Multi-granularity energy consumption characteristic parameter decoupling method and system based on constraint uncertainty set

Abstract

The invention provides a multi-granularity energy consumption characteristic parameter decoupling method and system based on a constraint uncertainty set, and the method comprises the steps: dividing the energy consumption characteristic of an industrial load into a plurality of intervals according to the energy consumption distribution and condition category probability, constructing a corresponding convex hull, and forming an uncertainty set based on a multi-interval convex hull; constructing a worst expectation four-layer robust optimization model by using the uncertainty set based on the energy consumption uncertainty and the condition category of the industrial load on different time scales; a Bregman alternating direction multiplier method is adopted to solve the non-convex problem of the worst expectation four-layer robust optimization model; and on the basis of the solution of the non-convex problem, performing continuous updating and decoupling optimization on the worst expectation four-layer robust optimization model by adopting a column and constraint generation algorithm of an alternating iteration strategy so as to realize decoupling of the multi-granularity energy consumption characteristics of the industrial load. According to the invention, the flexibility and adaptability of energy management are enhanced, and the energy utilization efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy management and optimization, in particular to a multi-granularity energy consumption characteristic parameter decoupling method and system based on constraint uncertain set, and also relates to a corresponding computer terminal and computer readable storage medium. BACKGROUND

[0002] With the rapid development of industrial production, the energy consumption demand of various industrial loads is becoming increasingly complex. These industrial loads not only have different energy consumption modes, but are also influenced by various external factors such as weather conditions, seasonal changes, market fluctuations, etc. Traditional methods have limitations in dealing with these uncertainties and complexities, and it is difficult to provide robust and efficient optimization solutions. Existing optimization models usually assume that energy consumption characteristics are deterministic or only consider uncertainty at a single time scale, which leads to their poor performance in the face of variability and diversity in actual industrial environments. Specifically, existing methods are often too conservative, sacrificing resource utilization efficiency in order to ensure the stability of the model; at the same time, these models have poor adaptability and cannot flexibly respond to changes in energy consumption demand at different times and scenarios, limiting their scope of application. In addition, for non-convex problem solving, existing methods also lack effective means to ensure the convergence of global optimal solutions. Therefore, there is an urgent need for an optimization method that can comprehensively consider energy consumption uncertainty and its category probability, and has efficient solving ability, to meet the needs of modern industrial load management. SUMMARY

[0003] The present application provides a multi-granularity energy consumption characteristic parameter decoupling method and system based on constraint uncertain set, and also provides a corresponding computer terminal and computer readable storage medium, to solve the above problems in the prior art.

[0004] According to one aspect of the present application, a multi-granularity energy consumption characteristic parameter decoupling method based on constraint uncertain set is provided, comprising:

[0005] Divide the energy consumption characteristics of industrial loads into multiple intervals according to energy distribution and condition category probability, and construct corresponding convex hulls to form a multi-interval convex hull-based uncertain set;

[0006] Using the uncertain set, construct a worst-case expected four-layer robust optimization model based on energy consumption uncertainty and condition category of industrial loads at different time scales;

[0007] Solve the non-convex problem of the worst-case expected four-layer robust optimization model using the Bregman alternating direction multiplier method;

[0008] Based on the solution of the non-convex problem, a column and constraint generation algorithm with an alternating iteration strategy is used to continuously update and decouple optimize the worst-case expected four-layer robust optimization model, so as to realize decoupling of the multi-granularity energy consumption characteristics of the industrial load.

[0009] According to another aspect of the present application, a multi-granularity energy consumption characteristic parameter decoupling system based on a constraint uncertainty set is provided, comprising:

[0010] An uncertainty set construction module is configured to divide energy consumption characteristics of an industrial load into multiple intervals according to energy consumption distribution and condition category probability, and construct a corresponding convex hull to form a multi-interval convex hull-based uncertainty set.

[0011] A robust optimization modeling module is configured to use the uncertainty set to construct a worst-case expected four-layer robust optimization model based on energy consumption uncertainty and condition categories of the industrial load at different time scales.

[0012] A non-convex problem solving module is configured to use a Bregman alternating direction multiplier method to solve a non-convex problem of the worst-case expected four-layer robust optimization model.

[0013] A dynamic update decoupling optimization module is configured to use a column and constraint generation algorithm with an alternating iteration strategy to continuously update and decouple optimize the worst-case expected four-layer robust optimization model based on the solution of the non-convex problem, so as to realize decoupling of the multi-granularity energy consumption characteristics of the industrial load.

[0014] According to a third aspect of the present application, a computer terminal is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method described above or run the system described above.

[0015] According to a fourth aspect of the present application, a computer readable storage medium is provided, which stores a computer program executable by a processor to implement the method described above or run the system described above.

[0016] Thanks to the above technical solutions, the present application has at least one of the following beneficial effects compared with the prior art:

[0017] The multi-granularity energy consumption characteristic parameter decoupling method and system based on a constraint uncertainty set provided by the present application enhances the flexibility and adaptability of energy management and improves energy utilization efficiency.

[0018] The application provides a multi-granularity energy consumption characteristic parameter decoupling method and system based on a constrained uncertain set.

[0019] The application provides a multi-granularity energy consumption characteristic parameter decoupling method and system based on a constrained uncertain set, and a multi-interval convex hull uncertain set method is provided, energy consumption characteristics of an industrial load are divided into multiple intervals according to distribution and probability, and a corresponding convex hull is constructed to form a set, so that the uncertainty of the energy consumption characteristics can be accurately described and the conservativeness of the model can be reduced.

[0020] The application provides a multi-granularity energy consumption characteristic parameter decoupling method and system based on a constrained uncertain set, and a Bregman alternating direction multiplier method is introduced to solve a non-convex problem, so that the effectiveness and efficiency of the optimization process are ensured.

[0021] The application provides a multi-granularity energy consumption characteristic parameter decoupling method and system based on a constrained uncertain set, and a column and constraint generation algorithm based on an alternating iteration strategy is designed to constantly update and decouple optimize the model, so that dynamic updating and decoupling optimization of the model are realized. BRIEF DESCRIPTION OF DRAWINGS

[0022] Other features, objects and advantages of the application will become more apparent after reading the detailed description of non-limiting embodiments with reference to the following drawings:

[0023] Figure 1 The figure is a work flow diagram of the multi-granularity energy consumption characteristic parameter decoupling method based on a constrained uncertain set in a preferred embodiment of the application.

[0024] Figure 2 The figure is a component module structure diagram of the multi-granularity energy consumption characteristic parameter decoupling system based on a constrained uncertain set in a preferred embodiment of the application.

[0025] Figure 3 The figure is a load curve in a specific application example.

[0026] Figure 4 The figure is an optimization operation result of an industrial load system under sunny conditions in a specific application example.

[0027] Figure 5 The figure is an optimization operation result of an industrial load system under overcast conditions in a specific application example.

[0028] Figure 6 The figure is an optimization operation result of an industrial load system under rainy conditions in a specific application example. DETAILED DESCRIPTION

[0029] The embodiments of the present application are described in detail below: The embodiments are implemented on the premise of the technical solutions of the present application, and detailed implementation manners and specific operation processes are given. It should be pointed out that, for those skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the protection scope of the present application.

[0030] The prior art has limitations in dealing with external factor uncertainty and complexity, and it is difficult to provide a robust and efficient optimization solution. In addition, the existing model usually assumes that the energy consumption feature is deterministic or only considers uncertainty at a single time scale, which leads to poor performance when facing the variability and diversity in actual industrial environments.

[0031] To solve the above problems, an embodiment of the present application provides a multi-granularity energy consumption feature parameter decoupling method and system based on constraint uncertainty set. The method constructs a worst-case expected four-layer robust optimization model, uses a multi-level optimization framework to ensure finding the optimal solution under various uncertain environments, introduces a multi-interval convex hull uncertainty set method to accurately describe the uncertainty of energy consumption features and reduce the conservatism of the model, uses the Bregman alternating direction multiplier method to solve non-convex problems to ensure the effectiveness and efficiency of the optimization process, and designs a column and constraint generation algorithm based on an alternating iteration strategy to realize dynamic updating and decoupling optimization of the model. Through a series of innovations, the flexibility and robustness of industrial load energy consumption management are improved.

[0032] Specifically, as shown in the figure, the multi-granularity energy consumption feature parameter decoupling method based on constraint uncertainty set provided by the embodiment can include: Figure 1

[0033] S1, dividing the energy consumption features of the industrial load into multiple intervals according to the energy consumption distribution and condition category probability, and constructing corresponding convex hulls to form a multi-interval convex hull-based uncertainty set;

[0034] S2, using the uncertainty set, constructing a worst-case expected four-layer robust optimization model based on the energy consumption uncertainty of the industrial load at different time scales and the condition category;

[0035] S3, using the Bregman alternating direction multiplier method to solve the non-convex problem of the worst-case expected four-layer robust optimization model;

[0036] S4, based on the solution of the non-convex problem, using a column and constraint generation algorithm based on an alternating iteration strategy to continuously update and decouple optimize the worst-case expected four-layer robust optimization model, for realizing decoupling of the multi-granularity energy consumption features of the industrial load. ​

[0037] In some preferred embodiments, S1, dividing the energy consumption characteristics of the industrial load into multiple intervals according to the energy distribution and the condition category probability, and constructing the corresponding convex hull, forms an uncertain set based on the multi-interval convex hull, can further include:

[0038] S11, obtaining actual energy consumption records of different industrial load types under various condition categories to obtain the energy consumption characteristics of the industrial load; the energy consumption characteristics include the actual power consumption and the gas consumption of the industrial load under different scenarios;

[0039] S12, dividing the energy consumption characteristics into multiple intervals according to the energy distribution and the condition category probability, constructing the corresponding convex hull, and forming an uncertain set U based on the multi-interval convex hull conv , which is expressed as:

[0040]

[0041] In the formula, represents the actual power consumption of the i-th industrial load at time t and category s; represents the gas consumption of the i-th industrial load at time t and category s; s represents the condition category, including: weather conditions, seasonal changes, market conditions, equipment states, and production plans; N s represents the number of condition categories.

[0042] In some preferred embodiments, S2, constructing a worst-case expected four-layer robust optimization model, can further include:

[0043] S21, establishing the architecture of the worst-case expected four-layer robust optimization model, including:

[0044] The first layer is a device operation state decision layer, which is used to optimize the start-stop logic and energy input-output strategy of the industrial load, and generates binary variables to control the operation state of the device;

[0045] The second layer is a worst probability distribution layer, which is used to search for the uncertainty probability distribution that deteriorates the objective function within the set condition category probability space;

[0046] The third layer is a multi-interval convex hull constraint layer, which divides the intervals based on the distribution of the energy consumption characteristics of the industrial load, and uses the convex hull set to represent the feasible region of the uncertain parameters;

[0047] The fourth layer is a real-time scheduling optimization layer, which is used to dynamically adjust the energy storage state and energy interaction parameters to ensure that the power limit, the change rate constraint, and the load demand balance are met;

[0048] S22, in the second stage, establishing the objective function of the worst-case expected four-layer robust optimization model based on the preset condition category probability distribution σs The system total cost (power consumption, start-stop loss, etc.) in the condition category probability space (Ω σ ) is maximized, and the worst case of the energy use feature combination is taken as the worst case to find the optimal solution of the objective function.

[0049] In some preferred embodiments, the above S22, the objective function, can further include:

[0050] The objective function can be represented as:

[0051]

[0052] In the formula, y i,t ,z i,t u i,t respectively represent the binary variables of the start, stop and running state of the i-th industrial load at time t; B buy,i,t ,B sell,i,t represent the binary variables of whether the i-th industrial load inputs power from the grid or outputs power to the grid; i represents the specific type of industrial load; represents the start / stop cost coefficient of the i-th industrial load; T represents the time interval index; N represents the number of industrial load types; N s represents the number of condition categories; σ s represents the probability of the condition category; Ω σ represents the condition category probability space; U conv represents the uncertainty set; α t ,β t represent the grid interaction loss coefficient and energy feedback efficiency coefficient; represents the actual power consumption of the i-th industrial load at time t and in category s; represents the gas consumption of the i-th industrial load at time t and in category s; represents the actual power input from the grid to the i-th industrial load at time t and in scenario s; represents the actual power input to the grid by the i-th industrial load at time t and in scenario s; SOC i,t,s represents the energy storage state of the i-th industrial load at time t;

[0053] The constraints of the objective function can include: device start-stop logic constraints, power output limit constraints, power change rate limit constraints, thermal energy storage constraints and power balance constraints; wherein:

[0054] The device start-stop logic constraint is represented as:

[0055] y i,t -z i,t =u i,t-u i,t-1

[0056] y i,t +z i,t ≤1

[0057] The power output limit constraint is represented as:

[0058]

[0059] wherein, represents the minimum power output of the i-th industrial load; represents the maximum power output of the i-th industrial load; P i,t,s represents the actual power output of the i-th industrial load at time t, category s;

[0060] The power change rate limit constraint is represented as:

[0061]

[0062] wherein, rubp i represents the rise rate limit of the i-th industrial load; rdbn i represents the fall rate limit of the i-th industrial load; represents the maximum power output of the i-th industrial load;

[0063] The thermal storage energy constraint is represented as:

[0064]

[0065] wherein, represents the storage energy state of the i-th industrial load at time t, category s; represents the charging amount of the i-th industrial load at time t, category i; represents the discharging amount of the i-th industrial load at time t, category s; represents the charging efficiency of the i-th industrial load; represents the discharging efficiency of the i-th industrial load;

[0066] The power balance constraint is represented as:

[0067]

[0068] wherein, represents the demand of the i-th industrial load at time t.

[0069] In some preferred embodiments, the specific types of the industrial loads in the above S22 can further include: electrolytic aluminum, ferroalloy, polysilicon, electric heating, and energy storage system.

[0070] In some preferred embodiments, S22, the probability distribution σ of the preset condition category is obtained by the following method: s

[0071] Based on the historical data of the energy consumption characteristics of the industrial load, it is obtained by multi-scene analysis (weather, season, market conditions, etc.) statistics and defined in the category probability space. Specifically, the probability distribution is generated by combining Monte Carlo simulation or non-parametric estimation method through the multi-time scale energy fluctuation characteristics (such as power consumption, gas input, and energy storage state change) in the actual operation record of the industrial load. The probability distribution reflects the influence weight of different external condition categories (such as extreme weather, market fluctuation) on the energy consumption parameters.

[0072] In some preferred embodiments, S3, the Bregman alternating direction multiplier method is used to solve the non-convex problem of the worst expected four-layer robust optimization model, and the method can further include:

[0073] S31, setting an initial point x 0 and a Lagrange multiplier u 0 ;

[0074] S32, updating the decision variable x, which is expressed as:

[0075]

[0076] In the formula, x k+1 represents the updated decision variable in the k+1 iteration; argmin represents the value of x that minimizes the expression; f(x) represents the objective function of the worst expected four-layer robust optimization model; A represents the coefficient matrix in the linear constraint; is used to map the decision variable x to a linear space; s represents the linear constraint coefficient matrix related to the auxiliary variable y; y k represents the value of the auxiliary variable in the k iteration; c is a constant vector representing the constant value on the right side of the linear constraint; u k represents the Lagrange multiplier or dual variable in the k iteration, which is used to ensure that the constraints of the worst expected four-layer robust optimization model are satisfied; ρ represents the penalty parameter, which is used to adjust the penalty degree when the constraints are violated;

[0077] S33, updating the Lagrange multiplier, which is expressed as:

[0078] u k+1 = u k + ρ(Ax k+1 + By k+1 -c)

[0079] In the formula, u k+1 represents the updated Lagrange multiplier in the k+1 iteration.​

[0080] In some preferred embodiments, the S4, based on the solution of the non-convex problem, uses the column and constraint generation algorithm with alternating iteration strategy to constantly update and decouple the optimization of the worst-case expected four-layer robust optimization model, and can further include:

[0081] S41, in each iteration of solving the non-convex problem of the worst-case expected four-layer robust optimization model, dynamically generate new decision variable columns and model constraints according to the current solution, denoted as:

[0082] Ax k+1 +By k+1 -c+u k+1 =0

[0083] In the formula, x k+1 is a new decision variable, y k+1 is a new auxiliary variable, and u k+1 is a new Lagrange multiplier;

[0084] S42, gradually add the new model constraints to the worst-case expected four-layer robust optimization model, constantly update the decoupled optimization model, until the model convergence condition is reached.

[0085] Based on the same inventive concept, an embodiment of the present application also provides a multi-granularity energy consumption characteristic parameter decoupling system based on constraint uncertainty set.

[0086] Specifically, as shown in Figure 2 , the multi-granularity energy consumption characteristic parameter decoupling system based on constraint uncertainty set provided by this embodiment can include:

[0087] An uncertainty set construction module, which is configured to divide the energy consumption characteristics of the industrial load into multiple intervals according to energy consumption distribution and condition category probability, and construct a corresponding convex hull to form a multi-interval convex hull-based uncertainty set;

[0088] A robust optimization modeling module, which is configured to use the uncertainty set to construct a worst-case expected four-layer robust optimization model based on energy consumption uncertainty and condition category of the industrial load at different time scales;

[0089] A non-convex problem solving module, which is configured to use the Bregman alternating direction multiplier method to solve the non-convex problem of the worst-case expected four-layer robust optimization model;

[0090] A dynamic update decoupling optimization module, which is configured to use the column and constraint generation algorithm with alternating iteration strategy to constantly update and decouple the optimization of the worst-case expected four-layer robust optimization model based on the solution of the non-convex problem, and is configured to realize the decoupling of the multi-granularity energy consumption characteristics of the industrial load.

[0091] The working content implementation of each function module of the system provided in the above embodiments of the application will be described in further detail below in combination with preferred embodiments.

[0092] The system provided in the above embodiments of the application includes the following parts: a first part, a robust optimization modeling module, for constructing a worst-case expected four-layer robust optimization model, considering the energy consumption uncertainty and its category probability of industrial loads such as electrolytic aluminum, ferroalloy, polysilicon, electric heating and energy storage system at different time scales; a second part, an uncertainty set construction module, for proposing a multi-interval convex hull uncertainty set method, by dividing the energy consumption characteristics of industrial loads into multiple intervals according to the distribution and probability, and constructing the corresponding convex hull to form the set; a third part, a non-convex problem solving module, for introducing a Bregman alternating direction multiplier method to solve the non-convex problem; and a fourth part, a dynamic updating decoupling optimization module, for designing a column and constraint generation algorithm based on an alternating iteration strategy to update the model continuously.

[0093] Further preferably, the robust optimization modeling module constructs a worst-case expected four-layer robust optimization model, and the specific implementation manner thereof includes:

[0094] A first stage: determining the start-stop plan (y i,t ), the power purchase and sale plan (B buy,i,t , B sell,i,t ), wherein i represents a specific industrial load type (electrolytic aluminum, ferroalloy, polysilicon, electric heating, energy storage system), establishing the basic architecture of the worst-case expected four-layer robust optimization model, including:

[0095] A first layer is a device operation state decision layer, for optimizing the start-stop logic and energy input-output strategy of industrial loads, to generate binary variables to control the operation state of the device;

[0096] A second layer is a worst probability distribution layer, for searching the uncertainty probability distribution that deteriorates the objective function (i.e., maximizes the total system cost) within the set condition category probability space;

[0097] A third layer is a multi-interval convex hull constraint layer, dividing intervals based on the distribution of industrial load energy consumption characteristics, and using the convex hull set to represent the feasible region of the uncertainty parameter;

[0098] A fourth layer is a real-time scheduling optimization layer, for dynamically adjusting the energy storage state and energy interaction parameters to ensure that the power limit, change rate constraint and load demand balance are met.

[0099] A second stage: based on the probability distribution of each category (σ s , finding the optimal solution of the objective function in the worst case.

[0100] Further, each category specifically includes: weather conditions, seasonal changes, market conditions, equipment status, production plans.

[0101] Further, the objective function is:

[0102]

[0103] In the above objective function, is the first layer; is the second layer; is the third layer; and is the fourth layer.

[0104] where y i,t ,z i,t u i,t : respectively represent the binary variables of the start, shutdown and running state of the i-th industrial load at time t. B buy,i,t ,B sell,i,t indicates the binary variable of whether the i-th industrial load inputs power from the grid or outputs power to the grid. Start-up / shutdown cost coefficient of the i-th industrial load. T: time interval index. N: number of industrial load types. N s : number of categories. σ s : probability of category. Ω σ : space of category probability. U conv : uncertainty set. α t ,β t : grid interaction loss coefficient, energy feedback efficiency coefficient. represents the actual power consumption of the i-th industrial load at time t under category s. Here s refers to different scenario or condition categories, such as different weather conditions, seasonal changes, etc. represents the gas consumption (such as hydrogen, nitrogen, etc.) of the i-th industrial load at time t under category s, which is suitable for industrial processes that require gas input (polysilicon production).

[0105] Further, the constraints of the objective function include: equipment start-stop logic constraints, power output limit constraints, power change rate limit constraints, thermal energy storage constraints, power balance constraints.

[0106] Further, the equipment start-stop logic constraint is:

[0107] y i,t -z i,t =u i,t -u i,t-1

[0108] y i,t +z i,t ≤1

[0109] Further, the power output limit constraint is:

[0110]

[0111] where, the minimum power output of the ith industrial load. the maximum power output of the ith industrial load. i,t,s : the actual power output of the ith industrial load at time t, category s.

[0112] Further, the power change rate limit constraint is:

[0113]

[0114] where, i : the rise rate limit of the ith industrial load. i : the fall rate limit of the ith industrial load. the maximum power output of the ith industrial load.

[0115] Further, the energy storage constraint is:

[0116]

[0117] where, the energy storage state of the ith industrial load at time t, category s. the charging amount of the ith industrial load at time t, category s. the discharging amount of the ith industrial load at time t, category s. the charging efficiency of the ith industrial load. the discharging efficiency of the ith industrial load.

[0118] Further, the power balance constraint is:

[0119]

[0120] where, denotes the demand of the ith industrial load at time t.

[0121] Further preferably, the uncertainty set construction module is based on a multi-interval convex hull uncertainty set method, the specific implementation of which includes:

[0122] The multi-interval convex hull uncertainty set is formed by constructing the corresponding convex hulls for the actual energy consumption records of different industrial load types (aluminum electrolysis, ferroalloy, polysilicon, electric heating, and energy storage systems) under various conditions.

[0123]

[0124] Further preferably, the non-convex problem solving module introduces the Bregman alternating direction multiplier method to solve the non-convex problem, and the specific implementation manner comprises:

[0125] (1) Set the initial point x 0 and the Lagrange multiplier λ 0 .

[0126] (2) Update the decision variable x:

[0127]

[0128] In the formula, x k+1 : the updated decision variable in the k+1th iteration. argminx: indicates finding the value of x that minimizes the expression. f(x): objective function. A: indicates the coefficient matrix in the linear constraint. It maps the decision variable x to a linear space. B: indicates the linear constraint coefficient matrix related to the auxiliary variable y k : the value of the auxiliary variable in the kth iteration. c: constant vector, indicating the constant value on the right side of the linear constraint. u k : the Lagrange multiplier (or dual variable) in the kth iteration, used to ensure that the constraint condition is satisfied. ρ: penalty parameter, used to adjust the penalty degree when the constraint is violated.

[0129] (3) Update the multiplier u

[0130] u k+1 = u k + ρ (Ax k+1 + By k+1 - c)

[0131] In the formula, u k+1 : the updated Lagrange multiplier in the k+1th iteration. u k : the Lagrange multiplier in the kth iteration. y: auxiliary variable.

[0132] Further preferably, the dynamic updating decoupling optimization module is designed based on the column and constraint generation algorithm based on the alternating iteration strategy to realize continuous updating and decoupling optimization of the model, and the specific implementation manner comprises:

[0133] (1) Gradually generate new columns and constraints: in each iteration, new columns (decision variables) and constraint conditions are dynamically generated according to the current solution, so that the model is closer to the actual energy use situation.

[0134] (2) Continuously update the optimization model: by gradually adding new constraint conditions, the optimization decoupling model is continuously updated until the convergence condition is reached. Column and constraint generation formula:

[0135] Axk+1 +By k+1 -c+u k+1 =0

[0136] It should be noted that the steps in the method provided by the application can be realized by using the corresponding components in the system, and those skilled in the art can refer to the technical solution of the system to realize the step flow of the method, or refer to the technical solution of the method to realize the composition of the system, that is, the embodiments in the system and the embodiments in the method can be understood as preferred examples, which will not be repeated here.

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

[0138] In this specific application example, the simulation model is composed of a typical industrial load system, which includes an aluminum electrolysis plant, a ferroalloy plant, a polysilicon plant, an electric heating facility, and an energy storage system. The load curve is as shown in the figure. Figure 3 The simulation running environment is a computer with a 2.8GHz processor, 6 Intel cores and 16GB memory, which uses GAMS / GUROBI solver for solving, and the time interval T is 1 hour.

[0139] A real application scenario of a certain large industrial park is selected for verification. The park has various types of industrial enterprises, including an aluminum electrolysis plant, a ferroalloy plant, a polysilicon plant, an electric heating facility, and an energy storage system. These enterprises have different power consumption modes and load characteristics, which are very suitable for applying the multi-granularity energy consumption characteristic parameter decoupling method based on constraint uncertainty set to improve the energy efficiency of the power system.

[0140] The multi-granularity energy consumption characteristic parameter decoupling method based on constraint uncertainty set involved in this specific application example includes the following steps:

[0141] Step S1: A multi-interval convex hull uncertainty set method is proposed, which divides the energy consumption characteristics of industrial loads into multiple intervals according to the distribution and probability, and constructs the corresponding convex hull to form the set.

[0142] Step S2: A worst-case expected four-layer robust optimization model is constructed, which considers the energy consumption uncertainty and its category probability of aluminum electrolysis, ferroalloy, polysilicon, electric heating and energy storage system industrial loads at different time scales.

[0143] Step S3: The Bregman alternating direction multiplier method is introduced to solve the non-convex problem.

[0144] Step S4: A column and constraint generation algorithm based on an alternating iteration strategy is designed to continuously update and decouple optimize the model.

[0145] The implementation process of each step and the implementation mode of each functional module are the same as those of the preferred embodiment of the method or system of the application, and will not be repeated here.

[0146] Figure 4 、 Figure 5 、 Figure 6 The optimization results of the industrial load system under different weather conditions (sunny, cloudy, rainy) are shown. In the time period t1-t4, the energy generation capacity within the system exceeds the power demand, so there is excess energy that can be sold to the external market. Conversely, during the peak electricity consumption period (such as t10-t12), even if the maximum power generating equipment is used, the system still needs to purchase additional energy from the external market to meet the demand. For thermal energy management: low load period (such as t1-t8): when the thermal energy supply is insufficient, the system uses efficient waste heat recovery technology to provide the required thermal energy, ensuring efficient use of energy. By absorbing low-temperature waste heat in the production process through the heat recovery unit, dependence on external heat sources is reduced. High load period (such as t10-t12): during peak demand periods, in addition to maximizing the use of existing power generation equipment, the system also supplements the thermal energy demand by flexibly using energy storage facilities to ensure stable supply. The thermal storage facility is flexibly adjusted according to weather conditions and load demand: in certain cases (such as t10-t16), when there is excess energy, the thermal storage facility will be charged for subsequent use. In other cases (such as during t18-t20 of the rainy category), the thermal storage is only discharged (such as at t3) to supplement the thermal energy demand when needed.

[0147] As verified by the above specific application examples, the application can flexibly adjust the energy production and storage scheme according to the changes in different weather conditions and load demand, ensuring efficient operation and economy of the system.

[0148] An embodiment of the application also provides a computer terminal, including a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to execute the method of any one of the above embodiments of the application, or run the system of any one of the above embodiments of the application.

[0149] Optionally, a memory for storing programs; the memory can include volatile memory (e.g., random-access memory (RAM) such as static random-access memory (SRAM), Double Data Rate Synchronous Dynamic Random Access Memory (DDR SDRAM), and the like) and / or non-volatile memory (e.g., flash memory). The memory is used to store computer programs (e.g., application programs, functional modules, and the like that implement the above-described methods), computer instructions, and the like, which can be stored in one or more memories in a partitioned manner. The computer programs, computer instructions, and the like described above can be invoked by the processor.

[0150] The processor is configured to execute the computer programs stored in the memory to implement the various steps in the methods or the various modules of the systems described above in the embodiments. Details can be found in the foregoing method and system embodiments.

[0151] The processor and the memory can be independent structures or integrated structures. When the processor and the memory are independent structures, the memory and the processor can be coupled by a bus.

[0152] An embodiment of the present application further provides a computer readable storage medium having stored thereon a computer program, which, when executed by a processor, is configured to perform the method of any one of the above-described embodiments of the present application or run the system of any one of the above-described embodiments of the present application.

[0153] The computer readable medium includes a computer storage medium and a communication medium. The communication medium includes any medium that facilitates transfer of a computer program from one place to another. A storage medium can be any available medium that can be accessed by a general purpose or special purpose computer. An exemplary storage medium is coupled to the processor such that the processor can read information from, and write information to, the storage medium. Of course, the storage medium can be a component of the processor. Suitable storage media include all volatile, nonvolatile, removable, and non-removable computer storage media, including but not limited to RAM, ROM, EEPROM, flash memory, and the like. Computer storage media includes physical computer storage media and communication media.

[0154] The multi-granularity energy feature parameter decoupling method and system based on the constraint uncertain set provided by the above embodiments of the present application enhance the flexibility and adaptability of energy management, and improve the energy utilization efficiency; by constructing a worst expected four-layer robust optimization model, the energy uncertainty and its category probability of electrolytic aluminum, ferroalloy, polysilicon, electric heating and energy storage system industrial loads on different time scales are considered, and through a multi-level optimization framework, the optimal solution under various uncertain environments is ensured; a multi-interval convex hull uncertain set method is proposed, the energy features of the industrial load are divided into multiple intervals according to the distribution and probability, and the corresponding convex hull is constructed to form a set, which can accurately describe the uncertainty of the energy features and reduce the conservativeness of the model; the Bregman alternating direction multiplier method is introduced to solve the non-convex problem, ensuring the effectiveness and efficiency of the optimization process; a column and constraint generation algorithm based on an alternating iteration strategy is designed, the model is constantly updated and decoupled, and the dynamic updating and decoupled optimization of the model are realized.

[0155] The details not described in the above embodiments of the present application are known in the art.

[0156] The specific embodiments of the present application are described above. It should be understood that the present application is not limited to the above specific embodiments, and those skilled in the art can make various modifications or changes within the scope of the claims, which does not affect the essential content of the present application.

Claims

1. A method for decoupling multi-granularity energy use feature parameters based on constrained uncertainty sets, characterized in that, The method comprises the following steps: dividing the energy consumption characteristics of the industrial load into multiple intervals according to the energy consumption distribution and the condition category probability, and constructing a corresponding convex hull to form an uncertain set based on the multi-interval convex hull; constructing a worst-case expected four-layer robust optimization model based on the energy consumption uncertainty of the industrial load at different time scales and the condition category; solving the non-convex problem of the worst-case expected four-layer robust optimization model by using a Bregman alternating direction multiplier method; based on the solution of the non-convex problem, using a column and constraint generation algorithm with an alternating iteration strategy to continuously update and decouple the optimization of the worst-case expected four-layer robust optimization model, so as to realize the decoupling of the multi-granularity energy consumption characteristics of the industrial load.

2. The method of claim 1, wherein, The step of dividing the energy consumption characteristics of the industrial load into multiple intervals according to the energy consumption distribution and the condition category probability, and constructing a corresponding convex hull to form an uncertain set based on the multi-interval convex hull comprises the following steps: obtaining actual energy consumption records of different industrial load types under various condition categories to obtain the energy consumption characteristics of the industrial load, wherein the energy consumption characteristics include actual power consumption and gas consumption of the industrial load under different scenarios; The energy consumption feature is divided into multiple intervals according to energy consumption distribution and condition category probability, a corresponding convex hull is constructed, and an uncertain set U based on the multi-interval convex hull is formed conv is expressed as: wherein represents the actual power consumption amount of the i-th industrial load at time t, category s; represents the gas consumption amount of the i-th industrial load at time t, category s; s represents a condition category, including: weather condition, seasonal change, market condition, equipment state, and production plan; N s represents the number of condition categories.

3. The method of claim 1, wherein, The step of constructing a worst-case expected four-layer robust optimization model comprises the following steps: establishing a basic framework of the worst-case expected four-layer robust optimization model, including: the first layer is a device operation state decision layer, which is used to optimize the start-stop logic and energy input-output strategy of the industrial load, and generate binary variables to control the operation state of the device; the second layer is a worst probability distribution layer, which is used to search for an uncertainty probability distribution that maximizes the total cost in a set condition category probability space; the third layer is a multi-interval convex hull constraint layer, which divides intervals based on the distribution of the energy consumption characteristics of the industrial load, and uses a convex hull set to represent the feasible region of the uncertainty parameter; the fourth layer is a real-time scheduling optimization layer, which is used to dynamically adjust the energy storage state and energy interaction parameters to ensure that the power limit, the change rate constraint and the load demand balance are met; establishing an objective function of a worst-case expected four-layer robust optimization model based on a preset probability distribution σ of the condition category s finding an optimal solution of the objective function in the worst case, which is a combination of energy features that maximizes the total cost of the system in the condition category probability space of 4. The method of claim 3, wherein, the objective function is represented as: where y i,t ,z i,t u i,t are binary variables representing the start-up, shut-down and running state of the i-th industrial load at time t; B buy,i,t ,B sell,i,t is a binary variable representing whether the i-th industrial load is importing power from the grid or exporting power to the grid; i represents the specific type of industrial load; is the start-up / shut-down cost coefficient of the i-th industrial load; T represents the time interval index; N represents the number of industrial load types; N s represents the number of condition categories; σ s represents the probability of the condition category; Ω σ represents the condition category probability space; U conv represents the set of uncertainties; α t ,β t represents the grid interaction loss coefficient and energy feedback efficiency coefficient; represents the actual power consumption of the i-th industrial load at time t in scenario s; represents the gas consumption of the i-th industrial load at time t in scenario s; represents the actual power imported from the grid by the i-th industrial load at time t in scenario s; represents the actual power exported to the grid by the i-th industrial load at time t in scenario s; SOC i,t,s represents the energy storage state of the i-th industrial load at time t; establishing constraints of the objective function, including: a device start-stop logic constraint, represented as: y i,t -z i,t = u i,t - u i,t-1 y i,t +z i,t ≤1 a power output limit constraint, represented as: wherein Pmin,i,s represents the minimum power output of the i-th industrial load of the s-th category; Pmax,i,s represents the maximum power output of the i-th industrial load of the s-th category; i,t,s Pact,i,s,t represents the actual power output of the i-th industrial load of the s-th category at time t; a power change rate limit constraint, represented as: wherein rubp i represents the rise rate limit of the i-th industrial load; rdbn i represents the fall rate limit of the i-th industrial load; represents the maximum power output of the i-th industrial load; a thermal energy storage constraint, represented as: wherein, represents the state of energy storage of the i-th industrial load at time t, category s; represents the charging amount of the i-th industrial load at time t, category s; represents the discharging amount of the i-th industrial load at time t, category s; represents the charging efficiency of the i-th industrial load; represents the discharging efficiency of the i-th industrial load; an electric power balance constraint, represented as: wherein denotes the demand of the i-th industrial load at time t.

5. The method of claim 4, wherein, the specific type of the industrial load includes electrolytic aluminum, ferroalloy, polycrystalline silicon, electric heating and energy storage system.

6. The method of claim 1, wherein: The step of solving the non-convex problem of the worst-case expected four-layer robust optimization model by using a Bregman alternating direction multiplier method comprises the following steps: Setting initial point x 0 and Lagrange multiplier u 0 ; updating the decision variable x, represented as: where x k+1 denotes the updated decision variable in the k+1th iteration; argmin denotes finding the value of x that minimizes the expression; f(x) denotes the objective function of the worst-case expected four-stage robust optimization model; A denotes the coefficient matrix in the linear constraints; x k denotes the value of the auxiliary variable in the kth iteration; c is a constant vector, denoting the constant value on the right side of the linear constraints; u k denotes the Lagrange multiplier or dual variable in the kth iteration, used to ensure that the constraints of the worst-case expected four-stage robust optimization model are satisfied; p denotes the penalty parameter, used to adjust the penalty level when the constraints are violated. updating the Lagrange multiplier, represented as: u k+1 = u k + p(Ax k+1 + By k+1 - c) where u k+1 denotes the updated Lagrange multiplier at the (k+1)th iteration.

7. The method of claim 1, wherein, The step of continuously updating and decoupling the optimization of the worst-case expected four-layer robust optimization model based on the solution of the non-convex problem by using a column and constraint generation algorithm with an alternating iteration strategy comprises the following steps: in each iteration of solving the non-convex problem of the worst-case expected four-layer robust optimization model, a new decision variable column and model constraint are dynamically generated according to the current solution, represented as: Ax k+1 +By k+1 -c+u k+1 =0 where x k+1 is a new decision variable, y k+1 is a new auxiliary variable, u k+1 is a new Lagrange multiplier; new model constraints are gradually added to the worst-case expected four-layer robust optimization model, and the optimization decoupling model is continuously updated until the model convergence condition is reached.

8. A multi-granularity energy feature parameter decoupling system based on constraint uncertain sets, characterized in that, The method comprises the following steps: An uncertainty set construction module, which is configured to divide energy consumption features of the industrial load into multiple intervals according to energy consumption distribution and condition category probability, and construct corresponding convex hulls to form a multi-interval convex hull-based uncertainty set; A robust optimization modeling module, which is configured to construct a worst-case expected four-layer robust optimization model based on energy consumption uncertainty of the industrial load at different time scales and condition categories by using the uncertainty set; A non-convex problem solving module, which is configured to solve a non-convex problem of the worst-case expected four-layer robust optimization model by using a Bregman alternating direction multiplier method; A dynamic updating decoupling optimization module, which is configured to constantly update and decoupling optimize the worst-case expected four-layer robust optimization model by using a column and constraint generation algorithm of an alternating iteration strategy based on a solution of the non-convex problem, so as to realize decoupling of multi-granularity energy consumption features of the industrial load.

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

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

Images