Multi-granularity energy consumption characteristic parameter decoupling method and system based on data driving

Abstract

The invention provides a multi-granularity energy consumption characteristic parameter decoupling method and system based on data driving, and the method comprises the steps: providing an energy consumption prediction model based on a deep neural network, taking multi-dimensional data in an actual operation period as the input of the energy consumption prediction model, and outputting energy flows of different industrial loads; an energy decoupling rule is defined, energy flows of different industrial loads are processed layer by layer based on the energy decoupling rule, a total cost minimization constraint is set, value functions of the different industrial loads are calculated through the total cost minimization constraint, and a control sequence is obtained by distributing loads among different energy sources; based on value functions of different industrial loads, establishing a global energy management function, defining control variables, and bringing system energy consumption and cost into an optimization function; and establishing an energy efficiency calculation model, quantifying the energy efficiency of each link, and realizing multi-granularity energy consumption characteristic parameter decoupling. According to the invention, the operation condition of the whole system can be accurately predicted, and real-time and effective energy distribution is realized.

Description

Technical Field

[0001] This invention relates to the field of energy management and optimization technology, and in particular to a data-driven method and system for decoupling multi-granularity energy consumption characteristic parameters, as well as a corresponding computer terminal and computer-readable storage medium. Background Technology

[0002] With the growth of global energy demand and the widespread application of renewable energy, efficient and intelligent management and optimization of energy use has become a key technology. Traditional energy management systems typically rely on static, rule-based methods, which have significant limitations when dealing with complex and dynamic energy demands. For example, traditional load forecasting and scheduling methods often fail to accurately capture changes in user behavior and the impact of environmental factors, resulting in large forecasting errors and poor optimization performance.

[0003] In recent years, with the advancement of information technology, data-driven methods have provided new insights into solving these problems. By analyzing large amounts of historical data, patterns and regularities hidden within the data can be discovered and utilized, thereby improving the accuracy of predictions and optimizing energy dispatch. However, existing methods still face some challenges in flexible resource energy modeling: high-dimensional data processing is difficult; traditional methods have high computational complexity when dealing with large-scale, high-dimensional energy data, making it difficult to achieve real-time or near-real-time optimization; the model generalization ability is insufficient; when encountering new or unseen data, traditional models have poor generalization ability and are prone to prediction bias; and user differences are ignored; existing modeling methods often overlook the differences and diversity among different users, making it difficult to achieve personalized energy management. Summary of the Invention

[0004] To address the aforementioned shortcomings in the prior art, this invention provides a data-driven method and system for decoupling multi-granularity energy consumption characteristic parameters, along with a corresponding computer terminal and computer-readable storage medium.

[0005] According to one aspect of the present invention, a data-driven decoupling method for multi-granularity energy consumption characteristic parameters is provided, characterized in that it includes:

[0006] A deep neural network-based energy consumption prediction model is provided, which takes multi-dimensional data during the actual operating cycle as input to the energy consumption prediction model and outputs energy flow under different industrial loads.

[0007] Define an energy decoupling rule, process the energy flow of different industrial loads layer by layer based on the energy decoupling rule, set a total cost minimization constraint, calculate the value function of different industrial loads through the total cost minimization constraint, and obtain the control sequence by allocating loads among different energy sources;

[0008] Based on the value functions of different industrial loads, a global energy management function is established, and the control variables of the global energy management function are defined. The energy consumption and cost of each industrial load are integrated to obtain the total energy input.

[0009] An energy efficiency calculation model is established. Using the control variables and total energy input, the effective output power of each industrial load is quantified through the calculation model, thereby achieving decoupling of multi-granularity energy consumption characteristic parameters.

[0010] According to another aspect of the present invention, a data-driven multi-granularity energy consumption characteristic parameter decoupling system is provided, comprising:

[0011] The energy consumption parameter prediction module provides an energy consumption prediction model based on a deep neural network. It takes multi-dimensional data from the actual operating cycle as input to the energy consumption prediction model and outputs the energy flow of different industrial loads.

[0012] An energy decoupling module is used to define an energy decoupling rule, process the energy flow of different industrial loads layer by layer based on the energy decoupling rule, set a total cost minimization constraint, calculate the value function of different industrial loads through the total cost minimization constraint, and obtain a control sequence by allocating loads among different energy sources.

[0013] The energy management module establishes a global energy management function based on the value function of different industrial loads, defines the control variables of the global energy management function, integrates the energy consumption and cost of each industrial load, and obtains the total energy input.

[0014] The energy efficiency assessment module is used to establish an energy efficiency calculation model. Using the control variables and total energy input, the calculation model quantifies the effective output power of each industrial load, thereby achieving decoupling of multi-granularity energy consumption characteristic parameters.

[0015] According to a third aspect of the present invention, a computer terminal is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it can be used to perform the methods described above in the present invention, or to run the system described above in the present invention.

[0016] According to a fourth aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, can be used to perform the methods described above in the present invention, or to run the system described above in the present invention.

[0017] By adopting the above technical solution, the present invention has at least one of the following beneficial effects compared with the prior art:

[0018] The present invention provides a data-driven multi-granularity energy consumption characteristic parameter decoupling method and system. By constructing a multi-granularity data-driven prediction model containing a convolutional neural network and a long short-term memory layer, it solves the problem that traditional energy management systems, which rely on static rules, have difficulty capturing dynamic load characteristics and environmental factors. This achieves the effect of high-precision prediction of industrial load energy demand and significantly improves the prediction accuracy of system operation status.

[0019] The present invention provides a data-driven multi-granularity energy consumption characteristic parameter decoupling method and system. By defining energy decoupling rules based on dynamic programming and hierarchical optimization strategies, it solves the problems of high computational complexity and difficulty in real-time response in the processing of complex energy flows by traditional methods. It realizes dynamic decoupling and optimal allocation of energy flows of different industrial loads, effectively reducing the total system cost and improving energy dispatching efficiency.

[0020] The data-driven multi-granularity energy consumption characteristic parameter decoupling method and system provided by this invention solves the limitations of traditional optimization models that ignore user differences and multi-objective collaboration by establishing a global energy management function and integrating multi-load value functions and weight coefficients. It achieves a comprehensive optimization effect that coordinates system energy consumption, cost and equipment life, and significantly improves the rationality of energy allocation and overall efficiency. Attached Figure Description

[0021] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0022] Figure 1 This is a flowchart illustrating the workflow of a data-driven multi-granularity energy consumption characteristic parameter decoupling method in a preferred embodiment of the present invention.

[0023] Figure 2 This is a schematic diagram of the constituent modules of a data-driven multi-granularity energy consumption characteristic parameter decoupling system in a preferred embodiment of the present invention.

[0024] Figure 3 These are the MSE and RMSE values ​​between the predicted load power and the actual load power in a specific application example of the present invention.

[0025] Figure 4 This is a comparison curve of the SOC change of the present invention with that of the traditional method in a specific application example.

[0026] Figure 5 This is a comparison chart of the power demand during the testing cycle of the present invention and the traditional method in a specific application example.

[0027] Figure 6 The graph shows the energy efficiency results of the present invention and the traditional method under different cycles in a specific application example. Detailed Implementation

[0028] The embodiments of the present invention are described in detail below: These embodiments are implemented based on the technical solution of the present invention, and provide detailed implementation methods and specific operation processes. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention.

[0029] Existing technologies often suffer from difficulties in processing high-dimensional data, insufficient model generalization ability, and neglect of user differences in flexible resource energy consumption modeling. To address these issues, one embodiment of the present invention provides a data-driven, multi-granularity energy consumption characteristic parameter decoupling method. This method preprocesses multi-dimensional data from the actual operating cycle, meticulously analyzes historical data, extracts key features, and establishes a more accurate prediction model. Simultaneously, it employs a multi-granularity analysis method to perform hierarchical decoupling of load behavior, reducing computational complexity and improving real-time optimization capabilities.

[0030] Specifically, such as Figure 1 As shown, the data-driven multi-granularity energy consumption characteristic parameter decoupling method provided in this embodiment may include:

[0031] S1 provides an energy consumption prediction model based on a deep neural network, which takes multi-dimensional data during the actual operating cycle as input to the energy consumption prediction model and outputs the energy flow of different industrial loads.

[0032] S2 defines an energy decoupling rule, processes the energy flow of different industrial loads layer by layer based on the energy decoupling rule, sets a total cost minimization constraint, calculates the value function of different industrial loads through the total cost minimization constraint, and obtains the control sequence by allocating loads among different energy sources;

[0033] S3, based on the value functions of different industrial loads, establishes a global energy management function, defines the control variables of the global energy management function, integrates the energy consumption and cost of each industrial load, and obtains the total energy input;

[0034] S4. Establish an energy efficiency calculation model, utilize control variables and total energy input, and quantify the effective output power of each industrial load through the calculation model to achieve decoupling of multi-granularity energy consumption characteristic parameters.

[0035] In some preferred embodiments, S1 above provides a deep neural network energy prediction model including a convolutional neural network and a long short-term memory layer, which uses multi-dimensional data from the actual operating cycle as input to the prediction model and outputs energy flows for different industrial loads. It may further include:

[0036] S11, acquire multi-dimensional time series data within the actual operating cycle, including: industrial load status information, environmental conditions and operating parameters;

[0037] S12, preprocessing multi-dimensional time series data;

[0038] S13 provides a deep neural network model containing convolutional neural network layers and long short-term memory layers. The preprocessed data is input into the neural network model to identify the operating parameters of different industrial loads, predict future energy consumption, and output the energy flow of different industrial loads.

[0039] In some preferred embodiments, the industrial load status information in S11 above may further include:

[0040] Electrolytic aluminum, including: current intensity I Al Efficiency factor η Al and total power demand P Al ;

[0041] Electric heating, including: set temperature T set Ambient temperature T env and total power demand P Heat ;

[0042] Energy storage system, including: charging and discharging current I bat Charging and discharging power P bat and State of charge (SOC) k ;

[0043] Environmental conditions, including temperature and humidity;

[0044] Operating parameters include: equipment operating mode.

[0045] In some preferred embodiments, S12 above defines an energy decoupling rule, processes the energy flow of different industrial loads layer by layer based on the energy decoupling rule, sets a total cost minimization constraint, calculates the value function of different industrial loads through the total cost minimization constraint, and obtains a control sequence by allocating loads among different energy sources. It may further include:

[0046] S21 defines an online energy decoupling rule, including:

[0047] Based on the Bellman optimality principle, an independent value function and constraints are defined for each industrial load. The optimal control variables for each industrial load are solved step by step through discrete-time dynamic programming. Finally, a dynamic control command sequence that satisfies the minimization of global total cost is generated. The command sequence includes a set of industrial load operating parameters that are dynamically adjusted according to the time step.

[0048] S22, based on energy decoupling rules, processes the energy flow of different industrial loads layer by layer, including:

[0049] The system is decomposed into multiple independent load layers such as electrolytic aluminum, electric heating, and energy storage. The optimal control variables for each layer are calculated independently based on dynamic programming according to the energy decoupling rules.

[0050] S23, based on the processed energy flow, guides the optimal operation of each industrial load according to the discrete-time dynamic programming optimization solver of Bellman's principle, defines the minimum cost constraint, and calculates the value function for different industrial loads by minimizing the cost constraint;

[0051] S24, based on the value function calculation results, allocates industrial loads among different energy sources to obtain control sequences.

[0052] In some preferred embodiments, the above-mentioned S23, which calculates the value function for different industrial loads by minimizing cost constraints, further includes any one or more of the following:

[0053] S231, Calculates the value function of the electrolytic aluminum industry load by minimizing cost constraints, including:

[0054]

[0055] In the formula, J Al The value function representing the load of the electrolytic aluminum industry; u k,Al =[η Al,k [x] is a control variable, representing the efficiency factor of the electrolyzer in stage k; k,Al L is a state variable representing the total power demand of the electrolytic aluminum load at time t; Al P is the instantaneous cost function used to measure energy consumption at each stage; elec,Al Instantaneous power consumption is represented by the efficiency factor η. Al and state variable x k,Al The decision is made by N, which represents the total number of stages in the discrete input sequence, i.e., the number of time steps for prediction.

[0056] The load constraints for electrolytic aluminum are established as follows:

[0057] I Al,min ≤I Al,k ≤I Al,max

[0058] In the formula, I Al,k I represents the current intensity in the k-th stage; Al,min Indicates the lower limit of the safe range for current intensity; I Al,max Indicates the upper limit of the safe range for current intensity;

[0059] S232, calculates the value function of industrial load for electric heating by minimizing cost constraints, including:

[0060]

[0061] In the formula, J Heat The value function representing the industrial load of electric heating; u k,Heat =[T set,k [x] is the control variable, representing the set temperature in stage k; k,Heat L is a state variable representing the total power demand of the electric heating load at time t; Heat P is the instantaneous cost function used to measure energy consumption at each stage; elec,Heat Indicates instantaneous power consumption, determined by the set temperature T. set The decision is made by N, which represents the total number of stages in the discrete input sequence, i.e., the number of time steps for prediction.

[0062] The constraints for establishing the electric heating load are as follows:

[0063] T set,min ≤T set,k ≤T set,max

[0064] T env,min ≤T env,k ≤T env,max

[0065] In the formula, T set,k T represents the set temperature for stage k; set,min This indicates the lower limit of the reasonable temperature range; T set,max This indicates the upper limit of the reasonable temperature range; T env Indicates ambient temperature; T env,min Indicates the lower limit of ambient temperature; T env,max Indicates the upper limit of ambient temperature;

[0066] S233, calculating the value function of the energy storage system load by minimizing cost constraints, including:

[0067]

[0068] In the formula, I Storage Cost represents the value function of the energy storage system load. ele Cost represents the energy consumed per kilowatt-hour. bat The aging cost of an energy storage system represents the cost increase per charge-discharge cycle. ele and cost bat The value is derived from actual operational data; N is the total number of stages in the discrete input sequence, i.e., the number of time steps for prediction; x k,storageThe state variable represents the current state of the energy storage system at time t, i.e., the state of charge (SOC); u k,storage =[P bat [k] is a control variable, representing the charging and discharging power in the k-th stage;

[0069] The constraints for establishing the load of the energy storage system are:

[0070] I bat,min ≤I bat,k ≤I bat,max

[0071] P bat,min ≤P bat,k ≤P bat,max

[0072] 0≤SOC k ≤SOC max

[0073] In the formula, I bat,k I represents the charging and discharging current in the k-th stage; bat,min Indicates the lower limit of the safe range for charging and discharging current; I bat,max Indicates the upper limit of the safe range of charging and discharging current; P bat,k P represents the charging and discharging power in the k-th stage; bat,min Indicates the lower limit of the device's permissible range; P bat,max Indicates the upper limit of the device's allowed range; SOC k Represents the state of charge at stage k; soc max This indicates the upper limit of the state of charge.

[0074] In some preferred embodiments, S3 above, based on the value functions of different industrial loads, establishes a global energy management function and defines the control variables of the global energy management function, integrates the energy consumption and cost of each industrial load, and obtains the total energy input. It may further include:

[0075] S31, Establish a global energy management function, expressed as:

[0076]

[0077] In the formula, J glabool This represents the global energy management function; M is the total industrial load; w i J represents the weighting coefficient, indicating the importance of each industrial load; i The value function for each industrial load;

[0078] S32 defines the control variables of the global energy management function, including: the state information of industrial loads, environmental conditions and operating parameters, so that the energy consumption and cost of each industrial load are reflected in the value function of the global energy management function; according to the weight coefficients, the energy consumption and cost of each industrial load are integrated to obtain the total energy input.

[0079] In some preferred embodiments, S4 above, establishing an energy efficiency calculation model, and using the control variables and total energy input, quantifying the effective output power of each industrial load through the calculation model, may further include:

[0080] The energy efficiency calculation model is expressed as follows:

[0081] Total energy efficiency index η global Calculated using the following formula:

[0082]

[0083] In the formula, η global P represents the overall energy efficiency index; output,i (t) represents the effective output power of the i-th industrial load; η sys,i P represents the operating efficiency coefficient of the corresponding subsystem; input,j (t) represents the total power of the j-th energy input; M represents the total number of industrial loads; R represents the number of energy input types.

[0084] In S4 above, the effective output power of each industrial load is quantified through a calculation model, that is, the energy efficiency of each link is quantified; where each link refers to different load types and energy conversion stages in the industrial energy system, mainly including the electrolytic aluminum production link, the electric heating operation link, and the energy storage system charging and discharging link.

[0085] Based on the same inventive concept, an embodiment of the present invention also provides a data-driven multi-granularity energy consumption characteristic parameter decoupling system.

[0086] Specifically, such as Figure 2 As shown, the data-driven multi-granularity energy consumption characteristic parameter decoupling system provided in this embodiment may include:

[0087] The energy consumption parameter prediction module provides an energy consumption prediction model based on a deep neural network. It takes multi-dimensional data from the actual operating cycle as input to the energy consumption prediction model and outputs the energy flow of different industrial loads.

[0088] The energy decoupling module defines an energy decoupling rule, processes the energy flow of different industrial loads layer by layer based on the energy decoupling rule, sets a total cost minimization constraint, calculates the value function of different industrial loads through the total cost minimization constraint, and obtains a control sequence by allocating loads among different energy sources.

[0089] The energy management module establishes a global energy management function based on the value function of different industrial loads, defines the control variables of the global energy management function, integrates the energy consumption and cost of each industrial load, and obtains the total energy input.

[0090] The energy efficiency assessment module is used to establish an energy efficiency calculation model. By using control variables and total energy input, the effective output power of each industrial load is quantified through the calculation model, thereby achieving decoupling of multi-granularity energy consumption characteristic parameters.

[0091] The following describes in detail, with reference to preferred embodiments, the specific implementation methods of each functional module constituting the system provided by the above embodiments of the present invention.

[0092] The system provided in the above embodiments of the present invention mainly includes the following parts: First, an energy consumption parameter prediction module, which proposes a multi-granularity data-driven energy consumption prediction method, constructs a deep neural network including a convolutional neural network and a long short-term memory layer, and uses multi-dimensional data from the actual operating cycle as input, including industrial load status information, environmental conditions, and operating parameters. Second, an energy decoupling module, which defines online energy decoupling rules, processes the energy flow of different industrial loads layer by layer based on the energy decoupling rules, sets a total cost minimization constraint, and obtains a control sequence by allocating loads among different energy sources. Third, an energy management module, which establishes a global energy management function, defines control variables, including the current energy configuration, the power allocation ratio among different loads, the impact power, and the charging and discharging power of energy storage devices, and incorporates system energy consumption and degradation costs into the global energy management function. Fourth, an energy efficiency evaluation module, which establishes an energy efficiency calculation framework and quantifies the energy efficiency of each link (i.e., quantifies the effective output power of each industrial load).

[0093] In a preferred embodiment, the energy consumption parameter prediction module proposes a multi-granularity data-driven energy consumption prediction method, which constructs a deep neural network containing a convolutional neural network and a long short-term memory layer, and uses multi-dimensional data from the actual operating cycle as input, including industrial load status information, environmental conditions and operating parameters.

[0094] More preferably, the multi-granularity data-driven energy consumption prediction method in the energy consumption parameter prediction module is specifically implemented as follows:

[0095] (1) Collect high-frequency time series data, including industrial load status information, environmental conditions and operating parameters;

[0096] (2) Preprocess the collected data, including cleaning, normalization and feature extraction;

[0097] (3) Construct a deep neural network containing convolutional neural network layers and long short-term memory layers to identify the operating parameters of different industrial loads and predict future energy consumption.

[0098] More preferably, high-frequency time-series data is collected, including multi-dimensional data on industrial load status, environmental conditions, and operating parameters. Specifically, the multi-dimensional data includes:

[0099] Industrial load status information:

[0100] Electrolytic aluminum: Current intensity I Al (t), efficiency factor η Al Total power demand P Al (t).

[0101] Electric heating: Set temperature T set Ambient temperature T env (t), total power demand P Heat (t).

[0102] Energy storage system: charging and discharging current I bat (t), charging and discharging power P bat (t), State of charge (SOC) k .

[0103] Environmental conditions: temperature and humidity.

[0104] Operating parameters: Equipment operating mode.

[0105] In a preferred embodiment, the energy decoupling module defines online energy decoupling rules, processes the energy flow of different industrial loads layer by layer based on the energy decoupling rules, sets a total cost minimization constraint, and obtains a control sequence by allocating loads among different energy sources.

[0106] More preferably, the online energy decoupling algorithm in the energy decoupling module is specifically implemented as follows:

[0107] (1) Process the energy flow of different industrial loads layer by layer based on energy decoupling rules;

[0108] (2) The discrete-time dynamic programming optimization solver based on Bellman's principle guides the optimal operation of each industrial load and calculates the value function to minimize the total cost and maximize the system benefits.

[0109] More preferably, the value function calculation of the electrolytic aluminum industry load is performed by minimizing cost constraints, and the specific implementation methods include:

[0110] For electrolytic aluminum loads, minimizing power consumption is achieved by optimizing operating conditions. This is primarily done by adjusting the current intensity I. Al (t) and efficiency factor η Al To optimize the operating efficiency of the electrolyzer. The state variable is the total power demand P. Al (t). The value function can be expressed as:

[0111]

[0112] Among them, u k,Al =[η Al,k ] is a control variable, representing the efficiency factor of the electrolyzer in stage k.

[0113] x k,Al It is a state variable, representing the total power demand of the electrolytic aluminum load at time t. L Al It is an instantaneous cost function used to measure energy consumption at each stage. P elec,Al Indicates instantaneous power consumption, expressed by η Al Other relevant parameters (i.e., state variable x) k,Al The number of stages in the discrete input sequence is determined by N, which is the number of time steps for prediction.

[0114] More preferably, the value function calculation for the industrial load of electric heating is performed by minimizing cost constraints, and the specific implementation methods include:

[0115] For electric heating loads, minimizing power consumption is achieved by optimizing the ambient temperature difference. Therefore, the control variable is the set temperature T. set The state variable is the total power demand P. Heat (t). The value function can be expressed as:

[0116]

[0117] Among them, u k,Heat =[T set,k ] is a control variable, representing the set temperature in stage k.

[0118] x k,Heat It is a state variable, representing the total power demand of the electric heating load at time t. L Heat It is an instantaneous cost function used to measure energy consumption at each stage. P elec,Heat Indicates instantaneous power consumption, expressed by T. set The parameters determine this. N is the total number of stages in the discrete input sequence, i.e., the number of time steps for prediction.

[0119] More preferably, the value function calculation of the energy storage system load is performed by minimizing cost constraints, and the specific implementation methods include:

[0120] After minimizing demand, the battery must meet the energy requirements to achieve optimal energy allocation in terms of cost. Because the energy storage system has a fast response time, it can quickly handle immediate situations; therefore, the control variable is defined as the charging and discharging power P. bat The value function can be expressed as:

[0121]

[0122] Among them, cost ele This is the cost of electricity, expressed as the energy consumed per kilowatt-hour. bat This is the aging cost of the energy storage system, representing the increase in cost per charge-discharge cycle. ele and cost bat The value can be derived from actual operational data. N is the total number of stages in the discrete input sequence, i.e., the number of time steps for prediction. k,storage It is a state variable, representing the current state of the energy storage system at time t, the state of charge (SOC). k,storage =[P bat [k] is a control variable, representing the charging and discharging power in stage k.

[0123] More preferably, in order to ensure that the control strategy conforms to the actual operational characteristics while meeting production needs, the following constraints are enforced:

[0124] Electrolytic aluminum load constraints: Current intensity I Al It must be kept within a safe range, i.e., I Al,min ≤I Al,k ≤I Al,max This is to avoid overload or insufficient power supply.

[0125] Electric heating load constraints: set temperature T set It must be kept within a reasonable range, i.e., T set,min ≤T set,k ≤T set,max To ensure indoor comfort and energy efficiency. Ambient temperature T env Actual weather conditions must be taken into account. env,min ≤T env,k ≤T env,max This is to ensure the accuracy of the model.

[0126] Energy storage system load constraints: charging and discharging current I bat It must be kept within a safe range, i.e., I bat,min ≤I bat,k ≤Ibat,max To avoid damaging the energy storage device. Charging / discharging power P bat It must be kept within the limits allowed by the equipment, i.e., P bat,min ≤P bat,k ≤P bat,max To ensure the safe operation of the battery, the state of charge (SOC) must be maintained within a reasonable range, i.e., 0 ≤ SOC. k ≤SOC max This is to ensure the safety and lifespan of energy storage devices.

[0127] In a preferred embodiment, the energy management module establishes a global energy management function, defines control variables, and incorporates system energy consumption and cost into the optimization function.

[0128] More preferably, the global energy management function established in the energy management module is constructed in the following way:

[0129]

[0130] Where M is the total industrial load. i This is a weighting factor, representing the importance of each industrial load. J i It is a value function for each industrial load, including: J Al J Heat and J Storage .

[0131] In a preferred embodiment, the energy efficiency assessment module establishes an energy efficiency calculation and assessment framework to quantify the energy efficiency of each stage.

[0132] More preferably, the energy efficiency calculation and evaluation framework in the energy efficiency assessment module is specifically implemented in the following ways:

[0133] Total energy efficiency index η global Calculated using the following formula:

[0134]

[0135] Among them, P output,i (t) represents the effective output power of the i-th industrial load; η sts,i P represents the operating efficiency coefficient of the corresponding subsystem; input,j (t) represents the total power of the j-th energy input; M represents the total number of industrial loads; R represents the number of energy input types.

[0136] It should be noted that the steps in the method provided by the present invention can be implemented using the corresponding components in the system. Those skilled in the art can refer to the technical solution of the system to implement the steps of the method, and can also refer to the technical solution of the method to implement the composition of the system. That is, the embodiments in the system and the embodiments in the method can be understood as preferred examples of each other, which will not be elaborated here.

[0137] The technical solution provided by the above embodiments of the present invention will be further described in detail below with reference to a specific application example.

[0138] In this specific application example, a real-world scenario of a large industrial park is selected for verification. This park houses various types of industrial enterprises, including aluminum smelters, electric heating facilities, and energy storage systems. These enterprises have different electricity consumption patterns and load characteristics, making them highly suitable for applying a data-driven, multi-granularity energy consumption characteristic parameter decoupling algorithm to improve the energy efficiency of the power system.

[0139] The specific application example involves a data-driven decoupling method for multi-granularity energy consumption characteristic parameters, which includes the following steps:

[0140] Step S1: A multi-granularity data-driven energy consumption prediction method is proposed. A deep neural network containing a convolutional neural network and a long short-term memory layer is constructed. Multi-dimensional data from the actual operating cycle is used as input, including industrial load status information, environmental conditions and operating parameters.

[0141] Step S2: Define online energy decoupling rules, process the energy flow of different industrial loads layer by layer based on the energy decoupling rules, set the total cost minimization constraint, and obtain the control sequence by allocating the load among different energy sources.

[0142] Step S3: Establish a global energy management function, define control variables, incorporate system energy consumption and cost into the global energy management function, integrate the energy consumption and cost of each industrial load, and obtain the total energy input.

[0143] Step S4: Establish an energy efficiency calculation framework and quantify the energy efficiency of each stage by using control variables and total energy input.

[0144] The specific implementation process of each of the above steps is the same as that of each step or the implementation method of each functional module given in the preferred embodiment of the above method or system of the present invention, and will not be repeated here.

[0145] To evaluate the accuracy of data-driven learning in industrial load forecasting, a randomized test period was selected as the input. Mean squared error (MSE) and root mean square error (RMSE) were used as evaluation metrics. Figure 3The MSE and RMSE values ​​between predicted load power and actual load power are shown. The MSE based on the method of this invention is only 1.58, which is 64.9% lower than that of LSTM network (2.98), indicating that the DNN model has stronger adaptability and higher prediction accuracy.

[0146] like Figure 4 As shown, this invention extends the lifespan of energy storage devices and improves energy recovery rates through decoupling algorithm optimization. The invention enables energy storage devices to maintain low SOC changes during frequent charge-discharge cycles. The maximum percentage SOC change in this invention is 9.8%, only one-quarter of the experimental results. This demonstrates that a framework based on data-driven and online energy decoupling rules can effectively control the charging and discharging process of energy storage devices, maintaining a relatively stable SOC trend, thereby significantly reducing energy storage device degradation.

[0147] To optimize the electricity demand of industrial loads (electrolytic aluminum, electric heating), minimum electricity demand is calculated using online energy decoupling rules based on forecast information. By optimizing the operating point, the overall efficiency of the system is improved. Figure 5 The study demonstrates a comparison of power demand during the test period. Through energy decoupling and online optimization, power demand was reduced from 41.8 MWh to 39.4 MWh. This improvement is particularly important for energy storage devices because it not only reduces the total energy required for charging but also extends the lifespan of the energy storage system.

[0148] Figure 6 The paper demonstrates carbon emissions under different cycles. By applying the energy decoupling framework proposed in this invention, the carbon emissions for the optimal cycle are 5224 kg, a reduction of 29.6% compared to the average experimental results. This indicates that the optimized system significantly reduces carbon emissions in each cycle.

[0149] This invention can accurately predict the operating status of the entire system, realize real-time and effective energy allocation, and improve energy efficiency by 29.6%.

[0150] An embodiment of the present invention also provides a computer terminal, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it can be used to execute any of the methods in the above embodiments of the present invention, or to run any of the systems in the above embodiments of the present invention.

[0151] Optionally, the memory is used to store programs; the memory may include volatile memory, such as random-access memory (RAM), such as static random-access memory (SRAM), double data rate synchronous dynamic random-access memory (DDR SDRAM), etc.; the memory may also include non-volatile memory, such as flash memory. The memory is used to store computer programs (such as application programs, functional modules, etc. that implement the above methods), computer instructions, etc., and the aforementioned computer programs, computer instructions, etc., can be partitioned and stored in one or more memories. Furthermore, the aforementioned computer programs, computer instructions, data, etc., can be accessed by the processor.

[0152] A processor is used to execute computer programs stored in memory to implement the various steps of the methods or various modules of the systems involved in the above embodiments. For details, please refer to the relevant descriptions in the preceding method and system embodiments.

[0153] The processor and memory can be separate structures or integrated structures. When the processor and memory are separate structures, they can be coupled together via a bus.

[0154] An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, can be used to perform the method of any of the above embodiments of the present invention, or to run the system of any of the above embodiments of the present invention.

[0155] Computer-readable media include computer storage media and communication media, wherein communication media include any medium that facilitates the transfer of computer programs from one place to another. Storage media can be any available medium accessible to a general-purpose or special-purpose computer. An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and storage medium can reside in an ASIC. Alternatively, the ASIC can reside in a user device. Of course, the processor and storage medium can also exist as discrete components in a communication device.

[0156] The data-driven multi-granularity energy consumption characteristic parameter decoupling method and system provided in the above embodiments of the present invention solves the problem that traditional energy management systems, which rely on static rules, are unable to capture dynamic load characteristics and environmental factors by constructing a multi-granularity data-driven prediction model containing convolutional neural networks and long short-term memory layers. This achieves high-precision prediction of industrial load energy demand and significantly improves the prediction accuracy of system operating status. By defining energy decoupling rules and hierarchical optimization strategies based on dynamic programming, the method solves the problems of high computational complexity and difficulty in real-time response in complex energy flow processing of traditional methods. It realizes dynamic decoupling and optimal allocation of energy flows of different industrial loads, effectively reducing the total system cost and improving energy dispatch efficiency. By establishing a global energy management function and integrating multi-load value functions and weight coefficients, the method overcomes the limitations of traditional optimization models that ignore user differences and multi-objective collaboration. This achieves a comprehensive optimization effect that coordinates system energy consumption, cost, and equipment lifespan, significantly improving the rationality and overall efficiency of energy allocation.

[0157] Any matters not covered in the above embodiments of the present invention are well-known in the art.

[0158] The specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various modifications or variations within the scope of the claims, which do not affect the essence of the present invention.

Claims

1. A data-driven decoupling method for multi-granularity energy consumption characteristic parameters, characterized in that, include: A deep neural network-based energy consumption prediction model is provided, which takes multi-dimensional data during the actual operating cycle as input to the energy consumption prediction model and outputs energy flow under different industrial loads. Define energy decoupling rules, process the energy flow of different industrial loads layer by layer based on the energy decoupling rules, set total cost minimization constraints, calculate the value function of different industrial loads through the total cost minimization constraints, and obtain the control sequence by allocating loads among different energy sources; Based on the value functions of different industrial loads, a global energy management function is established, and the control variables of the global energy management function are defined. The energy consumption and cost of each industrial load are integrated to obtain the total energy input. An energy efficiency calculation model is established. Using the control variables and total energy input, the effective output power of each industrial load is quantified through the calculation model, thereby achieving decoupling of multi-granularity energy consumption characteristic parameters.

2. The data-driven multi-granularity energy consumption characteristic parameter decoupling method according to claim 1, characterized in that, The provided energy consumption prediction model based on a deep neural network takes multi-dimensional data from the actual operating cycle as input and outputs energy flows for different industrial loads, including: Acquire multi-dimensional time series data within the actual operating cycle, including: industrial load status information, environmental conditions, and operating parameters; The multi-dimensional time series data is preprocessed; A deep neural network model containing convolutional neural network layers and long short-term memory layers is provided. Preprocessed data is input into the neural network model to identify the operating parameters of different industrial loads, predict future energy consumption, and output the energy flow of different industrial loads.

3. The data-driven multi-granularity energy consumption characteristic parameter decoupling method according to claim 2, characterized in that, The status information of the industrial load includes: Electrolytic aluminum, including: current intensity I Al Efficiency factor η Al and total power demand P Al ; Electric heating, including: set temperature T set Ambient temperature T env and total power demand P Heat ; Energy storage system, including: charging and discharging current I bat Charging and discharging power P bat and State of charge (SOC) k ; The environmental conditions include: temperature and humidity; The operating parameters include: device operating mode.

4. The data-driven multi-granularity energy consumption characteristic parameter decoupling method according to claim 1, characterized in that, The defined energy decoupling rule processes the energy flow of different industrial loads layer by layer, sets a total cost minimization constraint, calculates the value function of different industrial loads through the total cost minimization constraint, and obtains a control sequence by allocating loads among different energy sources, including: Define online energy decoupling rules, including: Based on the Bellman optimality principle, an independent value function and constraints are defined for each industrial load. The optimal control variables for each industrial load are solved in stages through discrete-time dynamic programming, and finally a dynamic control command sequence that satisfies the minimization of global total cost is generated. The command sequence includes a set of industrial load operating parameters that are dynamically adjusted according to the time step. Based on the aforementioned energy decoupling rules, the energy flow of different industrial loads is processed layer by layer, including: The system is decomposed into multiple independent load layers, and the optimal control variables for each layer are calculated independently based on dynamic programming using energy decoupling rules. Based on the processed energy flow, the discrete-time dynamic programming optimization solver according to Bellman's principle guides the optimal operation of each industrial load, defines the minimum cost constraint, and calculates the value function for different industrial loads by minimizing the cost constraint. Based on the value function calculation results, the industrial load is allocated among different energy sources to obtain the control sequence.

5. The data-driven multi-granularity energy characteristic parameter decoupling method according to claim 4, characterized in that, The calculation of the value function for different industrial loads by minimizing cost constraints also includes any one or more of the following: - Value function calculation of the electrolytic aluminum industry load by minimizing cost constraints, including: In the formula, J Al The value function representing the load of the electrolytic aluminum industry; u k,Al =[η Al,k [x] is a control variable, representing the efficiency factor of the electrolyzer in stage k; k,Al L is a state variable representing the total power demand of the electrolytic aluminum load at time t; Al P is the instantaneous cost function used to measure energy consumption at each stage; elec,Al This represents instantaneous power consumption; N is the total number of stages in the discrete input sequence, i.e., the number of time steps for prediction. The load constraints for electrolytic aluminum are established as follows: I Al,min ≤I Al,k ≤I Al,max In the formula, I Al,k I represents the current intensity in the k-th stage; Al,min Indicates the lower limit of the safe range for current intensity; I Al,max Indicates the upper limit of the safe range for current intensity; - Calculate the value function for industrial loads of electric heating by minimizing cost constraints, including: In the formula, J Heat The value function representing the industrial load of electric heating; u k,Heat =[T set,k [x] is the control variable, representing the set temperature in stage k; k,Heat L is a state variable representing the total power demand of the electric heating load at time t; Heat P is the instantaneous cost function used to measure energy consumption at each stage; elec,Heat Indicates instantaneous power consumption, determined by the set temperature T. set The decision is made by N, which represents the total number of stages in the discrete input sequence, i.e., the number of time steps for prediction. The constraints for establishing the electric heating load are as follows: T set,min ≤T set,k ≤T set,max T env,min ≤T env,k ≤T env,max In the formula, T set,k T represents the set temperature for stage k; set,min This indicates the lower limit of the reasonable temperature range; T set,max This indicates the upper limit of the reasonable temperature range; T env Indicates ambient temperature; T env,min Indicates the lower limit of ambient temperature; T env,max Indicates the upper limit of ambient temperature; - Calculate the value function of the energy storage system load by minimizing cost constraints, including: In the formula, J Storage Cost represents the value function of the energy storage system load. ele Cost represents the energy consumed per kilowatt-hour. bat The aging cost of an energy storage system represents the cost increase per charge-discharge cycle. ele and cost bat The value is derived from actual operational data; N is the total number of stages in the discrete input sequence, i.e., the number of time steps for prediction; x k,storage The state variable represents the current state of the energy storage system at time t, i.e., the state of charge (SOC); u k,storage =[P bat [k] is a control variable, representing the charging and discharging power in the k-th stage; The constraints for establishing the load of the energy storage system are: I bat,min ≤I bat,k ≤I bat,max P bat,min ≤P bat,k ≤P bat,max 0≤SOC k ≤SOC max In the formula, I bat,k I represents the charging and discharging current in the k-th stage; bat,min Indicates the lower limit of the safe range for charging and discharging current; I bat,max Indicates the upper limit of the safe range of charging and discharging current; P bat,k P represents the charging and discharging power in the k-th stage; bat,min Indicates the lower limit of the device's permissible range; P bat,max Indicates the upper limit of the device's allowed range; SOC k State of charge (SOC) represents the state of charge at stage k. max This indicates the upper limit of the state of charge.

6. The data-driven multi-granularity energy consumption characteristic parameter decoupling method according to claim 1, characterized in that, Based on the value functions of different industrial loads, a global energy management function is established, and the control variables of the global energy management function are defined. The energy consumption and cost of each industrial load are integrated to obtain the total energy input, including: The function for establishing global energy management is expressed as follows: In the formula, J glabool This represents the global energy management function; M is the total industrial load; w i J represents the weighting coefficient, indicating the importance of each industrial load; i The value function for each industrial load; The control variables of the global energy management function are defined, including the state information of the industrial load, environmental conditions, and operating parameters, so that the energy consumption and cost of each industrial load are reflected in the value function of the global energy management function; based on the weighting coefficients, the energy consumption and cost of each industrial load are integrated to obtain the total energy input.

7. The data-driven multi-granularity energy consumption characteristic parameter decoupling method according to claim 1, characterized in that, The establishment of the energy efficiency calculation model, utilizing the control variables and total energy input, quantifies the effective output power of each industrial load through the calculation model, including: The energy efficiency calculation model is expressed as follows: Total energy efficiency index η global Calculated using the following formula: In the formula, η global P represents the overall energy efficiency index; output,i (t) represents the effective output power of the i-th industrial load; η sys,i P represents the operating efficiency coefficient of the corresponding subsystem; input,j (t) represents the total power of the j-th energy input; M represents the total number of industrial loads; R represents the number of energy input types.

8. A data-driven multi-granularity energy consumption characteristic parameter decoupling system, characterized in that, include: The energy consumption parameter prediction module provides an energy consumption prediction model based on a deep neural network. It takes multi-dimensional data from the actual operating cycle as input to the energy consumption prediction model and outputs the energy flow of different industrial loads. An energy decoupling module is used to define an energy decoupling rule, process the energy flow of different industrial loads layer by layer based on the energy decoupling rule, set a total cost minimization constraint, calculate the value function of different industrial loads through the total cost minimization constraint, and obtain a control sequence by allocating loads among different energy sources. The energy management module establishes a global energy management function based on the value function of different industrial loads, defines the control variables of the global energy management function, integrates the energy consumption and cost of each industrial load, and obtains the total energy input. The energy efficiency assessment module is used to establish an energy efficiency calculation model. Using the control variables and total energy input, the calculation model quantifies the effective output power of each industrial load, thereby achieving decoupling of multi-granularity energy consumption characteristic parameters.

9. A computer terminal, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it can be used to perform the method of any one of claims 1-7, or to run the system of claim 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program can be used to perform the method of any one of claims 1-7, or to run the system of claim 8.

Images