Multi-scenario air conditioner load prediction and flexible regulation method and system

Abstract

This invention belongs to the field of air conditioning load optimization and scheduling technology, and provides a method and system for multi-scenario air conditioning load prediction and flexible control. The method includes: firstly, constructing a multi-scenario air conditioning load characteristic model based on the equivalent thermal parameter ETP model, dividing the scenarios, and calculating the adjustable potential of air conditioning in each scenario; then, constructing a multi-objective optimization mathematical model with the objectives of optimal prediction accuracy, lowest control cost, and minimal loss of user comfort; next, using the Snow Ablation Algorithm (SAO) to construct the optimization architecture, achieving a balance between global search and local optimization through dynamic partitioning of exploration / utilization subpopulations; finally, introducing a non-dominated sorting algorithm to handle multi-objective conflicts, determining the optimal strategy through constraint-objective hierarchical sorting and congestion sorting. This invention adapts to the characteristics of air conditioning loads in multiple scenarios, improves prediction accuracy and control efficiency, and takes into account the grid peak-shaving needs, load aggregator revenue, and user comfort.

Description

Technical Field

[0001] This invention belongs to the field of air conditioning load optimization and scheduling technology, and specifically relates to a method and system for multi-scenario air conditioning load prediction and flexible control. Background Technology

[0002] Air conditioning loads are being widely adopted on the user side due to their cleanliness, efficiency, and controllability. However, the disorderly integration of a large number of air conditioning loads, while contributing to the low-carbon energy transition, also puts enormous pressure on the power grid's peak shaving. Current air conditioning load aggregation models and response potential assessment models have some problems, often ignoring the differences in user response characteristics and treating the characteristics of the group load as homogeneous, resulting in coarse aggregation models and inaccurate assessments of adjustable potential. Although load aggregators (LAs) have been introduced to participate in management, existing strategies mainly focus on the interaction between the power grid and LAs, with limited research on the effective game interaction and interest balance between LAs and a large number of dispersed users. At the same time, existing models are insufficient in terms of refined management, lacking systematic solutions for finely balancing user comfort and adjustable potential, and for optimizing aggregation to match the differentiated dispatch needs of the power grid based on response characteristics. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for predicting and flexibly controlling air conditioning loads in multiple scenarios, so as to solve the problems of low accuracy of existing multi-scenario air conditioning load prediction, insufficient optimization of control strategies, and poor coordination among multiple objectives.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method for predicting and flexibly controlling air conditioning loads in multiple scenarios, including: An operating model for a single air conditioner is established based on an equivalent thermal parameter model. Multiple scenario sets are divided according to date type, weather conditions, and user comfort preferences. The adjustable potential of the air conditioner is calculated within the comfort range corresponding to each scenario, forming a basic model of adjustable potential. Based on the adjustable potential basic model, with load forecast value, adjustable power scheduling value and control duration as optimization variables, and user comfort, equipment operation, scenario adaptation and system power balance as constraints, a multi-objective optimization problem is established with the goal of maximizing load forecast accuracy, optimizing control economy and minimizing user comfort loss. An improved snow melting algorithm is used to solve the multi-objective optimization problem and obtain the optimal solution set. Based on the optimal solution set, the system outputs the optimal prediction results of air conditioning load under multiple scenarios, a flexible control strategy including adjustable power scheduling values ​​and control duration, as well as corresponding electricity price incentive strategies and suggestions for adjusting user electricity consumption behavior.

[0005] Furthermore, the operation model of a single air conditioner is established based on the equivalent thermal parameter model. Multiple scenario sets are divided according to date type, weather conditions, and user comfort preferences. The adjustable potential of the air conditioner is calculated within the comfort range corresponding to each scenario, forming a basic model of adjustable potential, including: Single air conditioner operation model: Establishing the temperature dynamic equation:

[0006] in, Let t be the room temperature in scenario s at time s. Outdoor temperature , For the equivalent thermal resistance and heat capacity of the room in scenario s, This refers to the rated power of the air conditioner. Solar radiation power, The time step is used to characterize the temporal granularity of temperature changes, and the unit is hours. This refers to the operating status variables of the air conditioner. Multi-scene segmentation: Scene set ,in ={Weekdays, Weekends} ={Sunny, Rainy, Cloudy} ={High comfort, medium comfort, low comfort}, where the high comfort range is 23-25℃, the medium comfort range is 24-26℃, and the low comfort range is 25-27℃. The comfort range of scenario s is divided according to the principle of "user preset preference priority": if the user presets it to high comfort, it is classified as a high comfort scenario; if the user presets it to medium comfort, it is classified as a medium comfort scenario; and so on for low comfort scenarios. Adjustable potential calculation: comfort ranges corresponding to various scenarios Within, calculate the adjustable duration. And can be reduced in duration This allows for the determination of the adjustable power range. :

[0007]

[0008] Where H is the scheduling period; This is the minimum adjustable power of the air conditioner; This refers to the maximum adjustable power of the air conditioner, i.e., its rated power. , Let τ represent the air conditioner's operating status in scenario s at time τ, where 1 indicates the air conditioner compressor is running and 0 indicates the compressor is stopped.

[0009] Furthermore, based on the adjustable potential fundamental model, and using load forecast values, adjustable power scheduling values, and control duration as optimization variables, and user comfort, equipment operation, scenario adaptation, and system power balance as constraints, a multi-objective optimization problem is established with the objectives of maximizing load forecast accuracy, optimizing control economy, and minimizing user comfort loss. This includes: User comfort loss target

[0010] in, The user comfort loss in scenario s at time t; The actual room temperature at time τ under scenario s; , represents the upper and lower limits of the comfort range under scenario s; H is the scheduling period; Δt is the time step; max(0,a,b) means taking the maximum value among 0, a, and b, and only incurring a loss when the room temperature deviates from the comfort range; Prediction accuracy target:

[0011] in, Let be the total load forecast deviation for the entire scenario at time t; The load forecast value for scenario s at time t; Let be the actual load value under scenario s at time t; S is the set of all scenarios. Regulation of total cost function The specific expression is:

[0012] Among them, the electricity purchase cost from the grid by load aggregators for:

[0013] in, This refers to the electricity purchase price coefficient; The power purchased at time t under scenario s is expressed in kW. The maintenance cost of air conditioning equipment is :

[0014] in, , , All are operation and maintenance cost coefficients; The adjustable power scheduling value of the air conditioner in scenario s at time t; User response compensation cost is :

[0015] in, is the compensation coefficient for scenario s.

[0016] Furthermore, the specific expressions for each constraint are as follows: Comfort constraints:

[0017] in and These represent the upper and lower limits of the comfort range under scenario s; Equipment operating constraints:

[0018] and ,in , , which are the upper and lower limits of the rate of change of power; Scene adaptation constraints:

[0019] Let be the user response compensation coefficient in scenario s. Basic compensation coefficient, For scene adaptation coefficients, The value is based on the adjustable potential of the air conditioner under scenario s. The smaller the adjustable potential, the lower the value. The larger; Power balance constraints:

[0020] in, , where is the power imbalance in scenario s at time t; The total power supply in scenario s at time t; Let t be the total power load of the air conditioner in scenario s at time t.

[0021] Furthermore, the step of using an improved snow ablation algorithm to solve the multi-objective optimization problem and obtain the optimal solution set includes: Initialize the population matrix, dynamically divide the population into exploration / utilization subpopulations, and update the population evolution using Brownian motion and day-to-day models respectively: The specific formula and parameters for population initialization are set as follows:

[0022] Where Z is the initial population matrix with dimensions N×Dim, N is the population size, and Dim is the total number of optimization variables.

[0023] L is the lower bound vector of the optimization variables:

[0024] in =0kW This is the adjustable power lower limit. =0 hours; U is the upper bound vector of the optimization variables.

[0025] in =5kW, Adjustable power limit, =H; θ is a random parameter matrix uniformly distributed in the interval [0,1], with the same dimension as Z; The rules for dividing the subpopulation and the update formula are as follows: A three-stage dynamic adjustment strategy is adopted to adapt to the needs of "global search - balanced exploration and utilization - local optimization" in the iterative process: Initial Iteration Phase: Exploring Subpopulation Size =30, utilizing subpopulation size =20, by expanding the exploration range, premature convergence of the algorithm is avoided; Mid-term iteration phase: Exploring subpopulation size =25, utilizing subpopulation size =25, balancing global search breadth and local optimization accuracy; Late iteration phase: Exploring subpopulation size =20, utilizing subpopulation size =30, strengthen the local development of the region near the optimal solution, and improve the convergence accuracy; where the total population size N=50, the subpopulation size of each stage automatically switches with the number of iterations. (2) Exploring the formula for subpopulation renewal:

[0026] in, and These are the optimization strategy vectors for the i-th individual in the exploration subpopulation in rounds t and t+1, respectively. For the elite strategy in round t, a strategy vector is randomly selected with equal probability from two types of strategies: ① a strategy vector randomly selected from the top 3 best individuals in the population; ② the central strategy vector of the top 50% of individuals, as shown in the following formula:

[0027] Only the arithmetic mean of the top 50% of individuals is calculated; Let the vector be a Brownian motion vector, with elements following a normal distribution of mean 0 and standard deviation 1, and dimension 1. Consistent; The optimal individual strategy for round t; For the population-centric strategy in round t, ; ∈[0,1] are random parameters that follow a uniform distribution; This refers to the element-wise multiplication operation of vectors. The update formula and related parameters for the subpopulation are defined as follows:

[0028] Where M is the snowmelt rate, and its calculation formula is:

[0029]

[0030] This represents the maximum number of iteration rounds. ∈[0,1] is a random parameter that follows a uniform distribution; the snowmelt rate M is used to adjust the convergence speed of the subpopulation towards the optimal individual strategy.

[0031] Furthermore, obtaining the optimal solution set includes: Elite strategies and optimal individual strategies are determined by constraint-objective hierarchical ranking and crowding ranking: Constraint sub-level classification: Divided into 4 levels according to the number of constraints satisfied, with level 4 being the highest: satisfying all 4 constraints; level 3: satisfying 3 constraints; level 2: satisfying 2 constraints; and level 1 being the lowest: satisfying 1 or fewer constraints. Target sub-level classification: Within the same constraint sub-level, based on the multi-objective non-dominant relationship, for any two individuals a and b, if the prediction error of a is ≤ b, the regulation cost is ≤ b, and the comfort loss is ≤ b, and at least one of them is strictly less than b, then a dominates b; non-dominant individuals are classified into the same target sub-level. Congestion calculation formula:

[0032] Where p is the number of individuals within the target sub-level; To predict the target accuracy value, i.e. ; To regulate economic target values; is the target value for comfort; i is the sorting index of the individual within the target sub-level; coefficient 2 is used to standardize the range of crowding values, making the crowding levels of different target sub-levels comparable. Individuals are sorted from highest to lowest constraint sub-level, and then from highest to lowest crowding level within the same constraint sub-level. The top 20% of individuals are selected as the elite strategy pool for population updates in the next iteration.

[0033] Furthermore, based on the optimal solution set, the output of optimal air conditioning load prediction results for multiple scenarios, a flexible control strategy including adjustable power scheduling values ​​and control duration, and corresponding electricity price incentive strategies and user electricity behavior adjustment suggestions includes: The specific process and convergence criteria for iterative solution are as follows: (1) Iterative process: Each iteration executes "population fitness assessment → non-dominant ranking → elite strategy selection → subpopulation division → population evolution update" in sequence. The higher the score, the better the fitness. (2) Convergence criterion: If the change in the fitness score of the best individual is ≤1×10^{-3} and the crowding variance of the Pareto optimal solution set is ≤5×10^{-2} in 50 consecutive iterations, then it is considered converged; (3) Strategy output: After convergence, select the strategy with the highest comprehensive score from the Pareto optimal solution set, and output the load forecast value, adjustable power scheduling value, regulation execution time, and corresponding electricity price incentive strategy and user electricity behavior adjustment suggestions for each scenario.

[0034] Secondly, the present invention provides a multi-scenario air conditioning load prediction and flexible control system, comprising: The modeling module is used to establish an operating model of a single air conditioner based on an equivalent thermal parameter model. It divides multiple scenario sets according to date type, weather conditions and user comfort preferences, and calculates the adjustable potential of the air conditioner within the comfort range corresponding to each scenario, forming a basic model of adjustable potential. The optimization model building module is used to establish a multi-objective optimization problem based on the adjustable potential basic model, with load forecast value, adjustable power scheduling value and control duration as optimization variables, and user comfort, equipment operation, scenario adaptation and system power balance as constraints. The goal is to achieve the highest load forecast accuracy, the best control economy and the minimum loss of user comfort. The solution module is used to solve the multi-objective optimization problem using an improved snow ablation algorithm to obtain the optimal solution set; The execution module is used to output the optimal prediction results of air conditioning load under multiple scenarios, a flexible control strategy including adjustable power scheduling values ​​and control duration, and corresponding electricity price incentive strategies and suggestions for adjusting user electricity consumption behavior based on the optimal solution set.

[0035] Thirdly, the present invention provides a computer device, including 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 steps of the multi-scenario air conditioning load prediction and flexible control method.

[0036] Fourthly, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the multi-scenario air conditioning load prediction and flexible control method.

[0037] Compared with the prior art, the present invention has the following technical effects: This invention, through multi-scenario refined modeling, introduces a three-dimensional scenario segmentation factor of time period, weather, and preference, effectively overcoming the adaptability defects of traditional single-scenario models and significantly improving the load forecasting accuracy and control strategy applicability under multiple operating conditions.

[0038] This invention employs the Non-Dominated Sort Snow Melting Algorithm (NSAO), which effectively avoids the problem of traditional optimization algorithms easily getting trapped in local optima by exploring / utilizing the dynamic cooperative evolution mechanism of subpopulations, and significantly improves the solution efficiency and global optimization capability of large-scale air conditioning load group optimization problems.

[0039] This invention introduces a non-dominated sorting multi-objective processing mechanism. Through a collaborative strategy of constraint hierarchical, objective hierarchical, and congestion-based sorting, it achieves multi-objective coordinated optimization of prediction accuracy, control cost, and user comfort, ensuring the diversity and uniformity of Pareto solution sets.

[0040] This invention can improve the accuracy of load forecasting in multiple scenarios while optimizing the allocation of flexible resources for air conditioning loads, effectively reducing the pressure on power grid peak shaving and control costs, and balancing the revenue of load aggregators with user comfort and economy. It has high engineering application value and social benefits in the fields of virtual power plants and demand response. Attached Figure Description

[0041] Figure 1 This is a flowchart of a multi-scenario air conditioning load prediction and flexible control strategy based on a non-dominated sorting snow melting algorithm, provided by an exemplary embodiment; Figure 2 This is a schematic diagram of the hardware structure of a computer device provided in an exemplary embodiment; Figure 3 This is a block diagram of a multi-scenario air conditioning load prediction and flexible control device provided in an exemplary embodiment. Detailed Implementation

[0042] The present invention will be further described below with reference to the accompanying drawings: Example 1: This invention provides a method for predicting and flexibly controlling air conditioning loads in multiple scenarios, including: An operating model for a single air conditioner is established based on an equivalent thermal parameter model. Multiple scenario sets are divided according to date type, weather conditions, and user comfort preferences. The adjustable potential of the air conditioner is calculated within the comfort range corresponding to each scenario, forming a basic model of adjustable potential. Based on the adjustable potential basic model, with load forecast value, adjustable power scheduling value and control duration as optimization variables, and user comfort, equipment operation, scenario adaptation and system power balance as constraints, a multi-objective optimization problem is established with the goal of maximizing load forecast accuracy, optimizing control economy and minimizing user comfort loss. An improved snow melting algorithm is used to solve the multi-objective optimization problem and obtain the optimal solution set. Based on the optimal solution set, the system outputs the optimal prediction results of air conditioning load under multiple scenarios, a flexible control strategy including adjustable power scheduling values ​​and control duration, as well as corresponding electricity price incentive strategies and suggestions for adjusting user electricity consumption behavior.

[0043] This invention provides a multi-scenario air conditioning load prediction and flexible control strategy based on the non-dominated sorting snow melting algorithm. Through multi-scenario refined modeling, intelligent optimization algorithms and multi-objective collaborative processing mechanisms, it achieves accurate prediction and flexible optimization control of air conditioning load groups in multiple spatiotemporal dimensions, effectively improving multi-scenario adaptability and prediction accuracy, while taking into account grid demand, load aggregator revenue, user comfort and economy.

[0044] Example 2: This invention provides a method for predicting and flexibly controlling air conditioning loads in multiple scenarios, including: Step 1): Establish a basic model of the load characteristics of a single air conditioner, construct a physical model of air conditioner operation based on the equivalent thermal parameter (ETP) model, establish a temperature dynamic equation through parameters such as indoor and outdoor temperature, solar radiation, and rated power of the air conditioner to characterize the thermal balance mechanism of air conditioner operation; based on user comfort range constraints, calculate the adjustable / lowerable duration and adjustable power range of the air conditioner to form a basic model of adjustable potential.

[0045] Step 2): Construct a multi-scenario refined modeling framework, introduce three-dimensional scenario segmentation factors such as time period type, meteorological conditions, and user comfort preferences, and divide the load scenarios into weekday / weekend, sunny / rainy / cloudy, and high / medium / low comfort preference combinations to form a multi-scenario set; based on the single air conditioner characteristic parameters obtained in Step 1, establish a dynamic model of the air conditioner's adjustable potential under each scenario to achieve accurate characterization of load characteristics under different operating conditions.

[0046] Step 3): Construct a multi-objective optimization mathematical model, using the predicted air conditioning load, adjustable power scheduling value, and control execution duration under each scenario as core optimization variables to cover scheduling needs across multiple time scales; set comfort constraints, equipment operation constraints, scenario adaptation constraints, and power balance constraints; construct three major optimization objectives: minimize multi-scenario load prediction deviation, minimize total control cost, and minimize user comfort loss. Through the linkage mechanism between compensation coefficient and adjustable potential in the scenario adaptation constraint, achieve a balance between LA cost control and user response incentives, solving the problem of insufficient interaction between LA and a large number of dispersed users in existing strategies.

[0047] Step 4): Construct an optimization architecture based on the Snow Ablation Algorithm (SAO), and initialize the population with optimization variables as the dimension; adopt a dynamic subpopulation partitioning strategy to divide the population into an exploration subpopulation and a utilization subpopulation, and dynamically adjust the ratio of the two with the number of iterations to balance global exploration and local utilization; the exploration subpopulation expands the search range based on the Brownian motion model, and the utilization subpopulation strengthens the search near the optimal solution based on the day-to-day model, thereby improving optimization accuracy and convergence speed.

[0048] Step 5): The non-dominated sorting algorithm is used to solve multi-objective conflicts. First, the constraint sub-levels are divided according to the constraint satisfaction. Then, within each level, the target sub-levels are dynamically divided according to the target performance. The crowding degree of individuals within the same target sub-level is calculated and sorted in descending order to ensure strategy diversity. The precise ranking of the entire population is obtained by splicing the sub-levels, the Pareto optimal solution set is determined, and the optimal load prediction results and flexible control strategies under multiple scenarios are output.

[0049] Optionally, in step 2), the multi-scenario modeling framework achieves fine-grained scenario segmentation by combining three-dimensional scenario factors of time period, weather, and preference, which significantly improves the model's adaptability to different operating conditions and its prediction accuracy.

[0050] Optionally, in step 4), the subpopulation dynamic partitioning mechanism adaptively adjusts the ratio of exploring and utilizing subpopulations according to the iteration process. In the early stage, it strengthens global search to avoid premature convergence, and in the later stage, it enhances local development to improve optimization accuracy, effectively avoiding getting trapped in local optima.

[0051] Optionally, in step 5), the non-dominated sorting algorithm combines constraint hierarchies and crowding calculations to effectively handle multi-objective conflicts, ensure the uniform distribution of the Pareto front, and avoid strategies from clustering in local optima.

[0052] Example 3: This invention provides a method for predicting and flexibly controlling air conditioning loads in multiple scenarios, including: Step 101: Construct a multi-scenario air conditioning load characteristic model. Based on the equivalent thermal parameter (ETP) model, divide typical scenarios such as weekday / weekend, sunny / rainy / cloudy, and high / medium / low comfort preference, and calculate the adjustable potential of air conditioning in each scenario.

[0053] Step 102: Construct a multi-objective optimization mathematical model with the objectives of achieving optimal prediction accuracy, lowest control cost, and minimal loss of user comfort, and introduce constraints such as equipment operation, comfort, and power balance.

[0054] Step 103: Use the Snow Ablation Algorithm (SAO) to construct an optimized architecture. By dynamically dividing the exploration / utilization subpopulation, a balance between global search and local optimization is achieved.

[0055] Step 104: Introduce a non-dominated sorting algorithm to handle multi-objective conflicts. Determine the optimal flexible control strategy through constraint-objective hierarchical sorting and congestion sorting to adapt to the load characteristics of multiple scenarios and improve control efficiency.

[0056] Step 105: Iteratively solve and output strategies. Through multiple rounds of iterative optimization, output the optimal prediction results of air conditioning load under multiple scenarios, flexible control strategies (including electricity price incentive strategies), and suggestions for adjusting user electricity consumption behavior.

[0057] Furthermore, a multi-scenario three-dimensional factor combination is used to classify the scenarios. This scenario classification framework is constructed based on the three-dimensional factors of "time period type - meteorological conditions - comfort preference". Through the full combination, 18 typical scenarios (2×3×3) are formed to achieve fine-grained characterization of load characteristics. Among them, the time period type factor (S1) distinguishes between weekdays (users have regular work and rest schedules, and air conditioning use is concentrated from 8 am to 10 pm) and weekends (use time periods are dispersed and the duration is longer); the meteorological conditions factor (S2) covers three categories: sunny (strong solar radiation, high peak air conditioning load), rainy (weak radiation, relatively stable load), and cloudy (moderate radiation, load between sunny and rainy); the comfort preference factor (S3) is divided according to temperature range: high comfort (23-25℃), medium comfort (24-26℃), and low comfort (25-27℃). Overlapping ranges are defined according to the "user preset priority" principle. For example, when the user presets high comfort, 24-25℃ is classified as a high comfort scenario, ensuring that the scenario classification matches the actual needs of users.

[0058] Furthermore, an optimization architecture for the Snow Ablation Algorithm (SAO) and dynamic subpopulation partitioning were constructed. This optimization architecture, centered on the population matrix, balances global search and local optimization capabilities through three-stage dynamic partitioning of the exploration / utilization subpopulation. In the initial iteration phase (rounds 0-300), the size of the exploration subpopulation is expanded (30 individuals) to prevent premature convergence. In the mid-term iteration phase (rounds 301-700), the sizes of the two subpopulations are balanced (25 individuals each), taking into account both search breadth and optimization accuracy. In the later iteration phase (rounds 701-1000), the utilization subpopulation (30 individuals) is strengthened to improve local development efficiency near the optimal solution. The exploration subpopulation is updated based on a Brownian motion model to expand the search range. The subpopulation, combined with a day-to-day model and snowmelt rate adjustment, accelerates convergence towards the optimal solution. The overall population size is fixed at 50, allowing for adaptive iteration without manual intervention.

[0059] Further, a non-dominated ranking process and constraint-objective grading are implemented, consisting of two steps: constraint grading and objective grading. First, based on the number of constraints satisfied, four constraint sub-levels are established (Level 4: all four constraints satisfied; Level 3: three satisfied; Level 2: two satisfied; Level 1: one or fewer satisfied), prioritizing individuals with higher constraint levels. Within the same constraint level, objective sub-levels are established based on multi-objective non-dominated relationships, with dominant individuals (better at one objective and not inferior at others) assigned to higher levels. Finally, the crowding degree of individuals within each objective sub-level is calculated and sorted in descending order to ensure strategy diversity. The crowding degree calculation comprehensively considers the differences between three objectives: prediction accuracy, control costs, and comfort loss. Standardization ensures comparability between different sub-levels, ultimately selecting the top 20% of individuals to form an elite strategy pool for the next iteration.

[0060] In this embodiment, a cluster of 1000 household air conditioners in a certain province is used as the application object. The air conditioner types include fixed frequency and variable frequency, and the rated power (P) is... naten Both are 1.5kW, with an equivalent thermal resistance (R). s The value is taken as 2.8℃ / kW, and the equivalent heat capacity (C) s The value was set to 12 kWh / ℃, the time step (Δt) was set to 30 minutes, and the scheduling period (H) was set to 24 hours to verify the effectiveness of the strategy of the present invention.

[0061] Further, step 101: Construction of multi-scenario air conditioning load characteristic model and calculation of adjustable potential. (1) ETP model parameter setting Temperature dynamic equation based on the operation of a single air conditioner:

[0062] Among them, outdoor temperature Scene settings: 35℃ for sunny scenes, 28℃ for rainy scenes, and 32℃ for cloudy scenes; solar radiation power. : 0.8kW for sunny scene, 0.2kW for rainy scene, and 0.5kW for cloudy scene; Air conditioning operating status variables Based on actual operational data, under typical weekday sunny and comfortable conditions, The system is set to 1 (operating) from 8:00 AM to 10:00 PM, and to 0 (shutdown) during other times.

[0063] (2) Calculation of Adjustable Potential In the scenario of "weekday + sunny + high comfort level" (S=S 11 ×S 21 ×S 31 Taking [23℃, 25℃] as an example, the comfort range is calculated as follows: [Calculate the adjustable duration]. With adjustable duration :

[0064] Calculations show that in this scenario =8h, =16h; adjustable power range is [0.6kW, 1.5kW].

[0065] Similarly, in the scenario of "weekend + rain + low comfort" (S=S 12 ×S 22 ×S 33 Within this range, the comfort level is [25℃, 27℃]. =28℃, Q an , s (t) = 0.2kW, x s (t) The runtime is distributed (9 AM - 11 PM), and the calculation is as follows: =10h, =14h, the adjustable power range is also [0.6kW, 1.5kW].

[0066] Further, step 102, constructing a multi-objective optimization mathematical model.

[0067] (1) Optimize target setting Prediction accuracy target: Minimize the target value, in kW; Cost control target: Minimize the objective value, in units of yuan; Electricity purchase cost k9 i d is set at 0.35 yuan / kWh (peak hours); Operation and maintenance costs ; User compensation costs (High comfort scenario, λ) s =1.2); Comfort loss target: Minimize the target value, in °C h.

[0068] (2) Setting constraints Comfort constraints: [23℃, 25℃] (high comfort scenario); Equipment operating constraints: [0.6kW, 1.5kW], and -0.5kW ≤ P a dj, s t -P a dj, s (t-1)≤0.5kW; Scene adaptation constraints: High comfort scenarios s =1.2, moderate comfort λ s =1.0, low comfort λ s =0.8; Power balance constraints:

[0069] Further, in step 103, the Snow Ablation Algorithm (SAO) is optimized and its architecture is implemented.

[0070] 1) Population initialization The population matrix Z has a dimension of 50×54 (N=50, Dim=3×18=54, 18 scenarios × 3 optimization variables: predicted value, scheduling value, and control duration); the lower bound of the optimization variables is L=[0kW, 0.6kW, 0h]. , upper limit U=[5kW,1.5kW,24h] The random parameter matrix θ has elements that follow a uniform distribution in the range [0,1]. The initial population is Z = L + θ. (UL).

[0071] (2) Subpopulation dynamic update Exploring subpopulation updates (based on Brownian motion): Z i t+ ¹=Elite t +BM i t [θ1 (G1 t -Z i t )+(1-θ1) (Z_c t -Z i t )] Among them, Elite t For the t-round elite strategy, randomly select from the center vectors of the top 3 best individuals or the top 50% of individuals in the population; BM i t Let θ1 be a Brownian motion vector (mean 0, standard deviation 1); θ1∈[0,1] is uniformly distributed. Update using subpopulation (combined with a day-to-day model): Z i t+ ¹=M G1 t +BM i t [θ2 (G1 t -Zi t )+(1-θ2) (Z_c t -Z i t )] Wherein, the snowmelt rate M = (0.35 + 0.25) (e^(t / 1000)-1) / (e-1)) e^(t / 1000), where t is the current iteration number, max=1000; θ2∈[0,1] is uniformly distributed.

[0072] Further, in step 104, the non-dominated sorting algorithm is implemented.

[0073] Taking a certain round of iterative population as an example, the constraint classification results are as follows: 32 individuals satisfy all 4 constraints (Level 4), 12 individuals satisfy 3 constraints (Level 3), and 6 individuals satisfy 2 constraints (Level 2). Among the Level 4 individuals, the target sub-levels are divided based on non-dominant relationships: 8 individuals are non-dominant individuals (Target Sub-Level 1), and 24 individuals are dominated individuals (Target Sub-Level 2). The crowding degree of Target Sub-Level 1 is calculated, taking 3 individuals as an example: Individual A: prediction bias 1.2kW, control cost 850 yuan, comfort loss 2.3℃. h; Individual B: Prediction deviation 1.0kW, control cost 880 yuan, comfort loss 2.1℃ h; Individual C: Prediction deviation 1.3kW, control cost 830 yuan, comfort loss 2.5℃ h; Crowding degree calculated _A=2.8, _B=3.2, _C=2.6, sorted in descending order as B>A>C, the top 20% (2 individuals) are selected to enter the elite strategy pool.

[0074] Further, in step 105, iterative solution and strategy output.

[0075] Iterative convergence criterion: The change in the fitness score of the optimal individual over 50 consecutive rounds is ≤1×10. - ³, and the variance of the crowding degree of the Pareto optimal solution set is ≤5×10. - ². After simulation iterations, the algorithm converges in 820 rounds, outputting the optimal strategy for each scenario: Load forecast results: Forecast value P for the "weekday + sunny + high comfort" scenario at t=14 (peak load period) pred =1.42kW, actual value P a c t =1.45kW, deviation 0.03kW, average prediction deviation for the entire scenario ≤0.05kW; Flexible control strategy: Adjustable power dispatch value P during peak hours (10:00-14:00) a dj=1.2kW, during off-peak hours (0:00-6:00) P a dj=0.8kW; the electricity price incentive strategy is a peak-hour electricity purchase price of k9. i d = 0.4 yuan / kWh, 0.2 yuan / kWh during off-peak hours; user compensation coefficient k_c for high comfort scenarios. omp =0.24 yuan / kWh; User electricity usage behavior adjustment suggestions: On sunny weekdays with high comfort levels, it is recommended that users raise their air conditioner temperature setting from 24℃ to 25℃ (still within the high comfort range), which can reduce the daily comfort loss by 0.8℃. h, while reducing LA control costs by approximately 12%.

[0076] In another embodiment of the present invention, a multi-scenario air conditioning load prediction and flexible control system is provided, which can be used to implement the above-mentioned multi-scenario air conditioning load prediction and flexible control method. Specifically, the system includes: The modeling module is used to establish an operating model of a single air conditioner based on an equivalent thermal parameter model. It divides multiple scenario sets according to date type, weather conditions and user comfort preferences, and calculates the adjustable potential of the air conditioner within the comfort range corresponding to each scenario, forming a basic model of adjustable potential. The optimization model building module is used to establish a multi-objective optimization problem based on the adjustable potential basic model, with load forecast value, adjustable power scheduling value and control duration as optimization variables, and user comfort, equipment operation, scenario adaptation and system power balance as constraints. The goal is to achieve the highest load forecast accuracy, the best control economy and the minimum loss of user comfort. The solution module is used to solve the multi-objective optimization problem using an improved snow ablation algorithm to obtain the optimal solution set; The execution module is used to output the optimal prediction results of air conditioning load under multiple scenarios, a flexible control strategy including adjustable power scheduling values ​​and control duration, and corresponding electricity price incentive strategies and suggestions for adjusting user electricity consumption behavior based on the optimal solution set.

[0077] Figure 2 This is a schematic diagram of the hardware structure of a computer device provided in an exemplary embodiment. Please refer to it. Figure 2At the hardware level, the device includes a processor 502, an internal bus 504, a network interface 506, memory 508, and non-volatile memory 510, and may also include other hardware required for its functions. One or more embodiments of the present invention can be implemented in software, for example, the processor 502 reads the corresponding computer program from the non-volatile memory 510 into memory 508 and then runs it. Of course, in addition to software implementation, one or more embodiments of the present invention do not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0078] Please refer to Figure 3 A multi-scenario air conditioning load prediction and flexible control device can be applied to, for example... Figure 3 The device shown, in order to implement the technical solution of the present invention, includes: Modeling module 601 is used to build multi-scenario air conditioning load characteristic models and adjustable potential basic models. It takes data such as outdoor temperature, solar radiation, and air conditioning parameters as input and outputs adjustable potential parameters for each scenario. The optimization model building module 602 is used to design optimization variables, constraints and multi-objective functions, integrates parameter libraries such as scenario adaptation coefficient and electricity price coefficient, and outputs a complete multi-objective optimization mathematical model. The algorithm optimization module 603 is used to implement the population initialization, evolution update and non-dominated sorting of the snow melting algorithm. It has built-in hyperparameters such as population size and maximum number of iterations, and outputs the Pareto optimal solution set. Solver module 604 is used to execute the iterative solution process, determine the optimization state based on the convergence criterion, and output the optimal prediction result and control strategy; The execution module 605 is used to convert the control strategy into executable instructions and send them to the air conditioning cluster controller and load aggregator platform, and simultaneously output the electricity price incentive signal to the user terminal. The data synchronization module 606 is used to collect air conditioning operation data, meteorological data, and power grid dispatch demand data in real time, providing dynamic input for each module; at the same time, it records the control effect data, updates the core parameters regularly, and realizes closed-loop iterative optimization.

[0079] Optionally, the modeling module 601 is specifically used to: establish a dynamic equation for the temperature of a single air conditioner based on the equivalent thermal parameter (ETP) model, introduce three-dimensional scene factors of time period-meteorology-preference to divide 18 typical scenarios, calculate the adjustable / lowering duration and adjustable power range of the air conditioner under each scenario according to the comfort interval, and form a basic model of adjustable potential.

[0080] Optionally, the optimization model construction module 602 is specifically used to: use load prediction values, adjustable power scheduling values ​​and control duration under multiple scenarios as optimization variables, set comfort, equipment operation, scenario adaptation and power balance constraints, construct a multi-objective function with optimal prediction accuracy, lowest control cost and minimum user comfort loss, and achieve a balance between LA and user interests by linking and compensating costs and adjustable potential through scenario adaptation coefficients.

[0081] Optionally, the algorithm optimization module 603 is specifically used for: initializing the population matrix, dynamically dividing the exploration / utilization subpopulation according to the iteration process, updating the exploration subpopulation based on Brownian motion to expand the search range, utilizing the subpopulation in combination with snowmelt rate adjustment to achieve local optimization; and using constraint-objective hierarchical sorting and crowding calculation to handle multi-objective conflicts and screen the elite strategy pool.

[0082] Optionally, the solution module 604 is specifically used to: iteratively solve the problem according to the process of "population fitness assessment → non-dominant ranking → elite strategy selection → subpopulation division → population evolution update", determine the optimization state based on the convergence criterion, and select the strategy with the highest comprehensive score from the Pareto optimal solution set for output.

[0083] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer, which can take the form of a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email sending and receiving device, game console, tablet computer, wearable device, or any combination of these devices.

[0084] The module division in this embodiment of the invention is illustrative and represents only one logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional modules in the various embodiments of the invention can be integrated into a single processor, exist as separate physical entities, or be integrated into a single module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0085] In another embodiment of the present invention, a computer device is provided, comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in the computer storage medium to achieve a corresponding method flow or corresponding function. The processor described in this embodiment of the present invention can be used in the operation of a multi-scenario air conditioning load prediction and flexible control method.

[0086] In another embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the multi-scenario air conditioning load prediction and flexible control method in the above embodiments.

[0087] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0088] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0089] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0090] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0091] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for predicting and flexibly controlling air conditioning load in multiple scenarios, characterized in that, include: An operating model for a single air conditioner is established based on an equivalent thermal parameter model. Multiple scenario sets are divided according to date type, weather conditions, and user comfort preferences. The adjustable potential of the air conditioner is calculated within the comfort range corresponding to each scenario, forming a basic model of adjustable potential. Based on the adjustable potential basic model, with load forecast value, adjustable power scheduling value and control duration as optimization variables, and user comfort, equipment operation, scenario adaptation and system power balance as constraints, a multi-objective optimization problem is established with the goal of maximizing load forecast accuracy, optimizing control economy and minimizing user comfort loss. An improved snow melting algorithm is used to solve the multi-objective optimization problem and obtain the optimal solution set. Based on the optimal solution set, the system outputs the optimal prediction results of air conditioning load under multiple scenarios, a flexible control strategy including adjustable power scheduling values ​​and control duration, as well as corresponding electricity price incentive strategies and suggestions for adjusting user electricity consumption behavior.

2. The method for multi-scenario air conditioning load prediction and flexible control according to claim 1, characterized in that, The operation model of a single air conditioner is established based on the equivalent thermal parameter model. Multiple scenario sets are divided according to date type, weather conditions, and user comfort preferences. The adjustable potential of the air conditioner is calculated within the comfort range corresponding to each scenario, forming a basic adjustable potential model, including: Single air conditioner operation model: Establishing the temperature dynamic equation: in, Let t be the room temperature in scenario s at time s. Outdoor temperature , For the equivalent thermal resistance and heat capacity of the room in scenario s, This refers to the rated power of the air conditioner. Solar radiation power, The time step is used to characterize the temporal granularity of temperature changes, and the unit is hours. This refers to the operating status variables of the air conditioner. Multi-scene segmentation: Scene set ,in ={Weekdays, Weekends} ={Sunny, Rainy, Cloudy} ={High comfort, medium comfort, low comfort}, where the high comfort range is 23-25℃, the medium comfort range is 24-26℃, and the low comfort range is 25-27℃. The comfort range of scenario s is divided according to the principle of "user preset preference priority": if the user presets it to high comfort, it is classified as a high comfort scenario; if the user presets it to medium comfort, it is classified as a medium comfort scenario; and so on for low comfort scenarios. Adjustable potential calculation: comfort ranges corresponding to various scenarios Within, calculate the adjustable duration. And can be reduced in duration This allows for the determination of the adjustable power range. : Where H is the scheduling period; This is the minimum adjustable power of the air conditioner; This refers to the maximum adjustable power of the air conditioner, i.e., its rated power. , Let τ represent the air conditioner's operating status in scenario s at time τ, where 1 indicates the air conditioner compressor is running and 0 indicates the compressor is stopped.

3. The method for multi-scenario air conditioning load prediction and flexible control according to claim 1, characterized in that, Based on the adjustable potential fundamental model, and using load forecast values, adjustable power scheduling values, and control duration as optimization variables, and user comfort, equipment operation, scenario adaptation, and system power balance as constraints, a multi-objective optimization problem is established with the objectives of maximizing load forecast accuracy, optimizing control economy, and minimizing user comfort loss. This includes: User comfort loss target in, The user comfort loss in scenario s at time t; The actual room temperature at time τ under scenario s; , represents the upper and lower limits of the comfort range under scenario s; H is the scheduling period; Δt is the time step; max(0,a,b) means taking the maximum value among 0, a, and b, and only incurring a loss when the room temperature deviates from the comfort range; Prediction accuracy target: in, Let be the total load forecast deviation for the entire scenario at time t; The load forecast value for scenario s at time t; Let be the actual load value under scenario s at time t; S is the set of all scenarios. Regulation of total cost function The specific expression is: Among them, the electricity purchase cost from the grid by load aggregators for: in, This refers to the electricity purchase price coefficient; The power purchased at time t under scenario s is expressed in kW. The maintenance cost of air conditioning equipment is : in, , , All are operation and maintenance cost coefficients; The adjustable power scheduling value of the air conditioner in scenario s at time t; User response compensation cost is : in, is the compensation coefficient for scenario s.

4. The method for multi-scenario air conditioning load prediction and flexible control according to claim 3, characterized in that, The specific expressions for each constraint are as follows: Comfort constraints: in and These represent the upper and lower limits of the comfort range under scenario s; Equipment operating constraints: and ,in , , which are the upper and lower limits of the rate of change of power; Scene adaptation constraints: Let be the user response compensation coefficient in scenario s. Basic compensation coefficient, For scene adaptation coefficients, The value is based on the adjustable potential of the air conditioner under scenario s. The smaller the adjustable potential, the lower the value. The larger; Power balance constraints: in, , where is the power imbalance in scenario s at time t; The total power supply in scenario s at time t; Let t be the total power load of the air conditioner in scenario s at time t.

5. The method for multi-scenario air conditioning load prediction and flexible control according to claim 1, characterized in that, The method of using an improved snow ablation algorithm to solve the multi-objective optimization problem and obtain the optimal solution set includes: Initialize the population matrix, dynamically divide the population into exploration / utilization subpopulations, and update the population evolution using Brownian motion and day-to-day models respectively: The specific formula and parameters for population initialization are set as follows: Where Z is the initial population matrix with dimensions N×Dim, N is the population size, and Dim is the total number of optimization variables. L is the lower bound vector of the optimization variables: in =0kW This is the adjustable power lower limit. =0 hours; U is the upper bound vector of the optimization variables. in =5kW, Adjustable power limit, =H; θ is a random parameter matrix uniformly distributed in the interval [0,1], with the same dimension as Z; The rules for dividing the subpopulation and the update formula are as follows: A three-stage dynamic adjustment strategy is adopted to adapt to the needs of "global search - balanced exploration and utilization - local optimization" in the iterative process: Initial Iteration Phase: Exploring Subpopulation Size =30, utilizing subpopulation size =20, by expanding the exploration range, premature convergence of the algorithm is avoided; Mid-term iteration phase: Exploring subpopulation size =25, utilizing subpopulation size =25, balancing global search breadth and local optimization accuracy; Late iteration phase: Exploring subpopulation size =20, utilizing subpopulation size =30, strengthen the local development of the region near the optimal solution, and improve the convergence accuracy; where the total population size N=50, the subpopulation size of each stage automatically switches with the number of iterations. (2) Exploring the formula for subpopulation renewal: in, and These are the optimization strategy vectors for the i-th individual in the exploration subpopulation in rounds t and t+1, respectively. For the elite strategy in round t, a strategy vector is randomly selected with equal probability from two types of strategies: ① a strategy vector randomly selected from the top 3 best individuals in the population; ② the central strategy vector of the top 50% of individuals, as shown in the following formula: Only the arithmetic mean of the top 50% of individuals is calculated; Let the vector be a Brownian motion vector, with elements following a normal distribution of mean 0 and standard deviation 1, and dimension 1. Consistent; The optimal individual strategy for round t; For the population-centric strategy in round t, ; ∈[0,1] are random parameters that follow a uniform distribution; This refers to the element-wise multiplication operation of vectors. The update formula and related parameters for the subpopulation are defined as follows: Where M is the snowmelt rate, and its calculation formula is: This represents the maximum number of iteration rounds. ∈[0,1] is a random parameter that follows a uniform distribution; the snowmelt rate M is used to adjust the convergence speed of the subpopulation towards the optimal individual strategy.

6. The method for multi-scenario air conditioning load prediction and flexible control according to claim 5, characterized in that, The process of obtaining the optimal solution set includes: Elite strategies and optimal individual strategies are determined by constraint-objective hierarchical ranking and crowding ranking: Constraint sub-level classification: Divided into 4 levels according to the number of constraints satisfied, with level 4 being the highest: satisfying all 4 constraints; level 3: satisfying 3 constraints; level 2: satisfying 2 constraints; and level 1 being the lowest: satisfying 1 or fewer constraints. Target sub-level classification: Within the same constraint sub-level, based on the multi-objective non-dominant relationship, for any two individuals a and b, if the prediction error of a is ≤ b, the regulation cost is ≤ b, and the comfort loss is ≤ b, and at least one of them is strictly less than b, then a dominates b; non-dominant individuals are classified into the same target sub-level. Congestion calculation formula: Where p is the number of individuals within the target sub-level; To predict the target accuracy value, i.e. ; To regulate economic target values; is the target value for comfort; i is the sorting index of the individual within the target sub-level; coefficient 2 is used to standardize the range of crowding values, making the crowding levels of different target sub-levels comparable. Individuals are sorted from highest to lowest constraint sub-level, and then from highest to lowest crowding level within the same constraint sub-level. The top 20% of individuals are selected as the elite strategy pool for population updates in the next iteration.

7. The method for multi-scenario air conditioning load prediction and flexible control according to claim 6, characterized in that, Based on the optimal solution set, the system outputs optimal prediction results for air conditioning load under multiple scenarios, a flexible control strategy including adjustable power scheduling values ​​and control duration, and corresponding electricity price incentive strategies and suggestions for adjusting user electricity consumption behavior, including: The specific process and convergence criteria for iterative solution are as follows: (1) Iterative process: Each iteration executes "population fitness assessment → non-dominant ranking → elite strategy selection → subpopulation division → population evolution update" in sequence. The higher the score, the better the fitness. (2) Convergence criterion: If the change in the fitness score of the best individual is ≤1×10^{-3} and the crowding variance of the Pareto optimal solution set is ≤5×10^{-2} in 50 consecutive iterations, then it is considered converged; (3) Strategy output: After convergence, select the strategy with the highest comprehensive score from the Pareto optimal solution set, and output the load forecast value, adjustable power scheduling value, regulation execution time, and corresponding electricity price incentive strategy and user electricity behavior adjustment suggestions for each scenario.

8. A multi-scenario air conditioning load prediction and flexible control system, characterized in that, include: The modeling module is used to establish an operating model of a single air conditioner based on an equivalent thermal parameter model. It divides multiple scenario sets according to date type, weather conditions and user comfort preferences, and calculates the adjustable potential of the air conditioner within the comfort range corresponding to each scenario, forming a basic model of adjustable potential. The optimization model building module is used to establish a multi-objective optimization problem based on the adjustable potential basic model, with load forecast value, adjustable power scheduling value and control duration as optimization variables, and user comfort, equipment operation, scenario adaptation and system power balance as constraints. The goal is to achieve the highest load forecast accuracy, the best control economy and the minimum loss of user comfort. The solution module is used to solve the multi-objective optimization problem using an improved snow ablation algorithm to obtain the optimal solution set; The execution module is used to output the optimal prediction results of air conditioning load under multiple scenarios, a flexible control strategy including adjustable power scheduling values ​​and control duration, and corresponding electricity price incentive strategies and suggestions for adjusting user electricity consumption behavior based on the optimal solution set.

9. A computer device 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 implements the steps of the multi-scenario air conditioning load prediction and flexible control method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the multi-scenario air conditioning load prediction and flexible control method as described in any one of claims 1 to 7.

Images