Regional comprehensive energy operation optimization method considering flexible adjustment of multi-energy network

Abstract

The invention discloses a regional comprehensive energy operation optimization method considering flexible adjustment of a multi-energy network, and relates to the technical field of energy operation optimization, and the method comprises the following steps: S1, constructing an electric heating hybrid energy network cluster and a multi-DIES cluster; s2, constructing a regional power distribution network model and a regional heat supply network model; s3, constructing an electric heating hybrid energy network cluster division model according to the regional power distribution network model and the regional heat supply network model; s4, constructing a multi-DIES cluster double-layer collaborative operation optimization model; and S5, solving the electric heating hybrid energy network cluster division model and the multi-DIES cluster double-layer collaborative operation optimization model, and completing optimization. According to the two-stage optimization method provided by the invention, the power demand of each DIES on the power distribution network can be reduced in the load peak period, the main transformer overload risk is reduced, and the economical efficiency of system operation is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of energy operation optimization, and in particular to a regional integrated energy operation optimization method considering flexible adjustment of a multi-energy network. Background Art

[0002] The existing integrated energy system is distributed at different nodes in the distribution network and the district heating network. There are also other electric and thermal load nodes in the distribution network and the district heating network. These nodes are supplied with heat and electricity through ESP. When the load in the distribution network is large, the electric power flowing into each substation through the main transformer of the distribution network will surge, causing the main transformer to be overloaded. The overload of the main transformer will produce a series of adverse consequences, mainly including the following aspects: 1. Increased heat loss and temperature rise of the transformer, specifically (1) Increased heat loss: Overload operation will significantly increase the heat loss of the transformer. This is because when the transformer is working, it will generate heat due to resistance, core loss and other reasons. When the load exceeds the rated value, these losses will increase sharply, causing the transformer temperature to rise; (2) Temperature rise: Long-term overload operation will cause the internal temperature of the transformer to continue to rise, exceeding the maximum temperature allowed by its design. This will not only affect the insulation performance of the transformer, but also accelerate the aging of the insulation material and shorten the service life of the transformer. 2. Voltage fluctuation and power quality degradation, specifically (1) Voltage drop: Overload operation will cause the transformer output voltage to drop, because when the transformer is overloaded, its internal resistance and leakage reactance will increase, thus affecting the stable output voltage. Voltage fluctuations will affect the normal operation of power equipment and even cause equipment failure; (2) Power quality degradation: Overload operation may also cause problems such as increased harmonics and three-phase imbalance, further reducing power quality. 3. Increased safety hazards, specifically (1) Equipment damage risk: Long-term overload operation will cause excessive stress on the internal components of the transformer, which may cause winding deformation, insulation breakdown and other faults, and in severe cases, cause the transformer to burn out; (2) Fire and explosion risk: High temperature and insulation aging may cause serious accidents such as fire or explosion, posing a threat to personal and property safety. 4. Affected grid stability, specifically the risk of local grid collapse: If the main transformer is severely overloaded and lasts for a long time, it may affect the stability of the entire grid. In extreme cases, it may cause local grid collapse, affecting power supply over a large area.

[0003] In response to the overload problem of the main transformer in the distribution network, the existing response measures include the following aspects: (1) Strengthen equipment monitoring and maintenance: Regularly monitor and maintain key equipment such as transformers to promptly detect and deal with potential problems. At the same time, establish a complete emergency plan and fault handling mechanism to deal with emergencies. (2) Improve equipment performance: Select transformer equipment with excellent performance and high efficiency to reduce heat loss and improve power quality. At the same time, strengthen the heat dissipation design and maintenance management of the equipment to reduce the equipment temperature. (3) Strengthen load forecasting and management: Through scientific load forecasting and management methods, reasonably arrange the load distribution of the power grid to avoid a single transformer bearing excessive load. (4) Optimize the power grid structure: Rationally optimize the power grid structure to improve the flexibility and reliability of the power grid. For example, by adding transformers, adjusting the power grid layout, etc., the load pressure of the main transformer can be shared.

[0004] However, conventional transformer overload interlocking devices have two major problems in actual operation: 1) they cannot automatically adapt to changes in substation operating modes, which may lead to the incorrect disconnection of some load lines; 2) due to the overly simple selection strategy for the load lines to be disconnected, unnecessary over-cutting is often caused. Summary of the Invention

[0005] In order to solve the above problems, the present invention proposes a regional integrated energy operation optimization method considering the flexible adjustment of multi-energy networks.

[0006] The technical solution of the present invention is: a regional integrated energy operation optimization method considering the flexible adjustment of multi-energy networks comprises the following steps: S1. Build electric-thermal hybrid energy network cluster and multi-DIES cluster; S2. Construct regional distribution network model and regional heating network model; S3. Based on the regional distribution network model and the regional thermal network model, a cluster partitioning model of the electric-thermal hybrid energy network is constructed; S4. Construct a two-tier collaborative operation optimization model for multiple DIES clusters; S5. Solve the electric-thermal hybrid energy network cluster partition model and the multi-DIES cluster two-layer collaborative operation optimization model to complete the optimization.

[0007] Furthermore, in S2, the expression of the regional distribution network model is: ; Where, F ij Represented by the node i With node j The power flow caused by the change of power transmission between f 1i Represents a slave node iThe power flow change on branch 1 caused by the power flowing into and out of the reference node, f 1j Representation node j The power flow change on branch 1 caused by the power flowing into and out of the reference node, f 2i Representation node i The power flow change on branch 2 caused by the power flowing into and out of the reference node, f 2j Representation node j The power flow change on branch 2 caused by the power flowing into and out of the reference node, f li Represents a slave node i The branch caused by the power flowing into and out of the reference node l The trend changes, f lj Represents a slave node j The branch caused by the power flowing into and out of the reference node l changes in trends; In S2, the regional heating network model includes the total heat loss of the heating pipeline and the head loss along the pipeline; Total heat loss of heating pipes E ij The expression is: ; Where, L ij Represents a slave node i To Node j The length of the pipe, β represents the additional coefficient of heat loss, q id Indicates the heat loss per unit length of pipe under standard working conditions; Head loss along the pipeline H ij The expression is: ; Where, Y Indicates the pipe roughness, r represents the density of the transmission medium, d ij Indicates the inner diameter of the pipe, G Indicates the transmission medium traffic.

[0008] Furthermore, S3 includes the following sub-steps: S31. Calculate the distribution network modularity index of the electric-thermal hybrid energy network cluster based on the regional distribution network model; S32. Calculate the thermal network modularity index of the electric-thermal hybrid energy network cluster based on the regional thermal network model; S33, calculating the power balance index of the distribution network of the electric-thermal hybrid energy network cluster; S34. Calculate the heat network power balance index of the electric-heat hybrid energy network cluster; S35. Construct an electric-thermal hybrid energy network cluster partitioning model based on the distribution network modularity index, the heat network modularity index, the distribution network power balance index, and the heat network power balance index.

[0009] Furthermore, in S31, the distribution network modularity index D E The expression is: ; Where, m represents the sum of the weights of all edges in the network, w i Represents all nodes i The sum of the weights of the connected edges, w j The table represents all nodes j The sum of the weights of the connected edges, F ij Represented by the node i With node j The power flow caused by the change of power transmission between d E ( i , j ) represents the distribution network node ownership determination factor; In S32, the thermal network modularity index D H The expression is: ; Where, H ij represents the head loss along the pipeline, E ij Represents the total heat loss of the heating pipe, d H ( i , j ) represents the regional heating network node ownership determination factor; In S33, the power balance index of the distribution network Q P The expression is: ; Where, f P,k Indicates the kThe power balance index of a DIES cluster, K Indicates the number of multi-DIES clusters; In S34, the heat network power balance index Q H The expression is: ; Where, f H,k Indicates the k Thermal power balance index of a DIES cluster; In S35, a reinforcement learning model is used to build a cluster partitioning model for the electric-thermal hybrid energy network; Action Space of Cluster Partitioning Model for Electric-Thermal Hybrid Energy Network Action The expression is: ; State space of cluster partitioning model for electric-thermal hybrid energy network State The expression is: ; Where, p k,m,t Represents a cluster k midpoint m The electrical load demand, h k,m,t,t+Δt Represents a cluster node m In time Thermal power demand, Represents a cluster k exist t The electrical power at the moment, Represents a cluster k exist t Thermal power at the moment, M E Represents a cluster k The number of distribution network nodes in M H Represents a cluster k The number of grid nodes in the district heating network; Reward function for cluster partitioning model of electric-thermal hybrid energy network Reward The expression is: ; Where, w p represents the weight coefficient of the two indicators in the distribution network, D E represents the modularity index of the distribution network, D H represents the thermal network modularity index, wH represents the weight coefficient of the two indicators in the regional heating network, and max(·) represents the maximum value function; Action-value function of cluster partitioning model for electric-thermal hybrid energy network Q t ( S t , A t ) is: ; Where, R t Indicates immediate reward, Q t ( S t+1 , A t+1 ) represents the action-value function for the next time period, S t express t The state of the moment, A t express t The action of the moment, S t+1 Indicates the status of the next time period. A t+1 Indicates the action in the next time period.

[0010] Furthermore, in S4, the multi-DIES cluster two-layer collaborative operation optimization model includes a lower-layer DIES cluster operation optimization model and an upper-layer DIES cluster operation optimization model.

[0011] Furthermore, the expression of the lower-level DIES cluster operation optimization model is: ; ; ; ; ; ; Where, C IES,k Indicates the operating cost of a multi-DIES cluster, C en,k represents the energy cost of a multi-DIES cluster, C om,k Indicates the maintenance cost of a multi-DIES cluster, C E,k represents the interaction cost between the multi-DIES cluster and the grid, CNG,k represents the fuel cost of a multi-DIES cluster, C H,k represents the thermal energy cost of a multi-DIES cluster, T Indicates the optimization time period, represents the price at which the DIES cluster sells electricity to the ESP, represents the electricity selling price of ESP, represents the market heat price, represents the thermal power purchased from the regional heating network by the multi-DIES cluster, max(·) represents the maximum value function, P grid,k,t represents the power transaction power between the DIES cluster and the ESP, P CHP,k,t Indicates the k Power of CHP units in a DIES cluster, P GB,k,t Indicates the k GB unit power in a DIES cluster, c gas represents the natural gas price, or CHP,k Indicates the k The operating efficiency of CHP units in a DIES cluster, LHV ng Indicates the lower calorific value of natural gas. or GB,k Indicates the k The operating efficiency of GB units in a DIES cluster, c w Indicates the k Units in a DIES cluster w Unit operating power operation and maintenance cost, P k,w,t Indicates the k Units in a DIES cluster w Operating power, W Indicates the k The number of units in a DIES cluster; The constraint expression of the lower-level DIES cluster operation optimization model is: ; ; ; ; ; Where, SOC ES,t+1 Indicates energy storage t +1 moment state of charge, SOC ES,t Indicates energy storage t The state of charge at the moment, α l represents the energy loss rate of energy storage self-discharge, Indicates the charging power of the energy storage, represents the energy storage charging efficiency, Indicates the rated charge and discharge power of energy storage, Δ t represents the optimization time interval, Indicates the rated capacity of energy storage, represents the discharge power of the energy storage, represents the energy storage discharge efficiency, M represents the energy storage charging and discharging state variable, Indicates the minimum state of charge for energy storage, Indicates the maximum state of charge of the energy storage, SOC ES,1 represents the expected state of charge of the energy storage initial state of charge, SOC ES,T Indicates the expected state of charge at the end of the energy storage dispatch cycle; The expression of the upper-level DIES cluster operation optimization model is: ; ; Where, C ENO represents the total operating cost, max(·) represents the maximum value function, P Tr,t Indicates the output power of the main transformer, It indicates the price of electricity purchased by the distribution network from the upper main grid. C over represents the overload cost of the main transformer, Indicates the rated power of the main transformer, c Tr Indicates the overload loss coefficient of the main transformer; The constraint expression of the lower-level DIES cluster operation optimization model is: ; Where, Indicates the k The power transaction power between the DIES cluster and the distribution network, P Cwt,t Indicates the power generated by the wind farm cluster in the distribution network.

[0012] Furthermore, in S5, the xLSTM algorithm is used to determine the optimal action in a given state and complete the solution of the cluster partitioning model of the electric-thermal hybrid energy network; In S5, a genetic algorithm is used to solve the upper-level DIES cluster operation optimization model; In S5, the Cplex commercial solver is used to solve the underlying DIES cluster operation optimization model.

[0013] The beneficial effects of the present invention are: (1) This paper designs a distribution network main transformer overload solution and a two-stage operation optimization framework that takes into account the flexible adjustment of the power distribution and regional heating network structures. Compared with the existing topology adjustment strategy, it not only involves the topology structure and equipment related to the distribution network, but also considers the regional heating network and related electric-thermal coupling equipment coupled with the distribution network. It has more adjustable resources and more effective adjustment. (2) This paper establishes a model that takes into account the flexibility of the power distribution and regional heating network structure. In view of the complex scenario in which multiple devices such as distributed photovoltaics, electric thermal energy storage, electric boilers, heat pumps and micro-turbines, as well as new loads are connected to the distribution network at this stage, a hybrid electric-thermal network structure flexibility model is established, which can provide a basic model support for the topology adjustment strategy of the electric-thermal hybrid energy network; (3) The present invention establishes a reinforcement learning model for cluster division of electric-thermal hybrid energy networks that takes into account the coupling relationship between structural characteristics and power characteristics. The existing technology establishes a distribution network cluster division model or a distribution network reconstruction model based on traditional data models. The present invention simultaneously considers the distribution network topology reconstruction, regional heating network topology optimization, and the coupling relationship between the distribution network and the regional heating network, and establishes a reinforcement learning model for cluster division of electric-thermal hybrid energy networks that takes into account the coupling relationship between structural characteristics and power characteristics. The cluster division model is established through the reinforcement learning model.

[0014] (4) The present invention establishes a two-layer collaborative operation optimization model that takes into account the coordinated operation of each distributed cluster and the electric and thermal multi-energy network. The existing technology does not consider incorporating the cost of main transformer overload into the optimization target of the upper-level energy service provider. The present invention establishes a two-layer collaborative operation optimization model that takes into account the coordinated operation of each distributed cluster and the electric and thermal multi-energy network, and considers incorporating the cost of main transformer overload into the optimization target in the upper-level model, solving the main transformer overload problem through upper and lower linkage optimization; (5) The present invention designs an improved Q-learning algorithm based on the xLSTM architecture to solve the reinforcement learning model of the electric-thermal hybrid energy network cluster division. It is necessary to coordinate the aggregation of nodes in the distribution network and the regional heating network topology. There are many decision actions. The state of the electric-thermal hybrid energy network cluster division includes the load, power generation power, heating power and the number of nodes in the cluster corresponding to each cluster. The state is complex. The xLSTM algorithm is suitable for processing high-dimensional states and continuous action spaces, which is consistent with the characteristics of the electric-thermal hybrid energy network cluster division problem. The algorithm can continuously learn and adapt to environmental changes, optimize the cluster division action strategy and match the multi-state and multi-action electric-thermal hybrid energy network cluster division problem, and has great application potential. When using the Q-learning algorithm to solve the reinforcement learning model of the electric-thermal hybrid energy network cluster division, the present invention introduces the xLSTM algorithm. When selecting the initial action, the xLSTM algorithm is used to learn and select the best action under the given state, thereby improving the accuracy and reliability of the Q-learning algorithm solution. (6) The present invention designs a distributed solution method for a two-layer collaborative operation optimization model of multiple DIES clusters. Considering that there is an uninterrupted parameter transfer process between the upper model and the lower model, and each DIES performs operation optimization as an independent subject, a distributed solution algorithm is designed to solve the two-layer optimization model. The upper-layer multi-DIES collaborative optimization model involves discontinuous variables, which are difficult to solve using traditional mathematical methods. Therefore, a genetic algorithm (GA) is used to solve the upper-layer model; the lower-layer DIES cluster operation optimization model is a typical convex optimization model, which is solved using the commercial solver Cplex. In the GA solution process of the upper-layer model, the Cplex solver for solving the lower-layer model is embedded to construct a distributed solution process, and finally the global optimal solution is obtained. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A flow chart of the regional integrated energy operation optimization method considering the flexible adjustment of multi-energy networks; Figure 2 A schematic diagram of the structure of the research object and adaptation scenario; Figure 3 This is the temperature transfer diagram in the supply and return water pipes; Figure 4 is a Markov decision process diagram. DETAILED DESCRIPTION

[0016] The embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0017] like Figure 1 As shown, the present invention provides a regional integrated energy operation optimization method considering the flexible adjustment of multi-energy networks, comprising the following steps: S1. Build electric-thermal hybrid energy network cluster and multi-DIES cluster; S2. Construct regional distribution network model and regional heating network model; S3. Based on the regional distribution network model and the regional thermal network model, a cluster partitioning model of the electric-thermal hybrid energy network is constructed; S4. Construct a two-tier collaborative operation optimization model for multiple DIES clusters; S5. Solve the electric-thermal hybrid energy network cluster partition model and the multi-DIES cluster two-layer collaborative operation optimization model to complete the optimization.

[0018] The present invention proposes a method for optimizing the operation of a regional integrated energy system considering the flexible adjustment of a multi-energy network. The object of the invention is a regional integrated energy system. The research object and the adaptation scenario are as follows: Figure 2 shown.

[0019] like Figure 2 As shown in the figure, the Regional Integrated Energy System (RIES) consists of an Energy Management System (EMS) located within the power supply range of a 110kV substation, multiple Distribution Area Integrated Energy Systems (DIES) within the 10kV medium-voltage Regional Distribution Network (RDN), distributed photovoltaic clusters, and distributed wind farms. These systems are then connected to the Urban Energy System (UES). The RIES, located at the end of the 10kV / 380V distribution transformer, is equipped with its own power and heat subsystems to provide power and heat to users. This system is complemented by a Regional Heat Network (RHN), which handles heat transmission and distribution.

[0020] The power subsystem of the district-level integrated energy system includes distributed photovoltaic (PV) and electric energy storage (EES) equipment. The thermal subsystem includes various devices, including electric boilers (EB), heat pumps (HP), gas boilers (GB), combined heat and power (CHP), and heat energy storage (HES). The energy service provider (ESP) leads and collaborates with RIES (Resource Energy Services) (RIES) for operational optimization, responsible for the overall low-carbon economic operation of the system and the operation and maintenance of the electric and thermal energy networks. During system operation, the ESP can optimize the coordinated operation strategies with various DIES (Distributed Energy Systems), PV clusters, and wind turbine (WT) clusters by setting reasonable electricity prices, and optimize the power interaction strategies between the RIES, the power grid, and the thermal utility.

[0021] like Figure 2 As shown in Figure 1, the integrated energy systems of each substation are distributed at different nodes in the distribution network and the regional heating network. There are also other electric and thermal load nodes in the distribution network and the regional heating network, which are supplied with heat and power through ESP.

[0022] In the first phase, a modularity index representing the structural tightness of the energy network and a supply-demand balance index representing the matching degree of supply and demand of electric / thermal power were established for the electric / thermal hybrid energy network containing multiple integrated energy systems in different substations. A cluster division optimization model was constructed by weighting the indicators, dividing the electric / thermal hybrid energy network into multiple clusters centered on DIES. DIES is responsible for supplying power and heat to user load nodes within the cluster.

[0023] In the second phase, a two-layer collaborative operation optimization model was constructed for the divided DIES clusters, with the goal of reducing the risk of overload on the distribution network's main transformers. The lower-layer model established an operation optimization model for a single DIES cluster and optimized the DIES cluster's operation strategy. The upper-layer model also established a collaborative operation optimization model for multiple DIES clusters. The ESP dynamically adjusted the distribution network electricity price to optimize the total power demand of the multiple DIES clusters on the distribution network. Through iterative optimization between the upper and lower-layer models, the optimal operation strategy for each DIES cluster and the optimal operation strategy for the distribution network's main transformers was determined.

[0024] In the embodiment of the present invention, in S2, the expression of the regional distribution network model is: ; Where, F ijRepresented by the node i With node j The power flow caused by the change of power transmission between f 1i Represents a slave node i The power flow change on branch 1 caused by the power flowing into and out of the reference node, f 1j Representation node j The power flow change on branch 1 caused by the power flowing into and out of the reference node, f 2i Representation node i The power flow change on branch 2 caused by the power flowing into and out of the reference node, f 2j Representation node j The power flow change on branch 2 caused by the power flowing into and out of the reference node, f li Represents a slave node i The branch caused by the power flowing into and out of the reference node l The trend changes, f lj Represents a slave node j The branch caused by the power flowing into and out of the reference node l changes in trends; In S2, the regional heating network model includes the total heat loss of the heating pipeline and the head loss along the pipeline; The district heating network will have a large amount of energy loss during long-distance heat transmission, mainly including heat energy loss and head loss. Figure 3 This is the temperature transfer in supply and return pipes. Due to the inconsistent temperatures inside and outside the pipes, the medium being transported in the pipes undergoes a heat exchange process. Therefore, during this heat transfer process, there will be a certain amount of heat loss along the pipes, resulting in a lower outlet temperature than the inlet temperature. T s- Indicates the inlet temperature of the heating pipe. T s+ Indicates the outlet temperature of the heating pipe. T R- represents the inlet temperature of the hot return pipe, T R+ represents the outlet temperature of the hot return pipe, L Indicates the length of the pipe.

[0025] Total heat loss of heating pipes E ij The expression is: ; Where, L ijRepresents a slave node i To Node j The length of the pipe, β represents the additional coefficient of heat loss, q id Indicates the heat loss per unit length of pipe under standard working conditions; In addition to heat loss, head loss is another important indicator for describing pressure loss within a pipeline. Head loss is the mechanical energy loss per unit mass of liquid during the flow of water, caused by external resistance to the flow. Head loss includes both longitudinal and local head losses. For long-distance heat transmission in pipelines, local head losses can be neglected. The Darcy-Weisbach equation can be used to calculate the head loss along the pipeline.

[0026] Head loss along the pipeline H ij The expression is: ; Where, Y Indicates the pipe roughness, r represents the density of the transmission medium, d ij Indicates the inner diameter of the pipe, G Indicates the transmission medium traffic.

[0027] In this embodiment of the present invention, S3 includes the following sub-steps: S31. Calculate the distribution network modularity index of the electric-thermal hybrid energy network cluster based on the regional distribution network model; S32. Calculate the thermal network modularity index of the electric-thermal hybrid energy network cluster based on the regional thermal network model; S33, calculating the power balance index of the distribution network of the electric-thermal hybrid energy network cluster; S34. Calculate the heat network power balance index of the electric-heat hybrid energy network cluster; S35. Construct an electric-thermal hybrid energy network cluster partitioning model based on the distribution network modularity index, the heat network modularity index, the distribution network power balance index, and the heat network power balance index.

[0028] Cluster division primarily leverages the coordination and complementarity between nodes to improve the matching between the output power of coupled devices in a DIES cluster and actual load requirements. Furthermore, the divided clusters must also ensure sufficient internal structural strength. Therefore, this paper proposes cluster division metrics based on modularity and power balance.

[0029] In an embodiment of the present invention, in S31, for the distribution network, a weight matrix for the distribution network partitioning can be defined based on the power flow changes caused by power transmission between different nodes. According to the distribution network model, all branches with small power flow changes are divided into the same cluster, while branches with large power flow changes are divided into different clusters. When dividing the distribution network, the smaller the power flow change, the greater the edge weight. Distribution network modularity index D E The expression is: ; Where, m represents the sum of the weights of all edges in the network, w i Represents all nodes i The sum of the weights of the connected edges, w j The table represents all nodes j The sum of the weights of the connected edges, F ij Represented by the node i With node j The power flow caused by the change of power transmission between d E ( i , j ) represents the distribution network node ownership determination factor; In S32, for the district heating network, the heat medium in the pipeline needs to overcome resistance during the flow process, which will cause a certain amount of temperature loss and head loss. Therefore, the weight matrix of the district heating network division can be described by temperature loss and head loss. According to the established district heating network model, the pipelines with less temperature loss and head loss are divided into the same cluster, while the pipelines with greater temperature loss and head loss are divided into different clusters. When dividing the district heating network, the smaller the temperature loss and head loss of the pipeline, the greater the edge weight. Thermal network modularity index D H The expression is: ; Where, H ij represents the head loss along the pipeline, E ij Represents the total heat loss of the heating pipe, d H ( i , j ) represents the regional heating network node ownership determination factor; In S33, the power balance index is used to characterize the matching degree between the electric / thermal power output of the divided cluster and the actual electric / thermal load demand. QP The expression is: ; Where, f P,k Indicates the k The power balance index of a DIES cluster, K Indicates the number of multi-DIES clusters; In S34, the heat network power balance index Q H The expression is: ; Where, f H,k Indicates the k Thermal power balance index of a DIES cluster; In S35, first, under the premise of ensuring that the nodes within the formed cluster are closely connected and the cluster's supply capacity meets internal demand, the modularity index and the supply-demand matching index are used to establish the optimization objectives for the initial cluster division at the heating network and power grid levels, respectively. The cluster division results must, on the one hand, ensure the connection strength of the nodes within the cluster. That is, the larger the modularity index D value, the closer the internal connection of the cluster and the better the structural performance; on the other hand, they must ensure that the cluster has sufficient energy supply capacity. That is, the larger the power balance index Q, the stronger the internal energy supply capacity of the cluster and the better the functional performance.

[0030] like Figure 4 As shown, the present invention is based on the Markov decision process and models the first stage electric-thermal hybrid energy network cluster partition problem as a reinforcement learning model. The electric-thermal energy network cluster partition reinforcement learning model generally consists of a four-tuple ( Action , State , Reward , π )definition, π represents the agent's strategy set and represents the state space State To the action space Action The mapping, Reward Denotes the reward function. A cluster partitioning model for electric-thermal hybrid energy network is constructed using reinforcement learning model.

[0031] For the cluster division process of electric and thermal energy networks, the decision variable is the node division decision in the distribution network / regional heating network, that is, the node ownership determination factor in the modularity index is used. d ( i , j ) to cluster nodes and determine whether to group the nodes into the same cluster. i and j When in the same cluster d (i , j )=1, otherwise d ( i , j )=0. Action space of cluster partitioning model for electric-thermal hybrid energy network Action The expression is: ; For the cluster division process of electric and thermal energy networks, in the distribution network / regional heating network action space Action After execution, the distribution network / regional heating network is divided into multiple clusters centered on DIES. At this time, the load, power generation power, heating power and number of nodes in each cluster will change. The above changes will directly affect the agent reward function. Reward Therefore, the present invention defines the state space State The state space of the cluster partition model of the electric-thermal hybrid energy network is the set of load, power generation, heating power, and number of nodes in the partitioned cluster. State The expression is: ; Where, p k,m,t Represents a cluster k midpoint m The electrical load demand, h k,m,t,t+Δt Represents a cluster node m In time Thermal power demand, Represents a cluster k exist t The electrical power at the moment, Represents a cluster k exist t Thermal power at the moment, M E Represents a cluster k The number of distribution network nodes in M H Represents a cluster k The number of grid nodes in the district heating network; For the cluster division process of electric and thermal energy networks, in the distribution network / regional heating network action space Action After execution, the load, power generation, heating power, and number of nodes in each cluster of the distribution network / regional heating network will change, resulting in changes in its modularity index and power balance index. Therefore, the present invention constructs a reward function that includes both modularity index and power balance index by setting weight coefficients. Reward function of the cluster partitioning model of the electric and thermal hybrid energy network Reward The expression is: ; Where, w p represents the weight coefficient of the two indicators in the distribution network, D E represents the modularity index of the distribution network, D H represents the thermal network modularity index, w H represents the weight coefficient of the two indicators in the regional heating network, and max(·) represents the maximum value function; Agent strategy set reward function π ( S t , A t ) represents the state of the agent facing S t Select action A t The present invention adopts the ε-greedy method, which defines an exploration rate ε between 0 and 1. When selecting an action, the agent has a probability of ε to select a random action and a probability of (1-ε) to select the optimal action with the largest action value. The action value is determined by the immediate reward. R t The action-value function of the cluster partitioning model of the electric-thermal hybrid energy network is composed of Q t ( S t , A t ) is: ; Where, R t Indicates immediate reward, Q t ( S t+1 , A t+1 ) represents the action-value function for the next time period, S t express t The state of the moment, A t express t The action of the moment, S t+1 Indicates the status of the next time period. A t+1 Indicates the action in the next time period.

[0032] In the embodiment of the present invention, in S4, the multi-DIES cluster two-layer collaborative operation optimization model includes a lower-layer DIES cluster operation optimization model and an upper-layer DIES cluster operation optimization model.

[0033] In this embodiment of the present invention, the lower-level model establishes an economic operation optimization model for a single DIES cluster, optimizing the DIES cluster's operation strategy. The lower-level optimization model minimizes operating costs as its objective function. The operating costs of a DIES cluster include operating and maintenance costs, fuel costs, electricity purchase costs, and heat purchase costs. The lower-level DIES cluster operation optimization model is expressed as: ; ; ; ; ; ; Where, C IES,k Indicates the operating cost of a multi-DIES cluster, C en,k represents the energy cost of a multi-DIES cluster, C om,k Indicates the maintenance cost of a multi-DIES cluster, C E,k represents the interaction cost between the multi-DIES cluster and the grid, C NG,k represents the fuel cost of a multi-DIES cluster, C H,k represents the thermal energy cost of a multi-DIES cluster, T Indicates the optimization time period, represents the price at which the DIES cluster sells electricity to the ESP, represents the electricity selling price of ESP, represents the market heat price, represents the thermal power purchased from the regional heating network by the multi-DIES cluster, max(·) represents the maximum value function, P grid,k,t represents the power transaction power between the DIES cluster and the ESP, P CHP,k,t Indicates the k Power of CHP units in a DIES cluster, P GB,k,t Indicates the k GB unit power in a DIES cluster, c gas represents the natural gas price, orCHP,k Indicates the k The operating efficiency of CHP units in a DIES cluster, LHV ng Indicates the lower calorific value of natural gas. or GB,k Indicates the k The operating efficiency of GB units in a DIES cluster, c w Indicates the k Units in a DIES cluster w Unit operating power operation and maintenance cost, P k,w,t Indicates the k Units in a DIES cluster w Operating power, W Indicates the k The number of units in a DIES cluster; The constraints that the lower-level DIES operation optimization model needs to meet include electrical / thermal power balance constraints and unit operation constraints. The constraint expression of the lower-level DIES cluster operation optimization model is: ; ; ; ; ; Where, SOC ES,t+1 Indicates energy storage t +1 moment state of charge, SOC ES,t Indicates energy storage t The state of charge at the moment, α l represents the energy loss rate of energy storage self-discharge, Indicates the charging power of the energy storage, represents the energy storage charging efficiency, Indicates the rated charge and discharge power of energy storage, Δ t represents the optimization time interval, Indicates the rated capacity of energy storage, represents the discharge power of the energy storage, represents the energy storage discharge efficiency, M represents the energy storage charging and discharging state variable, Indicates the minimum state of charge for energy storage, Indicates the maximum state of charge of the energy storage, SOC ES,1 represents the expected state of charge of the energy storage initial state of charge, SOC ES,TIndicates the expected state of charge at the end of the energy storage dispatch cycle; The upper-level model uses multiple DIES clusters as the target and establishes a collaborative operation optimization model. The ESP optimizes the total power demand of multiple DIES clusters on the distribution network by dynamically adjusting the distribution network electricity price. The objective function of the upper-level multi-DIES cluster collaborative operation optimization model is to minimize the system operating cost, which includes the transaction costs between the ESP and the grid and the main transformer overload cost. The upper-level DIES cluster operation optimization model is expressed as: ; ; Where, C ENO represents the total operating cost, max(·) represents the maximum value function, P Tr,t Indicates the output power of the main transformer, It indicates the price of electricity purchased by the distribution network from the upper main grid. C over represents the overload cost of the main transformer, Indicates the rated power of the main transformer, c Tr Indicates the overload loss coefficient of the main transformer; The constraint expression of the lower-level DIES cluster operation optimization model is: ; Where, Indicates the k The power transaction power between the DIES cluster and the distribution network, P Cwt,t Indicates the power generated by the wind farm cluster in the distribution network.

[0034] In the embodiment of the present invention, in S5, the xLSTM algorithm is used to determine the best action in a given state to complete the solution of the electric-thermal hybrid energy network cluster partitioning model; In S5, a genetic algorithm is used to solve the upper-level DIES cluster operation optimization model; In S5, the Cplex commercial solver is used to solve the underlying DIES cluster operation optimization model.

[0035] In an embodiment of the present invention, Q-learning is a reinforcement learning algorithm that is not based on an environmental model and is value-based. The main idea of ​​Q-learning is to define a state-action value function, namely the Q function, and substitute the observed data into the following update formula to iteratively learn the Q function. The Q-learning algorithm learns the optimal strategy by continuous trial and error and updating the Q table. It does not need to know the complete model of the environment and is a model-free reinforcement learning method. The core of the Q-learning algorithm lies in the application of the Bellman equation and the use of the ε-greedy strategy, which together promote the exploration and utilization capabilities of the intelligent agent, enabling the intelligent agent to find the optimal or near-optimal behavior strategy in a complex environment.

[0036] xLSTM (Extended Long Short-Term Memory) is an extension and improvement of the traditional LSTM model, effectively improving LSTM's performance in processing time series data and sequence prediction tasks. xLSTM can scale to billions of parameters, unlocking new possibilities for language modeling and sequence processing tasks. In multiple benchmarks, xLSTM has demonstrated capabilities that surpass conventional models, particularly in scenarios requiring the understanding and generation of highly contextualized and complex text.

[0037] The reinforcement learning model for clustering of hybrid electric and thermal energy networks established by the present invention requires coordinating the aggregation of nodes in the distribution network and regional heating network topology, and involves a large number of decision-making actions. The states of clustering of hybrid electric and thermal energy networks include the load, power generation, heating power, number of nodes within the cluster, and complex states corresponding to each cluster. The xLSTM algorithm is suitable for processing high-dimensional states and continuous action spaces, which is consistent with the characteristics of the clustering problem of hybrid electric and thermal energy networks. The algorithm can continuously learn and adapt to environmental changes, optimize the clustering action strategy to match the multi-state, multi-action clustering problem of hybrid electric and thermal energy networks, and has great application potential.

[0038] Therefore, when using the Q-learning algorithm to solve the reinforcement learning model of electric-thermal hybrid energy network cluster partitioning, the present invention introduces the xLSTM algorithm. When the initial action A1 is selected, the xLSTM algorithm is used to learn and select the optimal action A1 under the given state S1, thereby improving the accuracy and reliability of the Q-learning algorithm solution.

[0039] A continuous parameter transfer process exists between the upper and lower-level models, and each DIES operates as an independent entity for optimization. Therefore, a distributed solution algorithm is constructed to solve the two-level optimization model. The upper-level multi-DIES collaborative optimization model involves discontinuous variables, which are difficult to solve using traditional mathematical methods. Therefore, a genetic algorithm (GA) is used to solve the upper-level model. The lower-level DIES cluster optimization model is a typical convex optimization model, solved using the commercial solver Cplex. The Cplex solver for the lower-level model is embedded in the GA solution of the upper-level model to construct a distributed solution process. First, a randomly generated ESP price is used to guide the operation optimization of each DIES in the lower-level layer through Cplex, resulting in the interaction power between each DIES and the distribution network. Then, a GA is executed to solve the upper-level multi-DIES collaborative optimization model, optimizing the operation strategies for the WT, PV, and main transformer in the distribution network. Through continuous iterations of the GA and Cplex, the global optimal solution is ultimately achieved.

[0040] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and it should be understood that the scope of protection of the present invention is not limited to such specific descriptions and embodiments. Those skilled in the art can make various other specific variations and combinations based on the technical teachings disclosed in the present invention without departing from the essence of the present invention, and such variations and combinations are still within the scope of protection of the present invention.

Claims

1. A regional integrated energy operation optimization method considering flexible adjustment of multi-energy networks, characterized by: The following steps are involved: S1. Build electric-thermal hybrid energy network cluster and multi-DIES cluster; S2. Construct regional distribution network model and regional heating network model; S3. Based on the regional distribution network model and the regional thermal network model, a cluster partitioning model of the electric-thermal hybrid energy network is constructed; S4. Construct a two-tier collaborative operation optimization model for multiple DIES clusters; S5. Solve the electric-thermal hybrid energy network cluster partition model and the multi-DIES cluster two-layer collaborative operation optimization model to complete the optimization.

2. The regional integrated energy operation optimization method considering flexible adjustment of multi-energy networks according to claim 1 is characterized in that: In S2, the expression of the regional distribution network model is: ； Where, F ij Represented by the node i With node j The power flow caused by the change of power transmission between f 1i Represents a slave node i The power flow change on branch 1 caused by the power flowing into and out of the reference node, f 1j Representation node j The power flow change on branch 1 caused by the power flowing into and out of the reference node, f 2i Representation node i The power flow change on branch 2 caused by the power flowing into and out of the reference node, f 2j Representation node j The power flow change on branch 2 caused by the power flowing into and out of the reference node, f li Represents a slave node i The branch caused by the power flowing into and out of the reference node l The trend changes, f lj Represents a slave node j The branch caused by the power flowing into and out of the reference node l changes in trends; In said S2, the regional heating network model includes the total heat loss of the heating pipeline and the head loss along the pipeline; The total heat loss of the heating pipeline E ij The expression is: ； Where, L ij Represents a slave node i To Node j The length of the pipe, β represents the additional coefficient of heat loss, q id Indicates the heat loss per unit length of pipe under standard working conditions; The head loss along the pipeline H ij The expression is: ； Where, Y Indicates the pipe roughness, ρ represents the density of the transmission medium, d ij Indicates the inner diameter of the pipe, G Indicates the transmission medium traffic.

3. The regional integrated energy operation optimization method considering flexible adjustment of multi-energy networks according to claim 1 is characterized in that: The S3 includes the following sub-steps: S31. Calculate the distribution network modularity index of the electric-thermal hybrid energy network cluster based on the regional distribution network model; S32. Calculate the thermal network modularity index of the electric-thermal hybrid energy network cluster based on the regional thermal network model; S33, calculating the power balance index of the distribution network of the electric-thermal hybrid energy network cluster; S34. Calculate the heat network power balance index of the electric-heat hybrid energy network cluster; S35. Construct an electric-thermal hybrid energy network cluster partitioning model based on the distribution network modularity index, the heat network modularity index, the distribution network power balance index, and the heat network power balance index.

4. The regional integrated energy operation optimization method considering flexible adjustment of multi-energy networks according to claim 3 is characterized in that: In S31, the distribution network modularity index D E The expression is: ； Where, m represents the sum of the weights of all edges in the network, w i Represents all nodes i The sum of the weights of the connected edges, w j The table represents all nodes j The sum of the weights of the connected edges, F ij Represented by the node i With node j The power flow caused by the change of power transmission between δ E ( i , j ) represents the distribution network node ownership determination factor; In S32, the thermal network modularity index D H The expression is: ； Where, H ij represents the head loss along the pipeline, E ij Represents the total heat loss of the heating pipe, δ H ( i , j ) represents the regional heating network node ownership determination factor; In S33, the power balance index of the distribution network Q P The expression is: ； Where, φ P,k Indicates the k The power balance index of a DIES cluster, K Indicates the number of multi-DIES clusters; In said S34, the heat network power balance index Q H The expression is: ； Where, φ H,k Indicates the k Thermal power balance index of a DIES cluster; In said S35, a reinforcement learning model is used to construct an electric-thermal hybrid energy network cluster partitioning model; The action space of the electric-thermal hybrid energy network cluster partitioning model Action The expression is: ； The state space of the electric-thermal hybrid energy network cluster partitioning model State The expression is: ； Where, p k,m,t Represents a cluster k midpoint m The electrical load demand, h k,m,t,t+Δt Represents a cluster node m In time Thermal power demand, Represents a cluster k exist t The electrical power at the moment, Represents a cluster k exist t Thermal power at the moment, M E Represents a cluster k The number of distribution network nodes in M H Represents a cluster k The number of grid nodes in the district heating network; The reward function of the electric-thermal hybrid energy network cluster partitioning model Reward The expression is: ； Where, w p represents the weight coefficient of the two indicators in the distribution network, D E represents the modularity index of the distribution network, D H represents the thermal network modularity index, w H represents the weight coefficient of the two indicators in the regional heating network, and max(·) represents the maximum value function; The action value function of the electric-thermal hybrid energy network cluster partitioning model Q t ( S t , A t ) is: ； Where, R t Indicates immediate reward, Q t ( S t+1 , A t+1 ) represents the action-value function for the next time period, S t express t The state of the moment, A t express t The action of the moment, S t+1 Indicates the status of the next time period. A t+1 Indicates the action in the next time period.

5. The regional integrated energy operation optimization method considering flexible adjustment of multi-energy networks according to claim 1 is characterized in that: In S4, the multi-DIES cluster two-layer collaborative operation optimization model includes a lower-layer DIES cluster operation optimization model and an upper-layer DIES cluster operation optimization model.

6. The regional integrated energy operation optimization method considering flexible adjustment of multi-energy networks according to claim 5 is characterized in that: The expression of the lower-layer DIES cluster operation optimization model is: ； ； ； ； ； ； Where, C IES,k Indicates the operating cost of a multi-DIES cluster, C en,k represents the energy cost of a multi-DIES cluster, C om,k Indicates the maintenance cost of a multi-DIES cluster, C E,k represents the interaction cost between the multi-DIES cluster and the grid, C NG,k represents the fuel cost of a multi-DIES cluster, C H,k represents the thermal energy cost of a multi-DIES cluster, T Indicates the optimization time period, represents the price at which the DIES cluster sells electricity to the ESP, represents the electricity selling price of ESP, represents the market heat price, represents the thermal power purchased from the regional heating network by the multi-DIES cluster, max(·) represents the maximum value function, P grid,k,t represents the power transaction power between the DIES cluster and the ESP, P CHP,k,t Indicates the k Power of CHP units in a DIES cluster, P GB,k,t Indicates the k GB unit power in a DIES cluster, c gas represents the natural gas price, η CHP,k Indicates the k The operating efficiency of CHP units in a DIES cluster, LHV ng Indicates the lower calorific value of natural gas. η GB,k Indicates the k The operating efficiency of GB units in a DIES cluster, c w Indicates the k Units in a DIES cluster w Unit operating power operation and maintenance cost, P k,w,t Indicates the k Units in a DIES cluster w Operating power, W Indicates the k The number of units in a DIES cluster; The constraint condition of the lower-layer DIES cluster operation optimization model is expressed as follows: ； ； ； ； ； Where, SOC ES,t+1 Indicates energy storage t +1 moment state of charge, SOC ES,t Indicates energy storage t The state of charge at the moment, α l represents the energy loss rate of energy storage self-discharge, Indicates the charging power of the energy storage, represents the energy storage charging efficiency, Indicates the rated charge and discharge power of energy storage, Δ t represents the optimization time interval, Indicates the rated capacity of energy storage, represents the discharge power of the energy storage, represents the energy storage discharge efficiency, M represents the energy storage charging and discharging state variable, Indicates the minimum state of charge for energy storage, Indicates the maximum state of charge of the energy storage, SOC ES,1 represents the expected state of charge of the energy storage initial state of charge, SOC ES,T Indicates the expected state of charge at the end of the energy storage dispatch cycle; The expression of the upper-layer DIES cluster operation optimization model is: ； ； Where, C ENO represents the total operating cost, max(·) represents the maximum value function, P Tr,t Indicates the output power of the main transformer, It indicates the price of electricity purchased by the distribution network from the upper main grid. C over represents the overload cost of the main transformer, Indicates the rated power of the main transformer, c Tr Indicates the overload loss coefficient of the main transformer; The constraint condition of the lower-layer DIES cluster operation optimization model is expressed as follows: ； Where, Indicates the k The power transaction power between the DIES cluster and the distribution network, P Cwt,t Indicates the power generated by the wind farm cluster in the distribution network.

7. The regional integrated energy operation optimization method considering flexible adjustment of multi-energy networks according to claim 1 is characterized in that: In S5, the xLSTM algorithm is used to determine the best action in a given state to complete the solution of the electric-thermal hybrid energy network cluster partitioning model; In said S5, a genetic algorithm is used to solve the upper layer DIES cluster operation optimization model; In S5, the Cplex commercial solver is used to solve the underlying DIES cluster operation optimization model.

Images