Multi-data fusion energy-carbon terminal energy-saving control method and system

Abstract

The invention discloses a multi-data fusion energy-carbon terminal energy-saving control method and a multi-data fusion energy-carbon terminal energy-saving control system. The method comprises the following steps: carrying out standardized preprocessing on various kinds of collected energy consumption data to obtain standardized data; inputting the standardized data and the historical energy consumption data into a multi-modal feature extraction model, and outputting a current energy consumption feature vector and an abnormal energy consumption condition; inputting the energy consumption feature vector and the abnormal energy consumption condition into a reinforcement learning model combined with a reward function, and outputting an energy-saving control strategy set; and according to the energy-saving control strategy set, adjusting the energy utilization equipment. According to the method, various energy consumption data can be comprehensively considered, the optimal control strategy is dynamically generated, and the energy-saving control strategy is adjusted in time when the environment changes.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent energy-saving control, and more specifically, relates to a multi-data fusion energy-saving control method and system for carbon terminals. Background Art

[0002] In the field of energy monitoring, especially energy efficiency monitoring, existing technologies mainly focus on monitoring the consumption of electric energy data. These terminal devices acquire, analyze and process electric energy consumption by collecting electrical information. However, these technologies have obvious limitations: they lack the ability to fully monitor other energy consumption such as water and gas. This means that although electric energy consumption is monitored, it is not enough for a comprehensive energy management system to monitor only electric energy. An effective energy management system needs to be able to monitor and manage multiple energy consumptions at the same time, including electricity, water and gas, in order to achieve multi-data fusion and comprehensive energy management.

[0003] Existing energy monitoring terminals are not only limited in monitoring capabilities, but also show deficiencies in data analysis and the execution of energy-saving control strategies. These terminals mainly focus on the analysis of electric energy data, but are unable to integrate multiple types of data such as electricity, water, and gas for unified data analysis. This leads to difficulties in formulating and executing overall energy-saving control strategies due to the lack of comprehensive understanding and analysis of various energy consumption data.

[0004] The invention patent with the prior art publication number CN118966697A proposes a resource allocation strategy determination method, device, equipment, storage medium and program product, the method includes: obtaining an environmental feature vector and a particle optimization environment; constructing a first environmental feature based on the environmental feature vector and the particle optimization environment; initializing a preset particle swarm model with N resource allocation strategies and the first environmental feature; for each resource allocation strategy, iterative optimization is performed according to the first environmental feature, the preset reward function and the preset action function to obtain the first optimization position and the first optimization speed of the resource allocation strategy after each iteration; the fitness is calculated according to the first optimization position and the first optimization speed; the fitness of each resource allocation strategy is traversed to determine the first individual optimal solution and the first global optimal solution; and the resource allocation strategy corresponding to the first individual optimal solution and the first global optimal solution is determined. This scheme mainly relies on the park environment feature vector and the particle swarm optimization environment, and the input features are relatively single, making it difficult to comprehensively consider the complex multivariate influence. Summary of the invention

[0005] In order to overcome the problem in the prior art that it is unable to comprehensively monitor and analyze the consumption of multiple energy sources and formulate and implement effective energy-saving control strategies, the present invention provides a multi-data fusion energy-carbon terminal energy-saving control method and system.

[0006] The primary purpose of the present invention is to solve the above technical problems. The technical solution of the present invention is as follows:

[0007] The first aspect of the present invention provides a multi-data fusion energy-saving control method for an energy-carbon terminal, comprising the following steps:

[0008] Perform standardized preprocessing on the collected various energy consumption data to obtain standardized data;

[0009] Input the standardized data and historical energy consumption data into a multimodal feature extraction model, and output a current energy consumption feature vector and abnormal energy consumption conditions;

[0010] The energy consumption feature vector and abnormal energy usage are input into the reinforcement learning model combined with the reward function, and a set of energy-saving control strategies is output;

[0011] Adjust energy-consuming equipment according to the set of energy-saving control strategies.

[0012] Furthermore, the standardized preprocessing comprises the following steps:

[0013] Collect various energy consumption data;

[0014] Use the anomaly detection algorithm based on isolation forest to detect and clean the outliers in the collected data and output the cleaned data;

[0015] The cleaned data is formatted with the standard format template, mapped using a normalization algorithm, and the standardized data is output.

[0016] Furthermore, the standardized data is processed using a multimodal feature extraction model, including the following steps:

[0017] The standardized data is input into the Transformer-based multimodal feature extraction model to extract the single-modal features of various energy consumption data;

[0018] The extracted single-mode features are integrated and analyzed with the corresponding historical energy consumption data to obtain the energy consumption feature vector;

[0019] Based on the energy consumption feature vector and historical energy consumption data, the abnormal energy consumption is calculated and classified and output.

[0020] Furthermore, the reinforcement learning model is a deep deterministic policy gradient model.

[0021] Furthermore, the method for generating a set of energy-saving control strategies using a reinforcement learning model combined with a reward function comprises the following steps:

[0022] The rule engine is used to match energy consumption feature vectors and abnormal energy usage with multiple rules in the preset energy-saving control rule library, and output a preliminary control strategy set;

[0023] The preliminary control strategy set and energy consumption feature vector are input into the reinforcement learning model combined with the reward function, and the optimized control strategy set is output.

[0024] Furthermore, the method for generating a preliminary control strategy set using a preset energy-saving control rule base comprises the following steps:

[0025] Obtain a preset energy-saving control rule library, each energy-saving control rule includes a condition part and a control strategy part;

[0026] The energy-saving control rule base assigns a priority to each energy-saving control rule according to the importance of current equipment operation and energy-saving requirements;

[0027] Match the energy consumption feature vectors and abnormal energy consumption conditions with the rule conditions in the energy-saving control rule base one by one, and output the matching rules according to the priority of the rules;

[0028] Generate a corresponding preliminary control strategy set based on the control strategy of the matching rules.

[0029] Furthermore, the preliminary control strategy set is optimized using a reinforcement learning model combined with a reward function, including the following steps:

[0030] Input historical energy consumption data, equipment operation parameters, environmental parameters, equipment physical characteristics, and industry standards into the hybrid modeling tool and output the simulation environment; splice the energy consumption feature vector, equipment operation parameter vector, and environmental parameter vector into a high-dimensional state vector to obtain the current energy consumption state vector X t , where t represents the time step and the current energy state vector X t Combined with the energy consumption feature vector V t , get the environmental state S t , the formula is as follows:

[0031] S t =[V t ,X t ];

[0032] The initial control strategy set and environmental state S t Input the policy network of the reinforcement learning model and output the control action a at the current moment t , combined with the equipment operation constraints to correct the action and obtain the optimized action

[0033] Will optimize actions Applied to the simulation environment, the updated state vector X is obtained t+1 and energy consumption characteristic vector V t+1 ;

[0034] The updated environment state S t+1 =[V t+1 ,X t+1 ] and optimize actions Input the value network of the reinforcement learning model and update the value function Use the value function to adjust the parameters of the policy network;

[0035] Leverage Optimize Actions and the state vector X t+1 Combined with the reward function to calculate the reward value R t , use the reward value to update the parameters of the value network;

[0036] Repeat the above steps. The policy network outputs the optimization action according to the current environment state and applies it to the simulation environment to maximize the cumulative reward function value. By summarizing the optimization actions, the optimized control policy set is output.

[0037] Furthermore, the calculation formula of the reward function is:

[0038] R=αΔE+βΔW+γΔG+...+δC

[0039] Among them, ΔE, ΔW, ΔG, ... are the optimization actions to be performed The changes in various energy consumption data obtained later, α, β, γ, ..., δ are weight coefficients, C is the comfort bonus value, and the calculation formula is:

[0040]

[0041] Among them, T current and T target From the state vector X t+1 The current equipment operating parameters and target values ​​of the operating parameters are obtained, and σ is a sensitivity parameter. Further, the energy-consuming equipment is adjusted according to the energy-saving control strategy, including the following steps:

[0042] Convert the optimized energy-saving control strategy into control instructions for the equipment, including the start and stop time of the equipment and the adjustment of operating parameters;

[0043] Send control instructions to energy-consuming devices and execute them.

[0044] The second aspect of the present invention provides a multi-data fusion energy-carbon terminal energy-saving control system, which is used to implement the steps of a multi-data fusion energy-carbon terminal energy-saving control method, and the system includes:

[0045] The data preprocessing module is used to perform standardized preprocessing on the collected various energy consumption data to obtain standardized data;

[0046] A feature extraction module, used to input the standardized data and historical energy consumption data into a multimodal feature extraction model, and output a current energy consumption feature vector and abnormal energy consumption;

[0047] The energy-saving control strategy generation module is used to input the energy consumption feature vector and abnormal energy consumption into the reinforcement learning model combined with the reward function, and output the energy-saving control strategy set;

[0048] The intelligent control module is used to adjust the energy-consuming equipment according to the set of energy-saving control strategies.

[0049] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0050] The present invention combines the real-time energy consumption characteristics of multi-data fusion to capture the global state and dynamic changes of the system; through the policy network and value network of the reinforcement learning model, combined with the real-time environmental status and feedback data, it can dynamically generate the optimal control strategy and adjust the strategy in time when the environment changes; by setting the reward function, it can flexibly adjust the weights of energy saving, comfort and operating efficiency to support the dynamic balance of multiple objectives. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to make the purpose and technical solution of the present invention clearer, the present invention provides the following drawings and descriptions:

[0052] Figure 1 A flow chart of a method provided by an embodiment of the present invention;

[0053] Figure 2 A schematic diagram of system modules provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0054] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0055] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0056] Embodiment 1:

[0057] The present invention provides a multi-data fusion energy-saving control method for carbon terminals, such as Figure 1 The figure shows a flow chart of a multi-data fusion energy-saving control method for an energy-carbon terminal provided by the present invention, and the specific steps are as follows:

[0058] S1: Perform standardized preprocessing on the collected various energy consumption data to obtain standardized data.

[0059] In this embodiment, the various energy consumption data include electricity, water energy and gas energy. The electricity monitoring module can collect key parameters including electricity, voltage, current, power, power factor, etc. The water energy monitoring module can collect water resource consumption data. The gas energy monitoring module can monitor the gas consumption data of the gas pipeline. Combined with the real-time energy consumption characteristics of multi-data fusion, the system can capture the global status and dynamic changes.

[0060] The specific process is:

[0061] Collect various energy consumption data;

[0062] The outliers in the collected data are detected and cleaned using an anomaly detection algorithm based on isolation forest, and the cleaned data is output.

[0063] The process of anomaly detection and data cleaning through the isolation forest algorithm includes:

[0064] An isolation forest model is constructed based on the collected data to identify outliers.

[0065] Calculate the anomaly score and judge the anomaly based on the set threshold.

[0066] Remove data points marked as abnormal and keep normal data for subsequent processing.

[0067] The cleaned data is formatted with the standard format template, and the normalized data is mapped to the [0,1] interval using the normalization algorithm, and the standardized data is output.

[0068] S2: Input the standardized data and historical energy consumption data into a multimodal feature extraction model, and output a current energy consumption feature vector and abnormal energy consumption conditions.

[0069] The specific process is:

[0070] The standardized data is input into the Transformer-based multimodal feature extraction model to extract the single-modal features of various energy consumption data;

[0071] The extracted single-mode features are integrated and analyzed with the corresponding historical energy consumption data to obtain the energy consumption feature vector;

[0072] Based on the energy consumption feature vector and historical energy consumption data, the abnormal energy consumption is calculated and classified and output.

[0073] S3: Input the energy consumption feature vector and abnormal energy usage into the deep deterministic policy gradient model combined with the reward function, and output a set of energy-saving control strategies.

[0074] The specific process is:

[0075] The rule engine is used to match energy consumption feature vectors and abnormal energy usage with multiple rules in the preset energy-saving control rule library, and output a preliminary control strategy set.

[0076] The specific process is:

[0077] Obtain a preset energy-saving control rule library. Each energy-saving control rule includes a condition part and a control strategy part. In this embodiment, an example of a control strategy is as follows:

[0078] Power consumption strategy control: When the power load exceeds 100% of the set load, the energy carbon terminal can negotiate to reduce the output power of energy-consuming equipment such as air conditioning and lighting, or control and reduce the power output of non-critical circuit equipment.

[0079] Water use strategy control: When the historical average value exceeds a certain threshold in a certain period of time, the energy carbon terminal program can control the water-saving device through protocol to reduce the water flow of non-critical water-using equipment.

[0080] Gas usage strategy control: When the gas flow exceeds a certain threshold of the historical average value in a certain period of time and continues for a certain period of time, the Energy Carbon terminal program can control the operating status of the gas device by protocol.

[0081] The energy-saving control rule base assigns a priority to each energy-saving control rule according to the importance of current equipment operation and energy-saving requirements;

[0082] Match the energy consumption feature vectors and abnormal energy consumption conditions with the rule conditions in the energy-saving control rule base one by one, and output the matching rules according to the priority of the rules;

[0083] Generate a corresponding preliminary control strategy set based on the control strategy of the matching rules.

[0084] The preliminary control strategy set and energy consumption feature vector are input into the reinforcement learning model combined with the reward function, and the optimized control strategy set is output.

[0085] The specific process is:

[0086] Input historical energy consumption data, equipment operation parameters, environmental parameters, equipment physical characteristics, and industry standards into the hybrid modeling tool and output the simulation environment; splice the energy consumption feature vector, equipment operation parameter vector, and environmental parameter vector into a high-dimensional state vector to obtain the current energy consumption state vector X t , where t represents the time step and the current energy state vector X t Combined with the energy consumption feature vector V t , get the environmental state S t , the formula is as follows:

[0087] St =[V t ,X t ]

[0088] Environmental status S t Used to describe the overall situation of current equipment operation and energy consumption;

[0089] The initial control strategy set and environmental state S t Input the policy network of the reinforcement learning model and output the control action a at the current moment t , combined with the equipment operation constraints to correct the action and obtain the optimized action

[0090] Will optimize actions Applied to the simulation environment, the updated state vector X is obtained t+1 and energy consumption characteristic vector V t+1 ;

[0091] The updated environment state S t+1 =[V t+1 ,X t+1 ] and optimize actions Input the value network of the reinforcement learning model, the value network estimates the value of the current state, and updates the value function Use the value function to adjust the parameters of the policy network;

[0092] Leverage Optimize Actions and the state vector X t+1 Combined with the reward function to calculate the reward value R t , use the reward value to update the parameters of the value network;

[0093] Repeat the above steps, the policy network outputs the optimization action according to the current environment state and applies it to the simulation environment to maximize the cumulative reward function value. By summarizing the optimization actions, the optimized control strategy set is output. Through the policy network and value network of the reinforcement learning model, combined with the real-time environment state and feedback data, the optimal control strategy can be dynamically generated and the strategy can be adjusted in time when the environment changes. The calculation formula of the reward function is:

[0094] R=αΔE+βΔW+γΔG+...+δC

[0095] Among them, ΔE, ΔW, ΔG, ... are the optimization actions to be performed The change amount of various energy consumption data obtained later, α, β, γ, ..., δ are weight coefficients. In this embodiment, the calculation formula of the reward function is:

[0096] R=αΔE+βΔW+γΔG+δC

[0097] Among them, ΔE, ΔW, and ΔG are the optimization actions to be performed. The changes in electricity, water and gas energy consumption are obtained later. α, β, γ and δ are weight coefficients, which are set to 0.4, 0.3, 0.2 and 0.1 respectively. The importance of each indicator can be adjusted according to specific needs. C is the comfort bonus value, and the calculation formula is:

[0098]

[0099] Among them, T current and T target From the state vector X t+1 The current equipment operating parameters and target values ​​of the operating parameters are obtained. σ is the sensitivity parameter, which is set to 2. Through the reward function, the weights of energy saving, comfort and operating efficiency can be flexibly adjusted to support the dynamic balance of multiple objectives.

[0100] S4: Adjust the energy-consuming equipment according to the energy-saving control strategy set.

[0101] The specific process is:

[0102] The optimized energy-saving control strategy is converted into control instructions for the equipment, including the start and stop time of the equipment and the adjustment of operating parameters;

[0103] Send control instructions to energy-consuming devices and execute them.

[0104] Embodiment 2:

[0105] This embodiment provides a multi-data fusion energy-saving carbon terminal control system, such as Figure 2 As shown, a multi-data fusion energy-saving control method for energy-carbon terminals described in Example 1 is adopted, and the system includes:

[0106] The data preprocessing module is used to perform standardized preprocessing on the collected various energy consumption data to obtain standardized data;

[0107] A feature extraction module, used to input the standardized data and historical energy consumption data into a multimodal feature extraction model, and output a current energy consumption feature vector and abnormal energy consumption;

[0108] The energy-saving control strategy generation module is used to input the energy consumption feature vector and abnormal energy consumption into the reinforcement learning model combined with the reward function, and output the energy-saving control strategy set;

[0109] The intelligent control module is used to adjust the energy-consuming equipment according to the set of energy-saving control strategies.

[0110] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A multi-data fusion energy-saving control method for carbon terminals, characterized in that: The steps include: Perform standardized preprocessing on the collected various energy consumption data to obtain standardized data; Input the standardized data and historical energy consumption data into a multimodal feature extraction model, and output a current energy consumption feature vector and abnormal energy consumption conditions; The energy consumption feature vector and abnormal energy usage are input into the reinforcement learning model combined with the reward function, and a set of energy-saving control strategies is output; Adjust energy-consuming equipment according to the set of energy-saving control strategies.

2. According to the multi-data fusion energy-saving control method of carbon terminal according to claim 1, it is characterized in that: The standardized preprocessing comprises the following steps: Collect various energy consumption data; Use the anomaly detection algorithm based on isolation forest to detect and clean the outliers in the collected data and output the cleaned data; The cleaned data is formatted with the standard format template, mapped using a normalization algorithm, and the standardized data is output.

3. According to the multi-data fusion energy-saving control method of carbon terminal according to claim 1, it is characterized in that: The multimodal feature extraction model is used to process the standardized data and historical energy consumption data, including the following steps: The standardized data is input into the Transformer-based multimodal feature extraction model to extract the single-modal features of various energy consumption data; The extracted single-mode features are integrated and analyzed with the corresponding historical energy consumption data to obtain the energy consumption feature vector; Based on the energy consumption feature vector and historical energy consumption data, the abnormal energy consumption is calculated and classified and output.

4. According to the multi-data fusion energy-saving control method of carbon terminal according to claim 1, it is characterized in that: The reinforcement learning model is a deep deterministic policy gradient model.

5. The multi-data fusion energy-saving control method for carbon terminals according to claim 1 is characterized in that: The method for generating a set of energy-saving control strategies using a reinforcement learning model combined with a reward function comprises the following steps: The rule engine is used to match energy consumption feature vectors and abnormal energy usage with multiple rules in the preset energy-saving control rule library, and output a preliminary control strategy set; The preliminary control strategy set and energy consumption feature vector are input into the reinforcement learning model combined with the reward function, and the optimized control strategy set is output.

6. The multi-data fusion energy-saving control method for carbon terminals according to claim 5 is characterized in that: The method for generating a preliminary control strategy set using a preset energy-saving control rule library comprises the following steps: Obtain a preset energy-saving control rule library, each energy-saving control rule includes a condition part and a control strategy part; The energy-saving control rule base assigns a priority to each energy-saving control rule according to the importance of current equipment operation and energy-saving requirements; Match the energy consumption feature vectors and abnormal energy consumption conditions with the rule conditions in the energy-saving control rule base one by one, and output the matching rules according to the priority of the rules; Generate a corresponding preliminary control strategy set based on the control strategy of the matching rules.

7. The multi-data fusion energy-saving control method for carbon terminals according to claim 5 is characterized in that: The preliminary control strategy set is optimized using a reinforcement learning model combined with a reward function, including the following steps: Input historical energy consumption data, equipment operation parameters, environmental parameters, equipment physical characteristics, and industry standards into the hybrid modeling tool and output the simulation environment; splice the energy consumption feature vector, equipment operation parameter vector, and environmental parameter vector into a high-dimensional state vector to obtain the current energy consumption state vector X t , where t represents the time step and the current energy state vector X t Combined with the energy consumption feature vector V t , get the environmental state S t , the formula is as follows: S t =[V t ,X t ]; The initial control strategy set and environmental state S t Input the policy network of the reinforcement learning model and output the control action a at the current moment t , combined with the equipment operation constraints to correct the action and obtain the optimized action Will optimize actions Applied to the simulation environment, the updated state vector X is obtained t+1 and energy consumption characteristic vector V t+1 ; The updated environment state S t+1 =[V t+1 ,X t+1 ] and optimize actions Input the value network of the reinforcement learning model and update the value function Use the value function to adjust the parameters of the policy network; Leverage Optimize Actions and the state vector X t+1 Combined with the reward function to calculate the reward value R t , use the reward value to update the parameters of the value network; Repeat the above steps. The policy network outputs the optimization action according to the current environment state and applies it to the simulation environment to maximize the cumulative reward function value. By summarizing the optimization actions, the optimized control policy set is output.

8. The multi-data fusion energy-saving control method for carbon terminals according to claim 7 is characterized in that: The calculation formula of the reward function is: R=αΔE+βΔW+γΔG+...+δC Among them, ΔE, ΔW, ΔG, ... are the optimization actions to be performed The changes in various energy consumption data obtained later, α, β, γ, ..., δ are weight coefficients, C is the comfort bonus value, and the calculation formula is: Among them, T current and T target From the state vector X t+1 The current equipment operating parameters and target values ​​of the operating parameters are obtained, and σ is the sensitivity parameter.

9. The multi-data fusion energy-saving control method for carbon terminals according to claim 1 is characterized in that: The energy-consuming equipment is adjusted according to the energy-saving control strategy, including the following steps: Convert the optimized energy-saving control strategy into control instructions for the equipment, including the start and stop time of the equipment and the adjustment of operating parameters; Send control instructions to energy-consuming devices and execute them.

10. A multi-data fusion energy-carbon terminal energy-saving control system, using a multi-data fusion energy-carbon terminal energy-saving control method according to any one of claims 1 to 9, characterized in that: include: The data preprocessing module is used to perform standardized preprocessing on the collected various energy consumption data to obtain standardized data; A feature extraction module, used to input the standardized data and historical energy consumption data into a multimodal feature extraction model, and output a current energy consumption feature vector and abnormal energy consumption; The energy-saving control strategy generation module is used to input the energy consumption feature vector and abnormal energy consumption into the reinforcement learning model combined with the reward function, and output the energy-saving control strategy set; The intelligent control module is used to adjust the energy-consuming equipment according to the set of energy-saving control strategies.

Images