditional system is 18.33%, while the system of the present invention reduces it to 7.83%, a reduction of 57.29%, significantly improving energy utilization.

[0152] In terms of high-load response capability, the system of the present invention performs particularly well. The response deviation of the traditional system after load balancing is 13.67%, while that of the system of the present invention is only 2.77%, a reduction of 79.72%, indicating that the present invention can better balance the energy load and avoid system instability caused by single-point overload.

[0153] In general, the system of the present invention overcomes the defects of traditional systems in data lag, slow response and waste of resources through real-time data collection, intelligent analysis and rapid response, significantly improves the accuracy, efficiency and adaptability of energy monitoring and management, and provides reliable guarantee for the optimized operation of complex energy systems.

[0154] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for energy monitoring based on big data, characterized in that: The steps include: S1, deploy a multimodal sensor network to collect energy flow data and environmental data and construct the original data set; S2. Preprocessing the original data set, including removing outliers, aligning time and space dimensions, integrating multimodal data, and optimizing the data storage structure using distributed storage technology to obtain a basic data set; S3. Building a dynamic energy balance collaborative module based on the basic data set, optimizing energy allocation in real time through a contextual multi-armed bandit algorithm driven by big data, and generating an energy allocation plan; S4. Establish an adaptive scenario generation module based on the energy allocation scheme, use historical data to train a generative learning model, construct typical scenarios under normal, abnormal and high-load conditions, and output scenario control schemes; S5. Based on the scenario control scheme, a nonlinear energy mapping and transmission optimization module is built, an energy mapping function is constructed, the energy transmission path is optimized, and the energy flow is adjusted; S6. Develop a multi-dimensional environmental coupling perception module to integrate environmental data and energy data, and use big data analysis technology to build a correlation model between environmental factors and energy use; S7. Build an event-triggered fine-tuning response module, combine real-time monitoring data and historical abnormal event data, and use anomaly detection algorithms based on big data classification and clustering to identify emergencies; S8. Trigger a dynamic response mechanism according to the emergency event, and build an energy dynamic scene interaction and visualization module to generate a real-time three-dimensional visualization picture to display the energy flow path, node load and adjustment effect.

2. The method for energy monitoring based on big data according to claim 1, characterized in that: The S3 specifically includes: S31, construct a dynamic energy balance collaborative module based on the basic data set, and model the energy system as a multi-node network, where each node represents a key point of energy generation, storage or consumption; each edge E ij Represents node N i and node N j The energy flow relationship between them, the edge weight reflects the efficiency and loss of energy flow, and the dynamic feature matrix of nodes and edges is generated by combining historical data; S32, extracting context information, wherein the context information includes node features, environmental features, and energy flow relationships, and constructing a context feature vector C t ={X t ,E t ,Env t }, where X t represents the node features at time step t, E t Indicates the environmental characteristics at time step t, Env t Represents the energy flow relationship at time step t; S33. Define the context-based reward function R(a,t) to evaluate the distribution effect of action a at time step t: Among them, η(t) represents energy utilization efficiency, S(t) represents node load balance, U(t0) represents environmental adaptability, L(t) represents energy transmission loss, E comp (t) represents the complexity cost of the allocation scheme, exp represents the exponential function, κ represents the adjustment coefficient, and ω represents the complexity weight parameter; S34, based on the contextual multi-armed bandit algorithm driven by big data, construct the action space A = {a1, a2, ..., a m }, each action a i Represents an energy allocation strategy, selects the optimal action a * : in, represents the historical average reward of action a, γ represents the decay factor, λ represents the context exploration factor, and N(a0 represents the number of selections of action a; S35, based on the selected optimal action a * Adjust the node allocation strategy in a multi-node network, including reallocating energy flow paths between nodes and optimizing energy priorities of high-load nodes; S36, monitor the adjusted node status and overall system performance, feed back the real-time node status to the basic data set, and update the context feature vector C t , through re-iterative optimization based on the contextual multi-armed bandit algorithm driven by big data, an energy allocation plan is finally generated.

3. The method for energy monitoring based on big data according to claim 1 is characterized in that: The S4 specifically includes: S41, constructing an adaptive scenario generation module based on the energy allocation scheme, extracting time series features, spatial features and allocation weights between nodes in the energy allocation scheme, and combining energy usage features, load distribution features and environmental data in historical data to form a multidimensional data set; S42, processing the multidimensional data set, dividing the data in the multidimensional data set into three categories: normal conditions, abnormal conditions, and high load conditions; each category specifically includes a node energy consumption trend dimension, a network load state dimension, and an environmental dynamic change dimension, extracting characteristic data of energy distribution, and generating a training sample set; S43, introducing the actual energy allocation strategy under historical high-load conditions and the energy flow records under abnormal conditions, and using the historical data to train a generative learning model, the generative learning model includes an encoder and a decoder structure, the encoder compresses the historical data features into a latent feature space, retains key information and removes redundancy; the decoder reconstructs the target scene data from the latent feature space, so that the generated scene is consistent with the actual situation; S44, optimizing the generative learning model in combination with the conditional information, and generating typical scenario data matching normal, abnormal, and high-load conditions based on different conditional constraints; S45. Design scenario control solutions using typical scenario data generated by generative learning models; for normal scenarios, output energy usage strategies under stable operating conditions; for abnormal scenarios, provide energy redistribution suggestions; for high-load scenarios, generate load balancing strategies to avoid imbalance caused by overload of a single node; S46. Output the generated scenario control plan, wherein the scenario control plan specifically includes a control strategy, an energy allocation ratio between nodes, and an expected load state description.

4. The method for energy monitoring based on big data according to claim 1, characterized in that: The S5 specifically includes: S51. Construct a nonlinear energy mapping and transmission optimization module based on the scenario control scheme, extract the generation, consumption and transmission characteristics of nodes in the energy system, including node load status, transmission path efficiency and coupling strength between nodes, and generate a dynamic relationship data set; S52. Based on the dynamic relationship data set, a nonlinear energy mapping function is defined to describe the efficiency and loss relationship of energy in the transmission process, wherein the input is a node feature matrix and a transmission path feature matrix, and the input is a transmission efficiency matrix and a loss matrix between nodes, which is used to optimize the energy transmission path and adjust the energy flow; S53. Based on the historical energy usage data and the transmission path performance data, the energy mapping function is modeled to establish the relationship between node characteristics, path characteristics and transmission efficiency: Among them, E ij represents the energy transmission efficiency from node i to node j, L ij represents the energy consumption from node i to node j, N i represents the characteristic information of node i, N j represents the characteristic information of node j, θ1, θ2 and θ3 represent exponential parameters, S E Represents the global parameters of the scene control scheme, P ij represents the transmission path characteristics between node i and node j, α E represents the scaling factor of transmission efficiency, β E represents the attenuation coefficient of the effect of node and path characteristics on efficiency, γ E Represents the weight coefficient of the global parameter on efficiency, δ L represents the influence coefficient of node characteristics on loss, λ L represents the adjustment coefficient of path characteristics on loss, η L represents the modulation factor of the path characteristic, exp represents the exponential function; S54, optimizing the energy transmission path and adjusting the energy flow based on the energy mapping function, calculating the priority of each path, giving priority to a path combination with high transmission efficiency and low loss, and generating an optimal path table; S55, dynamically updating the optimal path table, combining the real-time status of the node and the environmental conditions, capturing the energy fluctuation characteristics of the node and the path through a sliding window method, and dynamically adjusting the priority of the transmission path; S56. Adjust the transmission scheme in the energy system based on the updated optimal path table. The transmission scheme specifically includes reallocating the energy flow direction between nodes, adjusting the transmission rate, and optimizing the node load distribution, and finally optimizing the energy transmission path and adjusting the energy flow.

5. The method for energy monitoring based on big data according to claim 1 is characterized in that: The S6 specifically includes: S61. Based on the processed environmental data and energy data, a fusion data structure is constructed using a multi-dimensional environmental coupling perception module, the environmental feature dimension and the energy use dimension are defined, and a preliminary correlation data set is generated; S62, performing data mining on the preliminary correlation data set, extracting key influencing factors of environmental data on energy use, constructing a multivariate feature vector and performing dimensionality reduction and feature engineering processing, retaining variables that have a significant impact on energy use, and removing redundant data; S63. Based on historical data, a hierarchical modeling method is used to construct a correlation model between environmental factors and energy use, wherein the hierarchical modeling method includes an environmental factor layer, a dynamic relationship layer, and an output layer; the environmental factor layer is used to extract the single factor contribution of environmental characteristics to energy use; the dynamic relationship layer is used to capture the interaction between environmental factors and the dynamic impact on energy use; and the output layer is used to output the predicted value of energy use; S64, using big data analysis technology to train the association model, by gradually optimizing the association model parameters, and adjusting the association model in combination with the dynamic change weights of environmental variables and the energy distribution ratio of historical data; S65, real-time analysis of current environmental data, combined with the predictive capabilities of the correlation model, to generate energy usage trend data for future time periods, forming energy demand trends based on dynamic changes in environmental variables; S66. Feedback the predicted energy demand trend to the energy system, form control strategy recommendations based on environmental data, and output optimized energy allocation plans.

6. The method for energy monitoring based on big data according to claim 1, characterized in that: The S7 specifically includes: S71. Construct an event-triggered fine-tuning response module to generate an abnormal event dataset D based on historical abnormal event data and real-time monitoring data. event ={d1,d2,…,d n }, each data point d i Including timestamp, anomaly type, node status and energy usage characteristics, and completing the time series and removing noise for the abnormal event data set; S72, extracting event feature vector F event ={f1,f2,…,f m }, where each eigenvector f j Including node load change trends, energy flow status, environmental change parameters and abnormal duration, the features are mapped to a unified feature space through a standardized method to generate input data sets for classification and clustering; S73. Use big data classification algorithms to perform preliminary detection of abnormal events, and use supervised learning models combined with context information to build dynamic classification functions: Among them, C(d i ) represents the data point d i Whether it belongs to the abnormal category, F event (t) represents the event feature vector at time t, m represents the dimension of the event feature vector, represents the feature weight parameter, W j represents the feature weight matrix, b j represents the feature bias, σ represents the activation function, ξ represents the historical anomaly influence weight, exp represents the exponential function, κ represents the regularization parameter, ∥F event (t)∥ represents the norm of the event feature vector, θ c represents the time decay parameter, R(d i ,τ) represents the impact factor of abnormal events at time τ; S74. Combine the big data clustering algorithm to perform detailed analysis on the detection results, use the density clustering algorithm to cluster the classified abnormal data points, aggregate the data points with similar event features, and output the abnormal event cluster G. event ={g1,g2,…,g k }, each cluster g k Indicates a class of similar abnormal events; S75. Based on the comprehensive results of classification and clustering, calculate the event anomaly score and output the emergency event set.

7. A big data-based energy monitoring system, which executes the big data-based energy monitoring method according to any one of claims 1 to 6, characterized in that: Includes the following modules: Data collection and preprocessing module, used for data collection and preprocessing to generate basic data sets; Dynamic energy balance collaboration module, used to optimize energy allocation through contextual multi-armed bandit algorithm and generate energy allocation scheme; Adaptive scenario generation module, which uses generative learning models to build normal, abnormal, and high-load scenarios and output scenario control solutions; Nonlinear energy mapping and transmission optimization module, which is used to construct energy mapping functions, optimize energy transmission paths and adjust energy flows; Multi-dimensional environmental coupling perception module, which is used to integrate environmental data and energy data, build a correlation model between environmental factors and energy use, and predict energy demand trends; An event-triggered fine-tuning response module is used to combine real-time monitoring data and historical abnormal event data, use classification and clustering algorithms to identify emergencies and trigger responses; The energy dynamic scene interaction and visualization module is used to provide real-time three-dimensional visualization images to display energy flow paths, node loads and regulation effects.

Images