Energy scheduling method for service area integrated energy system in off-network state

Abstract

The invention relates to the field of energy scheduling, in particular to an off-network state-oriented service area integrated energy system energy scheduling method, which comprises the following steps of: acquiring load operation data of a service area in an off-network state, and establishing a load model of a road service area integrated energy system; obtaining a new energy supply type of the service area, and constructing various energy supply equipment models of the service area in an off-grid state so as to establish an energy supply model of the road service area integrated energy system; constructing a single-target optimization scheduling model; and performing a whale optimization algorithm on the single-target optimization scheduling model, and performing optimization according to a penalty term to obtain an energy scheduling scheme. According to the service area integrated energy system energy scheduling method oriented to the off-network state, power supply guarantee of the key load is maximized, normal operation of equipment with high priority level in the off-network state of the road service area is guaranteed, and the overall energy utilization efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of energy dispatching, and in particular to an energy dispatching method for integrated energy systems in service areas under off-grid conditions. Background Technology

[0002] In modern highway service areas, the energy supply and demand structure is becoming increasingly complex with the integration of functions such as electric vehicles, smart lighting, and information services. In remote areas or off-grid conditions caused by severe weather leading to disconnection from the main grid, service stations must rely on local energy systems to ensure continuous power supply to critical loads. Traditional energy allocation methods based on experience or fixed strategies cannot guarantee the stable operation of high-priority loads in the context of limited energy resources and variable loads. Summary of the Invention

[0003] To address the aforementioned issues, this invention provides an energy dispatching method for integrated energy systems in service areas under off-grid conditions. This method maximizes the power supply guarantee for critical loads, ensures the normal operation of high-priority equipment in off-grid conditions of highway service areas, and improves overall energy utilization efficiency.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0005] An energy dispatching method for a service area integrated energy system in an off-grid state includes the following steps:

[0006] S1. Collect load operation data of the service area in the off-grid state, classify the load operation data according to the importance level, and establish a load model of the integrated energy system of the highway service area;

[0007] S2. Obtain the types of new energy supply in the service area and construct models of various energy supply equipment in the service area under off-grid conditions to establish an energy supply model of the integrated energy system of the road service area and provide status information of multiple energy sources;

[0008] S3. Taking the comprehensive benefits of the load guarantee and energy utilization rate of the energy system as the objective function, the energy supply situation of the energy system load is selected as the decision variable, and according to the load model and the energy supply model, the importance level of various loads in the energy system and the power of energy supply allocation are selected as variables to construct a single-objective optimization scheduling model.

[0009] S4. The energy scheduling scheme is obtained by applying the whale optimization algorithm to the single-objective optimization scheduling model and optimizing according to the penalty term.

[0010] Further, in step S1, the power status of each load under the current state is obtained to collect load operation data. Operational features are extracted from the load operation data, including the current average operating power P of each load. i Daily operating time period T for each load i Daily start-up frequency f for each load i Key performance score for each load operation V i Based on the operational characteristics, a load feature vector x is constructed. i =[P i ,T i ,f i V i ];

[0011] A load model is constructed using the K-means clustering algorithm, with the number of clusters in the load model set to 3, so that the load operation data can be divided into high load level, medium load level and low load level according to importance. The load feature vector is processed by the load model to obtain the level label corresponding to each load, and the level label is used as the weight parameter.

[0012] Furthermore, the method for constructing the load model includes the following steps:

[0013] S1.1 Set up a dataset, which has multiple sample feature vectors, and normalize the sample feature vectors:

[0014]

[0015] Where, x i.k x is the k-th eigenvalue in the i-th eigenvector; k Let be the set of values ​​that the k-th feature can take in all sample feature vectors;

[0016] S1.2 Set the number of cluster centers K=3 to represent high load level, medium load level and low load level, and initialize the cluster centers using the K-means++ strategy;

[0017] S1.3 Perform the K-means iterative clustering process until the cluster centers converge;

[0018] S1.4 Sort the three clustering results by load power, frequency, and importance mean, and map them to level labels L. d ∈{1,2,3} represents high load level, medium load level, and low load level, in order to construct and obtain the load model;

[0019] S1.5 assigns an importance level L to each load. d As the output of the load model, the weighting parameter is

[0020] Furthermore, in step S1.3, the iterative process is as follows:

[0021] S1.3.1 For each sample x in the dataset i 'Calculate its distance to all current cluster centers μ' K Euclidean distance:

[0022]

[0023] Where, d iK x is the Euclidean distance; i ' represents the sample vector; μ K x′ represents the cluster center; d represents the feature dimension; x′ i,m μ is the m-th element of the sample vector. K,m It is the m-th element of the current cluster center;

[0024] S1.3.2 Assign each sample to the cluster S corresponding to the nearest cluster center. K S K The cluster number K takes the following values:

[0025] K = C i Formula (1.3)

[0026]

[0027] For each cluster S K Recalculate cluster center μ K To achieve optimal updates in the sense of minimizing the variance between sample points within a cluster, for each cluster S K Calculate cluster center μ K The method is as follows:

[0028]

[0029] S1.3.3 Repeat steps S1.3.1 to S1.3.2 until any of the following conditions are met: all cluster centers μ K The maximum number of iterations is reached when the change between two consecutive rounds is less than a certain threshold ε, the class assignment of each sample no longer changes, and the maximum number of iterations is reached.

[0030] Furthermore, in step S2, the constructed source supply models include a photovoltaic power generation system model, a wind power generation model, an energy storage model, and a diesel engine power generation model, and corresponding numbers and constraints are set for each source supply model.

[0031] Furthermore, in step S3, the single-objective optimization scheduling model is as follows:

[0032]

[0033] Where λ1 is the load guarantee weighting coefficient, representing the priority given to power supply to important loads; N is the total number of independent electrical loads requiring power supply in the service station, and each independent electrical load is assigned a number i = 1, 2, 3... N; w i For load protection weighting coefficient; y b To indicate whether the load is supplied with electricity, and y b ∈{0,1};w i y b To indicate the importance of the load; λ2 is the energy consumption penalty coefficient, representing the tendency to suppress total energy consumption; M represents the total number of different source supply model types available for power supply, with each type of source supply model assigned a corresponding number j = 1, 2, 3...M; p ij For source supply model type j, power allocation conforms to electrical load i, and p ij It is defined by the source supply model and the corresponding constraints.

[0034] Further, in step S4, the step of applying the whale optimization algorithm to the single-objective optimization scheduling model includes:

[0035] S4.1 Initialize parameters and population settings, and randomly generate the position of each individual when initializing the solution space, so that the power allocation constraint of source supply model type j conforming to electrical load i is p. ij ∈[0,C j max ], which is the maximum output power of the device corresponding to the source supply model of type j;

[0036] S4.2 A penalty term is introduced into the single-objective optimization scheduling model to construct the fitness function;

[0037] S4.3 Traverse the entire population to calculate the fitness value F of each individual in the fitness function, set the individual with the smallest fitness as the current best individual X*, and record its fitness value F*.

[0038] S4.4 Perform the corresponding position update method for each individual according to the preset probability p;

[0039] S4.5 corrects for individuals that violate power constraints;

[0040] S4.6 compares the current individual's new fitness value with the recorded historical best value and updates the current best individual accordingly;

[0041] S4.7 Repeat steps S4.3 to S4.5 until the maximum number of iterations is reached or the error convergence condition is met. Then, output the position of the current global best individual. The position of the global best individual is the optimal scheduling strategy.

[0042] Furthermore, in step S4.1, the core parameters of the whale optimization algorithm are set and the variable encoding format is configured. The core parameters include the initial spiral factor a and the maximum number of iterations T. max Population size n;

[0043] The variable encoding format is: [y1,…,y N ,p 11 ,…,p NM ]

[0044] Among them, y b ∈{0,1} indicates whether the load is supplied with electricity; p ij Power allocation for source supply model type j to electrical load i;

[0045] In step S4.2, the fitness function is:

[0046] F = -f + ρPenalty formula (4.1) where f is the single-objective optimization scheduling model; ρ is the penalty coefficient, used to adjust the penalty intensity; the penalty term includes supply and demand imbalance: Energy exceeded limit: Output exceeds maximum power:

[0047] Further, in step S4.4,

[0048] If p < 0.5, execute the prey encirclement behavior position update method, and perform shrinking encirclement update:

[0049]

[0050] in, r1, r2 ∈ [0, 1], a decreases linearly with each iteration;

[0051] (2) If p≥0.5, execute the position update method of spiral behavior and move around in a spiral:

[0052]

[0053] in, b is a constant, l∈[-1,1];

[0054] In step S4.5, if p ij If it exceeds the maximum value, then it will be cropped. If no corresponding energy is provided, y b Forced to 0; if energy is overdrawn, p is redistributed. ij To reduce the load;

[0055] In step S4.6, the spiral coefficient is updated simultaneously when updating the current optimal individual. T represents the current iteration number.

[0056] Furthermore, in step S4.7, the position of the current globally optimal individual is output as X*. z , corresponding to the load on / off control vector y b and energy power allocation matrix p ij .

[0057] The beneficial effects of this invention are as follows: Throughout the optimization process, this invention collects and analyzes operational data from various loads in the service area, extracts operational characteristics, and uses the K-means clustering algorithm to classify loads according to their importance. By inputting optimized scheduling weight coefficients based on load importance, and using the comprehensive benefit of load guarantee and energy utilization as the objective function, this invention fully considers prioritizing the energy supply of important loads under off-grid conditions, thus obtaining the energy scheduling scheme with the best comprehensive benefit. This invention uses the whale optimization algorithm to solve the optimized scheduling model, which has the advantages of strong convergence stability, adaptability to single-objective scheduling models, and suitability for problems involving multiple resource constraints. It is applicable to the energy scheduling of integrated energy systems in highway service areas under off-grid conditions with limited energy supply. When the optimal comprehensive benefit scheduling strategy is obtained, the strategy provides information on wind and solar power output and energy storage charging and discharging. Relevant departments can use this scheme to adjust energy supply, achieving the goal of ensuring the stable operation of important loads in the service area and guaranteeing the normal operation of high-priority equipment. This invention is highly adaptable to integrated energy systems in highway service areas under off-grid conditions with limited energy supply. Attached Figure Description

[0058] Figure 1 This is a flowchart of a preferred embodiment of the present invention.

[0059] Figure 2 This is a flowchart of a clustering algorithm according to a preferred embodiment of the present invention.

[0060] Figure 3 This is a flowchart of a preferred embodiment of the whale optimization algorithm of the present invention. Detailed Implementation

[0061] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0062] It should be noted that when a component is described as "fixed to" another component, it can be directly on the other component or may have a component in between. When a component is considered "connected to" another component, it can be directly connected to the other component or may have a component in between. When a component is considered "set on" another component, it can be directly set on the other component or may have a component in between. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0063] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0064] Please also see Figures 1 to 3 A preferred embodiment of the energy dispatching method for a service area integrated energy system under off-grid conditions according to the present invention includes the following steps:

[0065] S1. Collect load operation data of the service area in the off-grid state, classify the load operation data according to the importance level, and establish a load model of the integrated energy system of the highway service area.

[0066] In step S1, the power status of each load under the current state is obtained to collect load operation data. Operational characteristics are extracted from the load operation data, including the current average operating power P of each load. i Daily operating time period T for each load i Daily start-up frequency f for each load i Key performance score for each load operation V i Construct a load feature vector x based on operational characteristics. i =[P i ,T i ,f i V i ].

[0067] The load of a service area generally includes basic living support load, transportation service load, energy system operation load, and commercial operation load. Basic living support load is necessary to ensure the basic survival and safety needs of people. This includes lighting load (including indoor lighting, road lighting, and emergency lighting), water supply and drainage system load (including water supply systems, sewage discharge and treatment systems), ventilation and air conditioning system load to maintain smooth indoor ventilation and suitable temperature and humidity, and security and communication system load (including communication base stations and security monitoring systems). Transportation service load refers to the load of equipment related to traffic operation in highway service areas, ensuring the smooth flow of vehicles and people. This type of equipment includes traffic lights, electronic road signs, and other traffic information guidance equipment, electric vehicle charging equipment, and refueling service equipment. Energy system operation load includes energy storage system operation load, energy conversion equipment load such as inverters and heat exchangers, and control system load. Commercial operation load refers to all loads related to commercial activities within the service area.

[0068] In off-grid conditions, the energy supply capacity of the service area is limited, making it difficult to meet the simultaneous operation of all loads. By classifying the loads according to their importance and prioritizing them within the service area, the operation of important loads can be prioritized in off-grid conditions, maintaining the function of critical equipment and achieving optimal energy supply allocation.

[0069] In this embodiment, the K-means clustering algorithm is used to construct the load model. The number of clusters in the load model is set to 3 to divide the load operation data into high load level, medium load level and low load level according to their importance. The load feature vector is processed by the load model to obtain the level label corresponding to each load, and the level label is used as the weight parameter.

[0070] The method for constructing a load model includes the following steps:

[0071] S1.1 Set up the dataset, which has multiple sample feature vectors, and normalize the sample feature vectors:

[0072]

[0073] Where, x i.k x is the k-th eigenvalue in the i-th eigenvector; k Let be the set of values ​​that the k-th feature can take in all sample feature vectors.

[0074] S1.2 Set the number of cluster centers K=3 to represent high load level, medium load level and low load level, and initialize the cluster centers through the K-means++ strategy.

[0075] S1.3 Perform the K-means iterative clustering process until the cluster centers converge;

[0076] S1.3.1 For each sample x in the dataset i 'Calculate its distance to all current cluster centers μ' K Euclidean distance:

[0077]

[0078] Where, d iK x is the Euclidean distance; i ' represents the sample vector; μ K x′ represents the cluster center; d represents the feature dimension; x′ i,m μ is the m-th element of the sample vector. K,m It is the m-th element of the current cluster center;

[0079] S1.3.2 Assign each sample to the cluster S corresponding to the nearest cluster center. K S K The cluster number K takes the following values:

[0080] K = C i Formula (1.3)

[0081]

[0082] For each cluster S K Recalculate cluster center μ K To achieve optimal updates in the sense of minimizing the variance between sample points within a cluster, for each cluster S K Calculate cluster center μ K The method is as follows:

[0083]

[0084] S1.3.3 Repeat steps S1.3.1 to S1.3.2 until any of the following conditions are met: all cluster centers μ K The maximum number of iterations is reached when the change between two consecutive rounds is less than a certain threshold ε, the class assignment of each sample no longer changes, and the maximum number of iterations is reached.

[0085] S1.4 Sort the three clustering results by load power, frequency, and importance mean, and map them to level labels L. d ∈{1,2,3} represents high load level, medium load level, and low load level, in order to construct a load model.

[0086] S1.5 assigns an importance level L to each load. d As the output of the load model, the weighting parameters are:

[0087] In this embodiment, the K-means algorithm is used to design a clustering model. The standard K-means algorithm is employed, with the number of clusters set to K=3, representing that load importance is divided into three categories: high load level L... d =1, such as communication equipment, security monitoring, and fire protection systems, where L is the load level. d =2, such as medium-sized important equipment like store refrigeration equipment and air conditioning systems, with a low load level (L). d =3, such as non-critical loads like billboards and entertainment terminals.

[0088] The load feature vectors are standardized and then input into the clustering model; the clustering model results are output, and the model output L is... d As the level label corresponding to each load, and L d As weight parameters in the optimization model, they are used to schedule optimization schemes.

[0089] S2. Obtain the types of new energy supply in the service area and construct models of various energy supply equipment in the service area under off-grid conditions to establish an energy supply model for the integrated energy system of the road service area and provide status information of multiple energy sources.

[0090] In step S2, the constructed source supply models include a photovoltaic power generation system model, a wind power generation model, an energy storage model, and a diesel engine power generation model, and corresponding numbers and constraints are set for each source supply model.

[0091] In this embodiment, the energy supply side of the highway service area includes municipal power grid, photovoltaic power generation equipment, wind power generation equipment, energy storage equipment, and other equipment.

[0092] Photovoltaic power generation equipment uses solar panels to directly convert solar energy into electrical energy based on the photovoltaic effect.

[0093] Photovoltaic power generation system model:

[0094] P PV,t P represents the actual output power of the photovoltaic cell. PVstc L represents the output power of the photovoltaic cell under standard test conditions, and L represents the actual solar irradiance. stc Indicates solar irradiance under standard conditions; T, T stc These represent the ambient temperatures under actual and standard conditions, respectively, and γ represents the temperature degradation coefficient of the photovoltaic cell.

[0095] Wind power generation equipment primarily relies on wind power to drive the blades, converting mechanical energy into electrical energy. Its power output is directly proportional to the cube of the wind speed. Wind power generation model:

[0096] Pwind The output power of the wind power generation is represented by: ρ; the air density of the corresponding region is represented by: K; the performance parameters of the wind power generation equipment are represented by: S; the area swept by the wind turbine blades is represented by: v;

[0097] Energy storage devices are devices that convert electricity or other forms of energy into a storable form for storage and release it for use when needed. In the field of new energy applications, energy storage devices can also enable centralized scheduling and management of distributed clean energy sources such as solar photovoltaic and wind power. The energy storage capacity of an energy storage device is related to its charging and discharging power and efficiency.

[0098] Electric energy storage model: E S (t)=(1-λ S E S (t-1)+Δt[η charge P charge (t)-η discharge P discharge ];

[0099] E S (t) represents the energy storage capacity at time t, η charge P represents the charging efficiency of an energy storage device. charge P represents the charging power of an energy storage device. discharge η represents the discharge power of the energy storage device. discharge λ represents the discharge efficiency of an energy storage device. S This represents the self-loss coefficient of the battery;

[0100] In addition, the constraints for the energy storage model are set as follows:

[0101] 0≤P discharge ≤ε discharge E max

[0102] 0≤P charge ≤ε charge E max

[0103] β min E max ≤E S (t)≤β max E max

[0104] Where, ε charge ε represents the maximum charging efficiency of the energy storage device. discharge β represents the maximum discharge efficiency of the energy storage device. min β represents the minimum energy storage coefficient. max Indicates the maximum energy storage coefficient

[0105] Diesel engines offer advantages such as rapid start-up, reliable operation, and flexible adjustment. In energy system applications, especially in off-grid or unstable energy supply scenarios, diesel engines can serve as main power, backup power, or supplementary power, ensuring stable system operation. This helps improve power supply stability and system stability, thereby building a safe and controllable energy system.

[0106] Diesel engine generator model: P dg (t)=a dg V dg (t);

[0107] Among them, P dg (t) represents the output power of the diesel engine at time t, a dg V is the fuel coefficient of a diesel engine. dg (t) represents the amount of fuel consumed by the diesel engine at time t.

[0108] The diesel engine output constraints are set as follows:

[0109] P dg,min ≤P dg (t)≤P dg,max ;

[0110] Among them, P dg,min and P dg,max These are the minimum and maximum output of the diesel engine.

[0111] S3. Taking the comprehensive benefits of the load guarantee and energy utilization rate of the energy system as the objective function, the energy supply situation of the energy system load is selected as the decision variable. Based on the load model and energy supply model, the importance level of various loads in the energy system and the power of energy distribution are selected as variables to construct a single-objective optimization scheduling model.

[0112] In step S3, the single-objective optimization scheduling model is as follows:

[0113]

[0114] Where λ1 is the load guarantee weighting coefficient, representing the priority given to power supply to important loads; N is the total number of independent electrical loads requiring power supply in the service station, with each independent electrical load assigned a number i = 1, 2, 3...N; N represents the total number of independent electrical loads requiring power supply in the service station, such as lighting systems, air conditioners, cameras, advertising screens, etc.; w i For load protection weighting coefficient; y b To indicate whether the load is supplied with electricity, and y b ∈{0,1};w i y bTo indicate the importance of the load; λ2 is the energy consumption penalty coefficient, representing the tendency to suppress total energy consumption; M represents the total number of different source supply model types available for power supply, with each type of source supply model assigned a corresponding number j = 1, 2, 3...M; M represents photovoltaic, energy storage battery, wind power, diesel power generation, etc.; p ij For source supply model type j, power allocation conforms to electrical load i, and p ij It is defined by the source supply model and the corresponding constraints.

[0115] This embodiment considers the load guarantee and energy utilization rate of the integrated energy system, takes the combined benefit of the two as the objective function, uses the energy supply status of the loads in the system as the decision variable, and uses the importance level of various loads in the system and the power of energy allocation as variables, thereby constructing a single-objective optimization scheduling model.

[0116] S4. The energy scheduling scheme is obtained by applying the whale optimization algorithm to the single-objective optimization scheduling model and finding the optimal solution based on the penalty term.

[0117] In step S4, the steps of applying the whale optimization algorithm to the single-objective optimization scheduling model include:

[0118] S4.1 Initialize parameters and population settings, and randomly generate the position of each individual when initializing the solution space, so that the power allocation constraint of source supply model type j conforming to electrical load i is p. ij ∈[0,C j max ], which is the maximum output power of the device corresponding to the source supply model of type j;

[0119] In step S4.1, the core parameters of the whale optimization algorithm are set and the variable encoding format is configured. The core parameters include the initial spiral factor 'a' and the maximum number of iterations 'T'. max Population size n;

[0120] The variable encoding format is: [y1,…,y N ,p 11 ,…,p NM ]

[0121] Among them, y b ∈{0,1} indicates whether the load is supplied with electricity; p ij The power allocation for source supply model type j to electrical load i.

[0122] S4.2 Introduces a penalty term into the single-objective optimization scheduling model to construct the fitness function;

[0123] In step S4.2, the fitness function is:

[0124] F = -f + ρPenalty formula (4.1) where f is the single-objective optimization scheduling model; ρ is the penalty coefficient, used to adjust the penalty intensity; the penalty term includes supply and demand imbalance: Energy exceeded limit: Output exceeds maximum power:

[0125] S4.3 Traverse the entire population to calculate the fitness value F of each individual in the fitness function, set the individual with the smallest fitness as the current best individual X*, and record its fitness value F*.

[0126] S4.4 Perform the corresponding position update method for each individual according to the preset probability p;

[0127] If p < 0.5, execute the prey encirclement behavior position update method, and perform shrinking encirclement update:

[0128]

[0129] in, r1, r2 ∈ [0, 1], a decreases linearly with each iteration;

[0130] (2) If p≥0.5, execute the position update method of spiral behavior and move around in a spiral:

[0131]

[0132] in, b is a constant, l∈[-1,1].

[0133] S4.5 corrects for individuals that violate power constraints;

[0134] In step S4.5, if p ij If it exceeds the maximum value, then it will be cropped. If no corresponding energy is provided, y b Forced to 0; if energy is overdrawn, p is redistributed. ij To reduce the load.

[0135] S4.6 compares the current individual's new fitness value with the recorded historical best value and updates the current best individual accordingly;

[0136] S4.7 Repeat steps S4.3 to S4.5 until the maximum number of iterations is reached or the error convergence condition is met. Then, output the position of the current global best individual. The position of the global best individual is the optimal scheduling strategy.

[0137] In step S4.7, the position of the current globally optimal individual is output as X*. z , corresponding to the load on / off control vector yb and energy power allocation matrix p ij .

Claims

1. A method for energy scheduling of a service area integrated energy system in an off-grid state, characterized in that, Comprise the following steps: S1. Collect the load operation data of the service area in the off-grid state, grade divide each load operation data according to importance, and establish the load model of the comprehensive energy system of the highway service area; S2. Obtain the new energy supply type of the service area, and construct various energy supply equipment models of the service area in the off-grid state to establish the energy supply model of the comprehensive energy system of the highway service area, and provide the state information of the multi-source energy; S3. Take the comprehensive benefits of load guarantee degree and energy utilization rate of the energy system as the objective function, select the energy supply situation of the energy system load as the decision variable, and select the importance degree level and energy supply distribution power of each type of load in the energy system as the variable according to the load model and the energy supply model, to construct a single-objective optimization scheduling model; S4. Obtain the energy scheduling scheme by whale optimization algorithm on the single-objective optimization scheduling model, and according to the penalty term. 2.The energy scheduling method for a service area comprehensive energy system in an off-grid state according to claim 1, wherein: In step S1, the power status of each load under the current state is obtained to collect load operation data. Operational features are extracted from this load operation data, including the current average operating power P of each load. i Daily operating time period T for each load i Daily start-up frequency f for each load i Key performance score for each load operation V i Based on the operational characteristics, a load feature vector x is constructed. i =[P i ,T i ,f i V i ]; The K-means clustering algorithm is used to construct the load model, and the number of clustering categories of the load model is set to 3, so that the load operation data is divided into high load level, medium load level and low load level according to importance. The load characteristic vector is processed by the load model to obtain the corresponding grade label of each load, and the grade label is used as a weight parameter.

3. The method of claim 2, wherein: The construction method of the load model comprises the following steps: S1.1 Set up a data set, the data set has a plurality of sample feature vectors, and the sample feature vectors are normalized: wherein x i.k is the kth eigenvalue in the ith eigenvector; x k is the set of values of the kth eigenvalue in all sample eigenvectors; S1.2 Set the number of cluster centers K=3 to represent high load level, medium load level and low load level, and initialize the cluster centers by K-means++ strategy; S1.3 Perform K-means iterative clustering process until the cluster center converges; S1.4 Sort the three types of clustering results according to the mean values of load power, frequency and importance, and map them to hierarchical labels L d ∈{1,2,3}, representing high load level, medium load level and low load level, to construct a load model; S1.5 rank the importance level L of each load d As an output of the load model, the weight parameter is 4. The energy scheduling method for a service area integrated energy system in an off-grid state according to claim 3, characterized in that: In step S1.3, the iteration process is: S1.3.1 Compute the Euclidean distance of each sample x in the dataset to all current cluster centers μ i S1.3.2 Select the cluster center μ with the smallest distance to x K S1.3.3 Assign x to the selected cluster. where d iK is the Euclidean distance; x′ i is the sample vector; μ K is the cluster center; d is the feature dimension; x i ' ,m is the mth element of the sample vector; μ K,m is the mth element of the current cluster center; S1.3.2 Assign each sample to the cluster corresponding to the nearest cluster center S K , S K The cluster number K takes the value: K = C i Equation (1.3) For each cluster S K Recalculate cluster centers μ K For each cluster S K Calculate cluster centers μ K The method is as follows: S1.3.3 Repeat steps S1.3.1 to S1.3.2 until any of the following conditions is met: all cluster centers μ K less than a certain threshold ε between consecutive two rounds, the class assignment of each sample no longer changes, a maximum number of iterations is reached.

5. The method of claim 1, wherein: In step S2, the constructed source supply model includes photovoltaic power generation system model, wind power generation model, electric energy storage model, diesel engine power generation model, and each source supply model is set with corresponding number and constraint condition.

6. The energy scheduling method for a service area integrated energy system in an off-grid state according to claim 5, characterized in that: In step S3, the single-objective optimization scheduling model is: wherein λ1 is a load guarantee weight coefficient, representing the importance of the power supply priority of the important load; N is the total number of independent electrical loads needing power supply in the service station, each independent electrical load is provided with a number i=1, 2, 3...N; w i is a load guarantee weight coefficient; y b is used to represent whether the load is powered, and y b ∈{0, 1}; w i y b is used to represent the importance of the load; λ2 is an energy consumption penalty coefficient, representing the inhibitory tendency to the total energy consumption; M is used to represent the total number of different source supply model types available for power supply, each type of the source supply model is provided with a corresponding number j=1, 2, 3...M; p ij is the power allocation of the source supply model type j to the electrical load i, and p ij is defined by the source supply model and the corresponding constraint condition.

7. The energy scheduling method for a service area integrated energy system in an off-grid state according to claim 6, characterized in that: In step S4, the steps of whale optimization algorithm on the single-objective optimization scheduling model include: S4.1 Initialize parameters and population settings, and initialize the solution space by randomly generating the position of each individual, such that the source supply model type j meets the power allocation constraints for electrical load i as p ij ∈ [0, C j max ], which is the maximum output power of the device corresponding to the source supply model type j; S4.2 Introduce a penalty term into the single-objective optimization scheduling model to construct a fitness function; S4.3 Traverse the entire population to calculate the fitness value F of each individual in the fitness function, set the individual with the minimum fitness as the current optimal individual X*, and record its fitness value F*; S4.4 According to the preset probability p, each individual executes the corresponding position updating mode; S4.5 Correct the individual that violates the power constraint; S4.6 Compare the new fitness value of the current individual with the recorded historical optimal value, and update the current optimal individual; S4.7 Repeat steps S4.3 to S4.5 until the maximum iteration number is reached or the error convergence condition is met, output the position of the current global optimal individual, and the position of the global optimal individual is the optimal scheduling strategy. 8.The energy scheduling method for a service area comprehensive energy system in an off-grid state according to claim 7, wherein: In step S4.1, the core parameters of whale optimization algorithm are set, including initial spiral factor a, maximum iteration number T, population size n, and variable coding format is set. max , a population size n; The variable encoding format is: [y1,..., y N ,p 11 ,…,p NM ] where y b ∈ {0,1}, indicates whether the load is powered or not; p ij is the power allocation of source supply model type j to electrical load i. In step S4.2, the fitness function is: F = -f + pPenalty Equation (4.1) Wherein, f is a single-target optimization scheduling model; p is a penalty coefficient, used to adjust the punishment degree; the penalty term includes supply and demand imbalance: Energy exceeds the limit: Output exceeds the maximum power:

9. The energy scheduling method for a service area integrated energy system in an off-grid state according to claim 7, characterized in that: In step S4.4, If p < 0.5, the position updating mode of surround hunting behavior is executed, and the shrink surround updating is performed: wherein, a decreases linearly step by step with iteration; (2) If p ≥ 0.5, the position updating mode of spiral behavior is executed, and the spiral surrounding movement is performed: wherein, b is a constant, l e [-1, 1]; In step S4.5, if p ij is over the maximum, it is clipped to If no corresponding energy is provided, y b is forced to 0; if the energy is over budget, p ij is redistributed to lower the load; In step S4.6, when updating the current optimal individual, the helix coefficient is also updated T is the current iteration number.

10. The method of claim 7, wherein the method is a method of energy dispatching for an integrated energy system of a service area in an off-grid state, characterized in that: In step S4.7, the position of the current global optimum individual is output as X z corresponding to the load switching control vector y b and the energy power distribution matrix p ij .

Images