Distributed cold station planning optimization method for novel energy system

Abstract

The invention discloses a distributed cold station planning optimization method for a novel energy system, and the method comprises the steps: carrying out the cluster division of cold stations, constructing a distributed cold station cluster network based on the division result, carrying out the graph calculation of each cluster, solving the power flow of each line in the distributed cold station cluster network according to the graph calculation result, adding the line constraint, and carrying out the planning optimization of the distributed cold stations. Constructing a load demand prediction model based on historical load time sequence data and the flow distribution strategy, predicting a state space of a cold station according to a prediction result, obtaining a cold station operation strategy, and carrying out frequency conversion operation and start-stop time control on the distributed cold station cluster network based on the cold station operation strategy. The method not only can improve the energy efficiency and accuracy of the operation of the distributed cold station, but also has good interpretability, and is of great significance for improving the reliability of an energy system and reducing the operation cost.

Description

Technical Field

[0001] The present invention relates to the field of energy facility planning, and in particular to a distributed cooling station planning optimization method for a new energy system. Background Art

[0002] With the growth of global energy demand and increasingly severe environmental problems, the development and optimization of new energy systems have become crucial. As an important part of the new energy system, the planning and optimization of distributed cooling stations are of great significance for improving energy efficiency, reducing operating costs and reducing environmental impact.

[0003] Traditional cooling station planning optimization methods mainly rely on simple trend analysis for load forecasting, lack system benefits and overall optimization among equipment, and lack intelligent control methods and strategies for the operation and control of cooling stations.

[0004] At present, although there are some planning and optimization methods for cooling stations based on intelligent algorithms or big data, most of the research on configuration optimization is carried out from the system level. However, for regional distributed cooling stations of new energy systems in the planning stage, it is necessary to simultaneously consider the optimization methods at the decision-making level based on equipment performance. Summary of the invention

[0005] The purpose of the present invention is to provide a distributed cooling station planning optimization method for a new energy system.

[0006] To achieve the above object, the present invention is implemented according to the following technical solutions:

[0007] The present invention provides a distributed cooling station planning optimization method for a new energy system, comprising the following steps:

[0008] Cluster the cold stations and build a distributed cold station cluster network based on the clustering results;

[0009] Perform graph calculation on each cluster, and solve the flow of each line in the distributed cold station cluster network according to the graph calculation results. If there is a line exceeding the limit, add a line constraint. If not, end the solution process and obtain the flow distribution strategy;

[0010] Acquire historical load time series data of the cold station, build a load demand prediction model based on the historical load time series data and the flow distribution strategy, predict the state space of the cold station according to the prediction result of the load demand prediction model, and obtain the cold station operation strategy;

[0011] Based on the cold station operation strategy, the distributed cold station cluster network is operated with variable frequency and the start and stop time is controlled.

[0012] Furthermore, the method of clustering the cold stations and building a distributed cold station cluster network based on the clustering results includes:

[0013] Initialize each cold station as a separate cluster, give priority to selecting the cold stations closest to each other as the same cluster, and obtain a set of cluster division strategies;

[0014] From the cluster partitioning strategy set, select the partitioning strategy with the greatest inter-cluster complementarity as the cluster partitioning method, and calculate the inter-cluster complementarity. The expression is:

[0015]

[0016] in, Expressed as the number of cold stations, α h It is expressed as the load distribution ratio of the hth cooling station, R h It is expressed as the rated cooling capacity of the hth cooling station, S h It is expressed as the spare cooling capacity of the hth cooling station, σ(α h ,R h ,S h ) represents the load demand fluctuation of the hth cooling station, m represents the average rated cooling capacity of all cooling stations, Represents the total number of clusters after division, A k Denotes the cooling capacity utilization of the kth cluster, T k Denoted as the average temperature set point of the kth cluster;

[0017] In order to maximize the energy efficiency ratio, a cold energy transmission path is constructed between clusters. The clusters are used as nodes and the cold energy transmission path as edges to build a distributed cold station cluster network.

[0018] Furthermore, the method of performing graph calculation on each cluster includes:

[0019] Convert each cluster into a directed graph, where nodes represent cold stations and edges represent pipelines connecting cold stations;

[0020] Taking the pipeline pressure loss as the weight, assign weights to each edge, determine the inflow and outflow cold energy of each node, and construct the flow matrix;

[0021] The graph algorithm is used to obtain the minimum pipeline pressure loss path between cold stations, and the pipeline pressure loss of the minimum pipeline pressure loss path is used as the new weight to update the flow matrix.

[0022] Furthermore, the method for solving the flow of each line in the distributed cold station cluster network according to the graph calculation result includes:

[0023] Initialize the flow value of each path in the distributed cold station cluster network. The expression is:

[0024]

[0025] Among them, f ij Expressed as the mass flow rate of the path, C p,ij Expressed as the specific heat capacity of the refrigerant along the path, T s,i and T s,j Respectively represent the saturation temperature of the refrigerant at the starting node i and the ending node j on the path, η sys Expressed as heat exchange efficiency, ΔP ij Expressed as the pressure drop along the path, L ij Expressed as the length of the path, ρ ij Expressed as the density of the refrigerant on the path, V ij Denoted as the fluid velocity of the path, ξ ij and ∈ ij They are respectively represented as the friction coefficient and local resistance coefficient of the path, Q loss,ij Expressed as heat loss along the path.

[0026] Furthermore, if there is a line limit violation, the method of adding line constraints includes:

[0027] Get the constraint range of the line, check whether the flow of each edge exceeds the capacity limit, and add line constraints with flow restrictions for the lines that exceed the constraints;

[0028] After adding the constraints, recalculate the power flow until all lines are within the constraints. The expression for recalculating the power flow is:

[0029]

[0030] Among them, Q ij It is represented as the initial power flow value of the path, ΔP max,ij Expressed as the maximum allowable pressure drop of the path, ΔP ij Expressed as the pressure drop along the path, ΔP safety,ij Expressed as the safe pressure drop range of the path, T max,ij Expressed as the maximum allowable temperature of the refrigerant in the path, T ij It is expressed as the actual temperature of the refrigerant in the path, T safety,ij Expressed as the safe temperature range of the path.

[0031] Furthermore, the method for constructing a load demand prediction model based on the historical load time series data and the flow distribution strategy includes:

[0032] The historical load time series data is divided into a training set and a test set in a ratio of 7:3;

[0033] A load demand forecasting model is constructed based on a time series convolutional network, and a model architecture is defined, including at least one convolutional layer, at least one pooling layer, and one fully connected layer. The load demand forecasting model is trained using a training set.

[0034] A penalty term is constructed based on the line constraints of the traffic distribution strategy. A penalty term is added to the loss function. When a line exceeds the limit, the loss value is increased. The expression of the loss function is:

[0035]

[0036] Where N is the number of samples, y r is the real load demand, represents the load demand predicted by the model, λ represents the regularization coefficient, M represents the number of lines, α represents the penalty term coefficient, and F rs It is expressed as the predicted flow of the rth sample on the sth line, C max,s Expressed as the maximum capacity limit of the sth line, It is expressed as the square of the difference between the predicted flow of all samples on the sth line and the maximum capacity limit. If the difference is positive, it means that the line has exceeded the limit.

[0037] Furthermore, the method of predicting the state space of the cooling station according to the prediction result of the load demand prediction model to obtain the cooling station operation strategy includes:

[0038] Construct the state space model of the cold station, the expression is:

[0039]

[0040] Among them, X(t+1) represents the state vector at time step t+1, f represents the state transfer function, that is, the change of state over time, X(t) represents the state vector at time step t, U(t) represents the control input vector at time step t, E(t) represents the external disturbance vector at time step t, among which, C p,w Expressed as the specific heat capacity of water, V s Expressed as the volume of the evaporator, T s Expressed as the temperature of the water in the evaporator, T a Expressed as air temperature, η ch Expressed as the efficiency of the compressor, P c Expressed as the compressor power, Q l Expressed as the liquid flow rate in the evaporator, T r Expressed as the temperature of the refrigerant in the condenser, V r Expressed as the volume of the condenser, h ex (Q l ) is expressed as the efficiency function of the heat exchanger, Pdischarge and P suction They are respectively represented by the discharge pressure and suction pressure of the compressor, ρ is represented by the density of the refrigerant, A ρ Expressed as the area of ​​the compressor piston, V f Expressed as the volume flow rate of refrigerant, τ f It is expressed as the flow time constant of the refrigerant, γ1 is expressed as the ambient temperature influence coefficient, T ambient It is expressed as ambient temperature;

[0041] The historical spatial state and historical load time series data are used to train a state space model that can map load demand to operating state. The load demand predicted by the load demand prediction model is used as input to obtain the predicted value of the spatial state as the operating strategy of the cold station.

[0042] Furthermore, the method for performing frequency conversion operation and controlling start and stop time of the distributed cold station cluster network based on the cold station operation strategy includes:

[0043] The variable frequency control model is constructed using the adaptive control algorithm, and the expression is:

[0044]

[0045] Where u(t) represents the output of the variable frequency controller at time t, K p is expressed as proportional gain, r(t) is expressed as the cooling station operation strategy, i.e. the predicted control state, and y p (t) represents the current state space of the current cold station, K i Expressed as integral gain, K d Expressed as differential gain;

[0046] Convert the output of the frequency converter controller into the control signal of the frequency converter to adjust the operating frequency of the equipment in the cooling station;

[0047] Define decision variables for each device in the cold station, let x σ is the decision variable for the start and stop state of the σth device, where x σ =1 means the device is started, x σ =0 means the equipment remains stopped. A linear programming model is constructed to control the start and stop. The expression is:

[0048]

[0049] Among them, f1 represents the objective function of control start and stop, f2 represents the constraint condition of control start and stop, Z represents the target value to minimize the total energy consumption of start and stop, ψ represents the total number of equipment, E σ It is expressed as the energy consumption of the σth device per unit time, x σIt represents the start and stop status of the σth device, P σ is represented by the load power of the σth device, D is represented by the total load power required, and x σ,last It is represented by the start and stop status of the σth device in the previous decision cycle, f min,σ It is expressed as the minimum operating frequency of the σth device;

[0050] The linear model is solved based on the interior point algorithm to obtain the optimal solution. The start and stop decisions of each device are determined based on the optimal solution, and the start and stop decisions are converted into control signals to control the start and stop time.

[0051] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects:

[0052] (1) The present invention provides a distributed cooling station planning optimization method for a new energy system. A distributed cooling station cluster network is constructed and line constraints are added to solve the flow distribution strategy of the distributed cooling station cluster network. A load demand prediction model is further constructed to predict the state space of the cooling station to obtain a cooling station operation strategy. Finally, based on the cooling station operation strategy, the distributed cooling station cluster network is subjected to frequency conversion operation and start and stop time control. The operating frequency and start and stop time of the equipment are automatically adjusted according to the operation strategy. The overall intelligent control improves the energy utilization rate of the cooling station planning process and the response speed of the cooling station.

[0053] (2) The present invention uses graph computing to perform flow analysis on the cold station cluster network, which can accurately predict and distribute traffic, optimize energy distribution, avoid line over-limit, and improve the energy utilization efficiency and safety of cold station planning.

[0054] (3) The present invention predicts the state space of the cold station by constructing a load demand prediction model to obtain the cold station operation strategy, and can obtain a more accurate operation strategy. The prediction model can quickly respond to load changes and timely adjust the operation status of the cold station to meet demand, thereby improving the operation efficiency and response speed of the cold station. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 The present invention is a flowchart of the steps of a distributed cooling station planning optimization method for a new energy system. DETAILED DESCRIPTION

[0056] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0057] Reference Figure 1 As shown, the present invention provides a distributed cooling station planning optimization method for a new energy system, comprising:

[0058] Step S1, clustering the cold station, and building a distributed cold station cluster network based on the division results;

[0059] In the actual evaluation, the clusters are initialized according to the distances between cold stations, and a set of partitioning strategies is obtained, including: Strategy 1: {A, B}, {C}, {D}, {E}; Strategy 2: {A, C}, {B, D}, {E}; Strategy 3: {A, D, E}, {B, C}. The complementarity between clusters of each strategy is calculated, and Strategy 2 is selected as the cluster partitioning strategy. Further, a cold energy transmission path is constructed between clusters with the highest energy efficiency ratio, and a distributed cold station cluster network is constructed with clusters as nodes and cold energy transmission paths as edges. The energy efficiency ratio of cold energy transmission between clusters is calculated, and the expression is:

[0060]

[0061] in, is the number of clusters after division, EER h It is expressed as the local energy efficiency ratio of the h-th cluster, i.e., the cooling energy provided per unit energy consumption, C h represents the total amount of cooling energy provided to the hth cluster, Loss h is the heat loss ratio of the hth cluster, M is the number of cold energy transmission paths between clusters, and E j It is expressed as the cold energy transported on path m, L m Expressed as the transmission distance on path m, Loss m It is expressed as the pumping energy consumption of path m;

[0062] Specifically, the path from cluster {A, C} to cluster {B, D} is obtained. Cluster {A, C} has extra cooling capacity, and the cold energy is transported to cluster {B, D} with higher demand through the transmission pipeline; from cluster {B, D} to cluster {E}, cluster {B, D} has surplus cold energy, and the cold energy is transported to cluster {E} through the transmission pipeline.

[0063] Step S2: perform graph calculation on each cluster, and solve the flow of each line in the distributed cold station cluster network according to the graph calculation result. If there is a line exceeding the limit, add a line constraint. If not, end the solution process and obtain the flow distribution strategy;

[0064] In the actual evaluation, taking cluster {A, C} as an example, there are two pipelines, which are bidirectional pipelines from cold station A to cold station C. The inflow and outflow of cold energy flows between cold station A and cold station C are determined, and the cold energy supplied by cold station A to cold station C is obtained. The flow matrix F is created. Cold station A has additional cooling capacity and can supply cold energy to cold station C. The cold energy flow F from cold station A to cold station C is calculated. AC , taking the pressure loss of the pipeline during the transportation process as the weight of the edge, the Floyd-Warshall graph algorithm is used to find the shortest path between nodes, that is, the path that minimizes pressure loss;

[0065] Specifically, the initial state of the traffic matrix F of cluster {A, C} is F AC = 100 units, indicating that cooling station A supplies 100 units to cooling station C. Based on the graph algorithm, the pressure loss can be reduced by connecting cooling station A and cooling station C through the newly added pipeline. The flow matrix is ​​updated to F AC =120 units;

[0066] In the actual evaluation, the cooling capacity of cold station A is 1000RT, the cooling load demand of cold station C is 1200RT, the capacity limit of pipeline 1 is 1000RT, the capacity limit of pipeline 2 is 800RT, and the flow of pipeline 1 reaches the capacity limit. Line constraints are added: pipeline 1: water supply temperature 12℃, flow constraint 1000RT, pipeline 2: water supply temperature 10℃, flow constraint 800RT. Linear programming is used to redistribute the flow, with the goal of minimizing the total pressure loss, while meeting the following conditions: the total demand of cold station C meets 1200RT, and the pipeline flow does not exceed the respective capacity limits;

[0067] Specifically, the constraints are as follows: the chilled water supply temperature constraint is set to 6°C to 12°C, and the chilled water flow constraint is set to no more than 1100m 3 / h, and the fan coil air volume constraint is set to 500 to 700 cubic meters per hour (m 3 / h), the cooling tower fan air volume is set to 800 to 1000m 3 / h, the pressure constraints of the fan and cooling water pump are set to 0.3 to 0.5 MPa respectively, and the fresh air volume constraint of the fan coil unit is set to meet the indoor air quality standard, not less than 200m 3 / h, obtain the flow data of line A to C in cluster {A, C}, and detect that the air volume of the cooling tower fan is 700m 3 / h, exceeds the constraint range, add cooling tower fan air volume constraint conditions to line A to C, and further recalculate the power flow;

[0068] Specifically, the flow distribution strategy is: cooling station A delivers 50 kW of cooling capacity to cooling station C, cooling station A delivers 200 liters / minute of cooling water to cooling station B, cooling station B delivers 150 liters / minute of cooling water to cooling station C, cooling station C delivers 100 liters / minute of cooling water to cooling station D, cooling station D delivers 50 liters / minute of cooling water to cooling station E, and cooling station E discharges cooling water to the external environment.

[0069] Step S3, obtaining historical load time series data of the cooling station, building a load demand prediction model based on the historical load time series data and the flow distribution strategy, predicting the state space of the cooling station according to the prediction result of the load demand prediction model, and obtaining the cooling station operation strategy;

[0070] In the actual evaluation, taking cold station A as an example, the historical load time series data of cold station A was obtained and accurately divided into training set and test set in a ratio of 7:3. A prediction model was built based on the time series convolutional network, which contains a deep convolution layer to capture local patterns in time series data, a maximum pooling layer to reduce feature dimensions, and a fully connected layer to output the prediction results. The model was trained using the training set, and the line constraint penalty term was added to the loss function by adjusting the regularization coefficient and the penalty term coefficient to ensure that when the predicted flow exceeds the maximum capacity limit of the line, the loss value will increase accordingly, effectively avoiding the line over-limit situation.

[0071] In the actual evaluation, a state space model of the cooling station was constructed. The model takes into account the influencing factors such as the specific heat capacity of water, the volume of the evaporator and condenser, and the ambient temperature. The model was trained using historical space state and load data. The predicted load demand obtained by the load demand prediction model was used as input. The future operating state of the cooling station was predicted through the state space model, and the operating strategy of cooling station A was obtained.

[0072] Specifically, for cooling station A, when it is predicted that the daytime load demand will increase by 5% during the next weekday, our model predicts that the supply water temperature should be adjusted to 12°C, the compressor power should be increased to 120kW, and the liquid flow rate LL should be increased to 2.5m3 / h to adapt to the load changes and maintain optimal system operation.

[0073] Step S4: performing frequency conversion operation and controlling start and stop time on the distributed cold station cluster network based on the cold station operation strategy.

[0074] In the actual evaluation, taking cold station A as an example, there are 5 devices with a total load demand of D. The variable frequency control model is constructed using the adaptive control algorithm. The model uses proportional gain, integral gain and differential gain as parameters to adjust the operating frequency of the compressor in real time, respond to the operation strategy of the cold station and the current spatial state of the cold station, define decision variables to represent the start and stop state of each device, and construct a linear programming model for controlling start and stop. The objective function aims to minimize the total energy consumption of the start and stop operation while meeting the load demand of the cold station and the equipment operation constraints. The linear programming model is solved using the interior point algorithm to obtain the optimal solution, which provides the start and stop decision of each device.

[0075] Specifically, taking device 1 as an example, the current energy consumption of device 1 is 120 kWh, the proportional gain is set to 0.5, the integral gain is set to 0.05, the differential gain is set to 0.01, and the current operating frequency of device 1 is 50 Hz. According to the load demand prediction model, it is predicted that the load will increase by 10% in the next period of time, so the operating frequency of device 1 needs to be increased from 50 Hz to 55 Hz to cope with the increased load demand;

[0076] Specifically, devices 1 to 3 need to be started from 9 to 11 o'clock, while devices 4 and 5 can remain stopped, all devices remain started from 11 to 20 o'clock, devices 1 and 2 remain started from 20 to 4 o'clock, device 1 is started from 4 to 9 o'clock, and the remaining devices can remain stopped to meet the total load demand while minimizing energy consumption.

[0077] During the overall optimization process, the total power consumption before optimization was 14750kWh, and the total power consumption after optimization was 12401kWh, with energy saving reaching 2349kWh and energy saving rate reaching 15.9%. Through the optimization of cold station planning, the project area successfully reduced electricity consumption by 15.9%.

[0078] In this embodiment, the method of clustering the cold station and building a distributed cold station cluster network based on the clustering result includes:

[0079] Initialize each cold station as a separate cluster, give priority to selecting the cold stations closest to each other as the same cluster, and obtain a set of cluster division strategies;

[0080] From the cluster partitioning strategy set, select the partitioning strategy with the greatest inter-cluster complementarity as the cluster partitioning method, and calculate the inter-cluster complementarity. The expression is:

[0081]

[0082] in, Expressed as the number of cold stations, α h It is expressed as the load distribution ratio of the hth cooling station, R hIt is expressed as the rated cooling capacity of the hth cooling station, S h It is expressed as the spare cooling capacity of the hth cooling station, σ(α h ,R h ,S h ) represents the load demand fluctuation of the hth cooling station, m represents the average rated cooling capacity of all cooling stations, Represents the total number of clusters after division, A k Denotes the cooling capacity utilization of the kth cluster, T k Denoted as the average temperature set point of the kth cluster;

[0083] In order to maximize the energy efficiency ratio, a cold energy transmission path is constructed between clusters. The clusters are used as nodes and the cold energy transmission path as edges to build a distributed cold station cluster network.

[0084] In this embodiment, the method of performing graph calculation on each cluster includes:

[0085] Convert each cluster into a directed graph, where nodes represent cold stations and edges represent pipelines connecting cold stations;

[0086] Taking the pipeline pressure loss as the weight, assign weights to each edge, determine the inflow and outflow cold energy of each node, and construct the flow matrix;

[0087] The graph algorithm is used to obtain the minimum pipeline pressure loss path between cold stations, and the pipeline pressure loss of the minimum pipeline pressure loss path is used as the new weight to update the flow matrix.

[0088] In this embodiment, the method for solving the flow of each line in the distributed cold station cluster network according to the graph calculation result includes:

[0089] Initialize the flow value of each path in the distributed cold station cluster network. The expression is:

[0090]

[0091] Among them, Q ij It is represented as the initial power flow value of the path, ΔP max,ij Expressed as the maximum allowable pressure drop of the path, ΔP ij Expressed as the pressure drop along the path, ΔP safety,ij Expressed as the safe pressure drop range of the path, T max,ij Expressed as the maximum allowable temperature of the refrigerant in the path, T ij It is expressed as the actual temperature of the refrigerant in the path, T safety,ij Expressed as the safe temperature range of the path.

[0092] In this embodiment, the method of adding line constraints if there is a line limit violation includes:

[0093] Get the constraint range of the line, check whether the flow of each edge exceeds the capacity limit, and add line constraints with flow restrictions for the lines that exceed the constraints;

[0094] After adding the constraints, recalculate the power flow until all lines are within the constraints. The expression for recalculating the power flow is:

[0095]

[0096] Among them, Q ij Expressed as the initialization power flow, ΔP max,ij Expressed as the maximum allowable pressure drop of the path, ΔP ij Expressed as the pressure drop along the path, ΔP safety,ij Expressed as the safe pressure drop range of the path, T max,ij Expressed as the maximum allowable temperature of the refrigerant in the path, T ij It is expressed as the actual temperature of the refrigerant in the path, T safety,ij Expressed as the safe temperature range of the path.

[0097] In this embodiment, the method for constructing a load demand prediction model based on the historical load time series data and the flow distribution strategy includes:

[0098] The historical load time series data is divided into a training set and a test set in a ratio of 7:3;

[0099] A load demand forecasting model is constructed based on a time series convolutional network, and a model architecture is defined, including at least one convolutional layer, at least one pooling layer, and one fully connected layer. The load demand forecasting model is trained using a training set.

[0100] A penalty term is constructed based on the line constraints of the traffic distribution strategy. A penalty term is added to the loss function. When a line exceeds the limit, the loss value is increased. The expression of the loss function is:

[0101]

[0102] Where N is the number of samples, y r is the real load demand, represents the load demand predicted by the model, λ represents the regularization coefficient, M represents the number of lines, α represents the penalty term coefficient, and F rs It is expressed as the predicted flow of the rth sample on the sth line, C max,s Expressed as the maximum capacity limit of the sth line, It is expressed as the square of the difference between the predicted flow of all samples on the sth line and the maximum capacity limit. If the difference is positive, it means that the line has exceeded the limit.

[0103] In this embodiment, the method of predicting the state space of the cold station according to the prediction result of the load demand prediction model to obtain the cold station operation strategy includes:

[0104] Construct the state space model of the cold station, the expression is:

[0105]

[0106] Among them, X(t+1) represents the state vector at time step t+1, f represents the state transfer function, that is, the change of state over time, X(t) represents the state vector at time step t, U(t) represents the control input vector at time step t, E(t) represents the external disturbance vector at time step t, among which, C p,w Expressed as the specific heat capacity of water, V s Expressed as the volume of the evaporator, T s Expressed as the temperature of the water in the evaporator, T a Expressed as air temperature, η ch Expressed as the efficiency of the compressor, P c Expressed as the compressor power, Q l Expressed as the liquid flow rate in the evaporator, T r Expressed as the temperature of the refrigerant in the condenser, V r Expressed as the volume of the condenser, h ex (Q l ) is expressed as the efficiency function of the heat exchanger, P discharge and P suction They are respectively represented by the discharge pressure and suction pressure of the compressor, ρ is represented by the density of the refrigerant, A ρ Expressed as the area of ​​the compressor piston, V f Expressed as the volume flow rate of refrigerant, τ f It is expressed as the flow time constant of the refrigerant, γ1 is expressed as the ambient temperature influence coefficient, T ambient Expressed as ambient temperature;

[0107] The historical spatial state and historical load time series data are used to train a state space model that can map load demand to operating state. The load demand predicted by the load demand prediction model is used as input to obtain the predicted value of the spatial state as the operating strategy of the cold station.

[0108] In this embodiment, the method for performing frequency conversion operation and controlling start and stop time of the distributed cold station cluster network based on the cold station operation strategy includes:

[0109] The variable frequency control model is constructed using the adaptive control algorithm, and the expression is:

[0110]

[0111] Where u(t) represents the output of the variable frequency controller at time t, K p is expressed as proportional gain, r(t) is expressed as the cooling station operation strategy, i.e. the predicted control state, and y p (t) represents the current state space of the current cold station, K i Expressed as integral gain, K d Expressed as differential gain;

[0112] Convert the output of the frequency converter controller into the control signal of the frequency converter to adjust the operating frequency of the equipment in the cooling station;

[0113] Define decision variables for each device in the cold station, let x σ is the decision variable for the start and stop state of the σth device, where x σ =1 means the device is started, x σ =0 means the equipment remains stopped. A linear programming model is constructed to control the start and stop. The expression is:

[0114]

[0115] Among them, f1 represents the objective function of control start and stop, f2 represents the constraint condition of control start and stop, Z represents the target value to minimize the total energy consumption of start and stop, ψ represents the total number of equipment, E σ It is expressed as the energy consumption of the σth device per unit time, x σ It represents the start and stop status of the σth device, P σ is represented by the load power of the σth device, D is represented by the total load power required, and x σ,last It is represented by the start and stop status of the σth device in the previous decision cycle, f min,σ It is expressed as the minimum operating frequency of the σth device;

[0116] The linear model is solved based on the interior point algorithm to obtain the optimal solution. The start and stop decisions of each device are determined based on the optimal solution, and the start and stop decisions are converted into control signals to control the start and stop time.

[0117] The above contents are merely examples and explanations of the structure of the present invention. The technicians in this technical field may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the structure of the invention or exceed the scope defined by the claims, they should all fall within the protection scope of the present invention.

Claims

1. A distributed cooling station planning optimization method for a new energy system, characterized in that: The following steps are involved: Cluster the cold stations and build a distributed cold station cluster network based on the clustering results; Perform graph calculation on each cluster, and solve the flow of each line in the distributed cold station cluster network according to the graph calculation results. If there is a line exceeding the limit, add a line constraint. If not, end the solution process and obtain the flow distribution strategy; Acquire historical load time series data of the cold station, build a load demand prediction model based on the historical load time series data and the flow distribution strategy, predict the state space of the cold station according to the prediction result of the load demand prediction model, and obtain the cold station operation strategy; Based on the cold station operation strategy, the distributed cold station cluster network is operated with variable frequency and the start and stop time is controlled.

2. A distributed cooling station planning optimization method for a new energy system according to claim 1, characterized in that: The method of clustering the cold stations and building a distributed cold station cluster network based on the clustering results includes: Initialize each cold station as a separate cluster, give priority to selecting the cold stations closest to each other as the same cluster, and obtain a set of cluster division strategies; From the cluster partitioning strategy set, select the partitioning strategy with the greatest inter-cluster complementarity as the cluster partitioning method, and calculate the inter-cluster complementarity. The expression is: in, Expressed as the number of cold stations, α h It is expressed as the load distribution ratio of the hth cooling station, R h It is expressed as the rated cooling capacity of the hth cooling station, S h It is expressed as the spare cooling capacity of the hth cooling station, σ(α h ,R h ,S h ) represents the load demand fluctuation of the hth cooling station, m represents the average rated cooling capacity of all cooling stations, Represents the total number of clusters after division, A k Denotes the cooling capacity utilization of the kth cluster, T k Denoted as the average temperature set point of the kth cluster; In order to maximize the energy efficiency ratio, a cold energy transmission path is constructed between clusters. The clusters are used as nodes and the cold energy transmission path as edges to build a distributed cold station cluster network.

3. A distributed cooling station planning optimization method for a new energy system according to claim 1, characterized in that: The method for performing graph calculation on each cluster includes: Convert each cluster into a directed graph, where nodes represent cold stations and edges represent pipelines connecting cold stations; Taking the pipeline pressure loss as the weight, assign weights to each edge, determine the inflow and outflow cold energy of each node, and construct the flow matrix; The graph algorithm is used to obtain the minimum pipeline pressure loss path between cold stations, and the pipeline pressure loss of the minimum pipeline pressure loss path is used as the new weight to update the flow matrix.

4. A distributed cooling station planning optimization method for a new energy system according to claim 1, characterized in that: The method for solving the flow of each line in the distributed cold station cluster network according to the graph calculation result includes: Initialize the flow value of each path in the distributed cold station cluster network. The expression is: Among them, f ij Expressed as the mass flow rate of the path, C p,ij Expressed as the specific heat capacity of the refrigerant along the path, T s,i and T s,j Respectively represent the saturation temperature of the refrigerant at the starting node i and the ending node j on the path, η sys Expressed as heat exchange efficiency, ΔP ij Expressed as the pressure drop along the path, L ij Expressed as the length of the path, ρ ij Expressed as the density of the refrigerant on the path, V ij Expressed as the fluid velocity of the path, ξ ij and ∈ ij They are respectively represented as the friction coefficient and local resistance coefficient of the path, Q loss,ij Expressed as heat loss along the path.

5. The distributed cooling station planning optimization method for a new energy system according to claim 1 is characterized in that: The method of adding line constraints if there is a line limit violation includes: Get the constraint range of the line, check whether the flow of each edge exceeds the capacity limit, and add line constraints with flow restrictions for the lines that exceed the constraints; After adding the constraints, recalculate the power flow until all lines are within the constraints. The expression for recalculating the power flow is: Among them, Q ij It is represented as the initial power flow value of the path, ΔP max,ij Expressed as the maximum allowable pressure drop of the path, ΔP ij Expressed as the pressure drop along the path, ΔP safety,ij Expressed as the safe pressure drop range of the path, T max,ij Expressed as the maximum allowable temperature of the refrigerant in the path, T ij It is expressed as the actual temperature of the refrigerant in the path, T safety,ij Expressed as the safe temperature range of the path.

6. A distributed cooling station planning and optimization method for a new energy system according to claim 1, characterized in that: The method for constructing a load demand prediction model based on the historical load time series data and the flow distribution strategy includes: The historical load time series data is divided into a training set and a test set in a ratio of 7:3; A load demand forecasting model is constructed based on a time series convolutional network, and a model architecture is defined, including at least one convolutional layer, at least one pooling layer, and one fully connected layer. The load demand forecasting model is trained using a training set. A penalty term is constructed based on the line constraints of the traffic distribution strategy. A penalty term is added to the loss function. When a line exceeds the limit, the loss value is increased. The expression of the loss function is: Where N is the number of samples, y r is the real load demand, represents the load demand predicted by the model, λ represents the regularization coefficient, M represents the number of lines, α represents the penalty term coefficient, and F rs It is expressed as the predicted flow of the rth sample on the sth line, C max,s Expressed as the maximum capacity limit of the sth line, It is expressed as the square of the difference between the predicted flow of all samples on the sth line and the maximum capacity limit. If the difference is positive, it means that the line has exceeded the limit.

7. A distributed cooling station planning and optimization method for a new energy system according to claim 1, characterized in that: The method of predicting the state space of the cold station according to the prediction result of the load demand prediction model to obtain the cold station operation strategy includes: Construct the state space model of the cold station, the expression is: Among them, X(t+1) represents the state vector at time step t+1, f represents the state transfer function, that is, the change of state over time, X(t) represents the state vector at time step t, U(t) represents the control input vector at time step t, E(t) represents the external disturbance vector at time step t, among which, C p,w Expressed as the specific heat capacity of water, V s Expressed as the volume of the evaporator, T s Expressed as the temperature of the water in the evaporator, T a Expressed as air temperature, η ch Expressed as the efficiency of the compressor, P c Expressed as the compressor power, Q l Expressed as the liquid flow rate in the evaporator, T r Expressed as the temperature of the refrigerant in the condenser, V r Expressed as the volume of the condenser, h ex (Q l ) is expressed as the efficiency function of the heat exchanger, P discharge and P suction They are respectively represented by the discharge pressure and suction pressure of the compressor, ρ is represented by the density of the refrigerant, A ρ Expressed as the area of ​​the compressor piston, V f Expressed as the volume flow rate of refrigerant, τ f It is expressed as the flow time constant of the refrigerant, γ1 is expressed as the ambient temperature influence coefficient, T ambient Expressed as ambient temperature; The historical spatial state and historical load time series data are used to train a state space model that can map load demand to operating state. The load demand predicted by the load demand prediction model is used as input to obtain the predicted value of the spatial state as the operating strategy of the cold station.

8. A distributed cooling station planning and optimization method for a new energy system according to claim 1, characterized in that: The method for performing frequency conversion operation and controlling start and stop time on the distributed cold station cluster network based on the cold station operation strategy includes: The variable frequency control model is constructed using the adaptive control algorithm, and the expression is: Where u(t) represents the output of the variable frequency controller at time t, K p is expressed as proportional gain, r(t) is expressed as the cooling station operation strategy, i.e. the predicted control state, and y p (t) represents the current state space of the current cold station, K i Expressed as integral gain, K d Expressed as differential gain; Convert the output of the frequency converter controller into the control signal of the frequency converter to adjust the operating frequency of the equipment in the cooling station; Define decision variables for each device in the cold station, let x σ is the decision variable for the start and stop state of the σth device, where x σ =1 means the device is started, x σ =0 means the equipment remains stopped. A linear programming model is constructed to control the start and stop. The expression is: f1: f2: Among them, f1 represents the objective function of control start and stop, f2 represents the constraint condition of control start and stop, Z represents the target value to minimize the total energy consumption of start and stop, ψ represents the total number of equipment, E σ It is expressed as the energy consumption of the σth device per unit time, x σ Represents the start and stop status of the σth device, P σ is represented by the load power of the σth device, D is represented by the total load power required, and x σ,last It is represented by the start and stop status of the σth device in the previous decision cycle, f min,σ It is expressed as the minimum operating frequency of the σth device; The linear model is solved based on the interior point algorithm to obtain the optimal solution. The start and stop decisions of each device are determined based on the optimal solution, and the start and stop decisions are converted into control signals to control the start and stop time.

Images