Multi-energy-flow scheduling method and system of building energy system based on demand side response

Abstract

The invention discloses a multi-energy-flow scheduling method and system of a building energy system based on demand side response, and the method comprises the following steps: S1, building a physical feature model and an operation constraint condition of each related device in the building energy system; s2, based on the demand side response, establishing a target function by taking balance of system operation cost and scheduling cost as a target; s3, in combination with the objective function, the physical feature model of each related device and the operation constraint condition, constructing a self-adaptive depth deterministic strategy gradient algorithm fused with a limit gradient lifting classification mechanism; and S4, carrying out classification processing on the load data, training an adaptive depth deterministic strategy gradient algorithm based on the classified data to obtain a corresponding strategy network model, and carrying out optimal scheduling on multiple energy flows of the building energy system based on the strategy network model. Therefore, the response capability in a complex building scene is greatly improved, and the operation robustness of the system is relatively high.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of multi-energy flow scheduling, in particular to a multi-energy flow scheduling method for building energy systems based on demand side response and a multi-energy flow scheduling system for building energy systems based on demand side response. BACKGROUND

[0002] In the development trend of low-carbon building integrated energy systems, the deep integration of electricity, heat and hydrogen energy, especially the integration of renewable energy and clean hydrogen fuel cells, is gradually becoming an important path to achieve the "double carbon" goal. However, such systems generally face strong source-side volatility, high load-side uncertainty and high coupling between multi-energy flows, which brings significant challenges to the coordinated scheduling and optimal control of the system.

[0003] In related technologies, traditional optimization methods rely on battery energy storage systems and heat storage devices to absorb fluctuating renewable energy, in order to improve system flexibility, reduce dependence on fossil fuels and optimize operating costs to a certain extent. However, traditional optimization methods are difficult to cope with multiple uncertainties such as load fluctuations, energy price changes and user responses in building energy systems, resulting in poor adaptability of scheduling strategies, reduced efficiency and low robustness of system operation. SUMMARY

[0004] To solve the above technical problems, the present application provides a multi-energy flow scheduling method for building energy systems based on demand side response, which adopts an adaptive deep deterministic policy gradient algorithm that integrates a limit gradient boosting classification mechanism, can dynamically adjust the strategy according to the demand side response type and system state changes, improves the response capability in complex building scenarios, and has high robustness of system operation.

[0005] The technical solutions adopted by the present application are as follows:

[0006] A multi-energy flow scheduling method for building energy systems based on demand side response, comprising the following steps: S1, establishing physical characteristic models of each related device in the building energy system and operating constraint conditions; S2, establishing an objective function based on the demand side response and balancing system operating costs and scheduling costs; S3, combining the objective function, the physical characteristic models of each related device and the operating constraint conditions, constructing an adaptive deep deterministic policy gradient algorithm that integrates a limit gradient boosting classification mechanism; S4, classifying and processing load data, and training the adaptive deep deterministic policy gradient algorithm based on the classified and processed data to obtain a corresponding policy network model, and optimizing and scheduling the multi-energy flow of the building energy system based on the policy network model.

[0007] In an embodiment of the present application, the building energy system is an electricity-heat-hydrogen low-carbon building energy system, which comprises a photovoltaic power generation device, an electrolyzer, a hydrogen storage tank, a fuel cell, a heat pump, and a heat storage tank.

[0008] In an embodiment of the present application, the physical property model of the photovoltaic power generation device is:

[0009]

[0010] wherein, represents the photovoltaic power generation power, η PV represents the photovoltaic efficiency, η inv represents the inverter efficiency, A PV represents the effective illumination area of the photovoltaic power generation device, I0represents the solar radiation intensity;

[0011] The physical property model of the electrolyzer as the accommodation of daytime photovoltaic power generation is:

[0012]

[0013] wherein, H ET is the hydrogen production rate, η F is the Faraday efficiency, n ET is the number of electrolyzers, I ET is the working current, and F is the Faraday constant;

[0014] The physical property model of the hydrogen storage tank for storing the hydrogen produced by the electrolyzer and providing a hydrogen source for the fuel cell power generation is:

[0015]

[0016] wherein, TP t H represents the tank pressure of the hydrogen storage tank at time t, represents the tank pressure of the hydrogen storage tank at time t+1, represents the hydrogen flow rate of the hydrogen storage tank at time t, R represents the universal gas constant, T H represents the working temperature of the hydrogen storage tank, V H represents the volume of the hydrogen storage tank, z t represents the hydrogen compression coefficient corresponding to time t;

[0017] The physical property model of the fuel cell is:

[0018]

[0019] wherein, H FC is the hydrogen consumption rate of the fuel cell,I FC n is the operating current of the fuel cell, FC A is the number of cells, cell j is the current density of the fuel cell, P is the electric power of the fuel cell, T η is the thermal efficiency of the fuel cell, E η is the electric efficiency of the fuel cell, P is the thermal power of the fuel cell;

[0020] The physical property model of the heat pump is:

[0021]

[0022] wherein COP represents the coefficient of performance of the heat pump, h1 represents the first fitting coefficient, h2 represents the second fitting coefficient, h3 represents the third fitting coefficient, ΔT HP represents the difference between the target temperature and the current environmental temperature of the heat pump, P represents the thermal power of the heat pump, P represents the electric power of the heat pump, COP H represents the heating coefficient of performance, P represents the cooling power of the heat pump, COP C represents the cooling coefficient of performance;

[0023] The physical property model of the heat storage tank is:

[0024]

[0025] wherein HSD t+1 represents the heat storage state of the heat storage tank at t+1, HSD t represents the heat storage state of the heat storage tank at t, represents the charging or discharging power of the heat storage tank at t, H c represents the maximum heat storage amount of the heat storage tank, and ΔT is the time step.

[0026] In an embodiment of the present application, the operation constraint conditions include a safe operation constraint and a power balance constraint, wherein,

[0027] The safe operation constraint is:

[0028]

[0029] HSD min ≤ HSD t ≤ HSD max ,

[0030] TP min ≤TP t ≤TP max ,

[0031] wherein, represents the output power of the fuel cell at time t, and represent the minimum and maximum values of the output power of the fuel cell, respectively, represents the electricity consumption power of the heat pump at time t, and represent the minimum and maximum values of the electricity consumption power of the heat pump, respectively, represents the electrolysis power of the electrolyzer at time t, and represent the minimum and maximum values of the electrolysis power of the electrolyzer, respectively, represents the hydrogen consumption rate of the fuel cell at time t, and represent the minimum and maximum values of the hydrogen consumption rate of the fuel cell, respectively, represents the hydrogen production rate of the electrolyzer at time t, and represent the minimum and maximum values of the hydrogen production rate of the electrolyzer, respectively, t represents the heat storage state of the heat storage tank at time t, min and max represent the minimum and maximum values of the heat storage state of the heat storage tank, respectively, t represents the hydrogen storage state of the hydrogen storage tank at time t, min and max represent the minimum and maximum values of the hydrogen storage state of the hydrogen storage tank, respectively;

[0032] The power balance constraint is:

[0033]

[0034] wherein, represents the output power of the fuel cell at time t, represents the photovoltaic power generation power at time t, represents the electrical load of the building energy system at time t, represents the electricity consumption power of the heat pump at time t, represents the electrolysis power of the electrolyzer at time t, represents the heat production amount of the fuel cell at time t, represents the heat production amount of the heat pump at time t, represents the heat release / charge amount of the heat storage tank at time t, is positive, indicating the heat release amount of the heat storage tank at time t, is negative, indicating the heat charging amount of the heat storage tank at time t, denotes the heat load of the building energy system at time t, denotes the hydrogen production rate of the electrolytic cell at time t, denotes the hydrogen flow rate of the hydrogen storage tank at time t, denotes the hydrogen consumption rate of the fuel cell at time t, denotes the hydrogen load of the building energy system at time t.

[0035] In an embodiment of the present application, the objective function is:

[0036]

[0037] wherein J denotes the objective function, λ1 denotes the first adjustable parameter, λ2 denotes the second adjustable parameter, λ3 denotes the third adjustable parameter, λ4 denotes the fourth adjustable parameter, i represents the i-th device in the building energy system, OC i denotes the unit operating cost of the i-th device in the building energy system, P i,t denotes the output power of the i-th device at time t, HSD0 denotes the heat storage state of the heat storage tank at the beginning of the scheduling period, HSD T denotes the heat storage state of the heat storage tank at the end of the scheduling period, TP0 denotes the hydrogen storage state of the hydrogen storage tank at the beginning of the scheduling period, TP T denotes the hydrogen storage state of the hydrogen storage tank at the end of the scheduling period, T denotes the scheduling period, DR i denotes the demand response degree reward of the i-th device.

[0038] In an embodiment of the present application, step S3 specifically comprises:

[0039] The state space is established by the following formula:

[0040]

[0041] wherein s t denotes the state space, and t denotes time;

[0042] The action space is established by the following formula:

[0043]

[0044] wherein α t denotes the action space;

[0045] The reward function is established by the following formula:

[0046]

[0047] wherein R represents a reward function, λ5 represents a fifth adjustable parameter, k1 represents a first weight parameter, k2 represents a second weight parameter, k3 represents a third weight parameter, C out,t represents an additional added penalty term, represents a thermal imbalance penalty, represents a hydrogen imbalance penalty, represents an electrical imbalance penalty.

[0048] A multi-energy flow scheduling system of a building energy system based on demand side response, comprising: a first establishing module, the first establishing module is used for establishing physical characteristic models and operation constraint conditions of each related device in the building energy system; a second establishing module, the second establishing module is used for establishing a target function based on the demand side response and taking balancing system operation cost and scheduling cost as the target; a third establishing module, the third establishing module combines the target function, the physical characteristic models of each related device and the operation constraint conditions, and constructs an adaptive deep deterministic policy gradient algorithm combined with a limit gradient boosting classification mechanism; an optimization scheduling module, the optimization scheduling module is used for classifying and processing load data, and training the adaptive deep deterministic policy gradient algorithm based on the classified and processed data to obtain a corresponding policy network model, and optimizing and scheduling multi-energy flow of the building energy system based on the policy network model.

[0049] The beneficial effects of the present application are:

[0050] The adaptive deep deterministic policy gradient algorithm combined with the limit gradient boosting classification mechanism can dynamically adjust the policy according to the demand side response type and system state change, improve the response ability in the complex building scene, and the system has high operation robustness. BRIEF DESCRIPTION OF DRAWINGS

[0051] Figure 1 A flow chart of a multi-energy flow scheduling method of a building energy system based on demand side response of an embodiment of the present application;

[0052] Figure 2 A structural schematic diagram of an electricity-hydrogen-heat low-carbon building energy system of one specific embodiment of the present application;

[0053] Figure 3 A sunny day light intensity and overcast day light intensity curve diagram of one specific embodiment of the present application;

[0054] Figure 4 A block schematic diagram of a multi-energy flow scheduling system of a building energy system based on demand side response of an embodiment of the present application. DETAILED DESCRIPTION

[0055] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the present application.

[0056] Figure 1 A flowchart of the multi-energy flow scheduling method of the demand-side response based building energy system according to an embodiment of the present application.

[0057] As shown in Figure 1 , the multi-energy flow scheduling method of the demand-side response based building energy system according to an embodiment of the present application can include the following steps:

[0058] S1, a physical characteristic model of each related device in the building energy system and a running constraint condition are established.

[0059] In an embodiment of the present application, the building energy system is an electricity-heat-hydrogen low-carbon building energy system, which includes a photovoltaic power generation device, an electrolyzer, a hydrogen storage tank, a fuel cell, a heat pump and a heat storage tank. Specifically, as shown in Figure 2 , the electrolyzer can be a PEM electrolyzer, the fuel cell can be a PEM fuel cell, the heat pump can be an air source heat pump, and the heat storage tank can be a steam heat storage tank.

[0060] In an embodiment of the present application, the physical characteristic model of the photovoltaic power generation device is:

[0061]

[0062] wherein, represents the photovoltaic power generation power, η PV represents the photovoltaic efficiency, η inv represents the inverter efficiency, A PV represents the effective illumination area of the photovoltaic power generation device, and I0represents the solar radiation intensity.

[0063] The electrolyzer is used for consumption of photovoltaic power generation during the day, and the physical characteristic model is:

[0064]

[0065] wherein, H ET is the hydrogen production rate, η F is the Faraday efficiency, n ET is the number of electrolyzers, I ET is the working current, and F is the Faraday constant.

[0066] The hydrogen storage tank is used for storing hydrogen produced by the electrolytic cell, and provides a hydrogen source for fuel cell power generation. It should be noted that in the present application, a high-pressure hydrogen storage tank can be used, and the tank pressure of the hydrogen storage tank dynamically changes with the storage and release of hydrogen. Correspondingly, the physical property model is:

[0067]

[0068] wherein TP t H represents the tank pressure of the hydrogen storage tank at time t, represents the tank pressure of the hydrogen storage tank at time t+1, represents the hydrogen flow rate of the hydrogen storage tank at time t, and R represents a universal gas constant, T H represents the working temperature of the hydrogen storage tank, V H represents the volume of the hydrogen storage tank, and z t represents the hydrogen compression coefficient corresponding to time t.

[0069] wherein the hydrogen compression coefficient z t corresponding to time t can be calculated based on NIST data and according to the Lemmon equation, that is,

[0070]

[0071] wherein a i , b i , and c i are constants, T0 represents a reference temperature, for example, T0 = 100 K, and R = 8.3145 J / mol·K. It should be noted that i represents the serial number in the Lemmon equation, and when the value of i changes, a i , b i , and c i also change accordingly. For example, the values of a i , b i , and c i may be as shown in Table 1:

[0072] Table 1

[0073]

[0074]

[0075] The physical property model of the fuel cell is:

[0076]

[0077] wherein H FC is the hydrogen consumption rate of the fuel cell, I FC is the working current of the fuel cell, and nFC is the number of cells, that is, the hydrogen consumption of the proton exchange membrane fuel cell can be calculated by the above formula based on Faraday's law; A cell is the reaction area of the cell unit, and j is the current density of the fuel cell, that is, the reaction rate of the fuel cell can be represented by the current density j of the fuel cell based on the above formula; is the electric power of the fuel cell, η T is the thermal efficiency of the fuel cell, η E is the electric efficiency of the fuel cell, is the thermal power of the fuel cell;

[0078] The coefficient of performance COP of the heat pump is related to temperature, and therefore, the physical property model can be:

[0079]

[0080] wherein COP represents the coefficient of performance of the heat pump, h1 represents a first fitting coefficient, h2 represents a second fitting coefficient, h3 represents a third fitting coefficient, ΔT HP represents a difference between the target temperature of the heat pump refrigeration / heat and the current ambient temperature, represents the thermal power of the heat pump, represents the electric power of the heat pump, COP H represents the heating performance coefficient, represents the refrigeration power of the heat pump, COP C represents the cooling performance coefficient;

[0081] The physical property model of the heat storage tank is:

[0082]

[0083] wherein HSD t+1 represents the heat storage state of the heat storage tank at t+1, HSD t represents the heat storage state of the heat storage tank at t, HSD t and HSD t+1 have a value range of [0, 1], represents the charging power or discharging power of the heat storage tank at t, is a negative value representing the charging power of the heat storage tank at t, is a positive value representing the discharging power of the heat storage tank at t, H c represents the maximum heat storage amount of the heat storage tank, and ΔT is a time step.

[0084] S2, based on demand side response and establishing an objective function aiming to balance system operation cost and scheduling cost.

[0085] In one embodiment of the present invention, the operation constraint conditions include: safe operation constraint and power balance constraint, wherein:

[0086] The safe operation constraints are:

[0087]

[0088]

[0089] TP min ≤TP t ≤TP max ,

[0090] in, represents the output power of the fuel cell at time t, and Represent the minimum and maximum output power of the fuel cell, represents the electrical power consumed by the heat pump at time t, and Represent the minimum and maximum values ​​of the heat pump power consumption, represents the electrolytic power of the electrolytic cell at time t, and Respectively represent the minimum and maximum values ​​of the electrolytic power of the electrolytic cell, represents the hydrogen consumption rate of the fuel cell at time t, and Respectively represent the minimum and maximum values ​​of the fuel cell hydrogen consumption rate, represents the hydrogen production rate of the electrolyzer at time t, and Respectively represent the minimum and maximum hydrogen production rates of the electrolyzer, HSD t Indicates the heat storage state of the heat storage tank at time t, HSD min and HSD max Respectively represent the minimum and maximum values ​​of the heat storage state of the heat storage tank, TP t Indicates the hydrogen storage state of the hydrogen storage tank at time t, TP min and TP max Respectively represent the minimum and maximum values ​​of the hydrogen storage state of the hydrogen storage tank;

[0091] The power balance constraint is:

[0092]

[0093]

[0094] in, represents the output power of the fuel cell at time t, represents the photovoltaic power generation power at time t, represents the electrical load of the building energy system at time t, represents the electricity consumption of the heat pump at time t, represents the electrolysis power of the electrolyzer at time t, represents the heat production of the fuel cell at time t, represents the heat production of the heat pump at time t, represents the heat discharge / charge of the thermal storage tank at time t, wherein, represents the heat discharge of the thermal storage tank at time t when the value is positive, represents the heat charge of the thermal storage tank at time t when the value is negative, represents the thermal load of the building energy system at time t, represents the hydrogen production rate of the electrolyzer at time t, represents the hydrogen flow rate of the hydrogen storage tank at time t, represents the hydrogen consumption rate of the fuel cell at time t, represents the hydrogen load of the building energy system at time t.

[0095] In an embodiment of the present application, the objective function is:

[0096]

[0097] wherein J represents the objective function, λ1 represents the first adjustable parameter, λ2 represents the second adjustable parameter, λ3 represents the third adjustable parameter, λ4 represents the fourth adjustable parameter, i represents the i-th device in the building energy system, OC i represents the unit operation cost of the i-th device in the building energy system, P i,t represents the output power of the i-th device at time t, HSD0 represents the thermal storage state of the thermal storage tank at the beginning of the scheduling period, HSD T represents the thermal storage state of the thermal storage tank at the end of the scheduling period, TP0 represents the hydrogen storage state of the hydrogen storage tank at the beginning of the scheduling period, TP T represents the hydrogen storage state of the hydrogen storage tank at the end of the scheduling period, T represents the scheduling period, DR i represents the demand response degree reward of the i-th device.

[0098] S3, in combination with the objective function, the physical characteristic model of each related device and the operation constraint condition, constructs an adaptive deep deterministic policy gradient algorithm fused with a limit gradient boosting classification mechanism.

[0099] In an embodiment of the present application, step S3 specifically comprises:

[0100] The state space is established by the following formula:

[0101]

[0102] wherein st S represents a state space, t represents time;

[0103] The action space is established by the following formula:

[0104]

[0105] Wherein, a t S represents an action space;

[0106] The reward function is established by the following formula:

[0107]

[0108] Wherein, R represents a reward function, λ5 represents a fifth adjustable parameter, k1 represents a first weight parameter, k2 represents a second weight parameter, k3 represents a third weight parameter, C out,t S represents an additional penalty term, S represents a thermal imbalance penalty, S represents a hydrogen imbalance penalty, S represents an electrical imbalance penalty.

[0109] S4, the load data is classified and processed, and the adaptive deep deterministic policy gradient algorithm is trained based on the classified and processed data to obtain a corresponding policy network model, and the multi-energy flow of the building energy system is optimized and dispatched based on the policy network model.

[0110] Specifically, the extreme gradient boosting is a classification method based on the idea of ensemble learning, which combines weak learners, iterative optimization and gradient descent. The weak learner usually uses a shallow decision tree (such as CART tree) as the basic classifier; the residual error is fitted step by step, that is, each tree learns the error of the previous round, that is, the gain is increased in the place where the model is insufficient; the gradient descent idea means that each iteration of the model is in the direction of minimizing the gradient of the loss function. At the beginning of training, a constant model is used for prediction, then a tree is trained in each iteration to fit the negative gradient of the loss function, the prediction results of multiple trees are added up to form the final classification model, and finally the softmax or sigmoid classification judgment is performed after the output category probability.

[0111] Specifically, first, data preparation is performed, such as Figure 3 As shown in the figure, two types of samples (i.e. load data, high and low light intensity data) are constructed according to typical daily data, that is, sunny light intensity curve (strong, sharp) and cloudy light intensity curve (weak, flat), wherein each sample is a 24-dimensional vector representing the irradiance every hour in a day.

[0112] Secondly, data shuffling and modeling are carried out, a Gradient Boosting Classifier is used to create a model, a fit() function is used to train the model, and the pattern of each type of lighting data is learned. The main parameters involved include: n_estimators, learning_rate, max_depth, min_samples_split, min_samples_leaf, subsample, max_features, loss and random_state, wherein n_estimators is used to set the number of weak learners (usually decision trees), which determines the complexity and training depth of the model; learning_rate is the weight of each tree on the final prediction result, and a smaller learning rate usually improves the robustness of the model, but more weak learners are needed; max_depth is used to control the maximum depth of a single tree, thereby affecting the fitting ability of the model to features and preventing overfitting; min_samples_split and min_samples_leaf limit the minimum number of samples for internal nodes and leaf nodes, respectively, which can effectively balance the complexity and generalization ability of the model; subsample represents the proportion of new samples used for training each iteration, and appropriately reducing this value can enhance the robustness of the model and reduce the risk of overfitting; max_features is used to limit the number of features considered for each node division, which helps to speed up training and improve model stability; in terms of loss function, the log loss function (loss='log_loss') is usually selected for classification tasks, which is suitable for binary and multi-classification scenarios; in order to ensure the reproducibility of the experiment, the random_state parameter is also commonly used to control the random behavior inside the model.

[0113] Finally, model prediction and visualization are carried out, the training set itself is used for prediction, PCA dimensionality reduction is used, and the distribution of samples in two-dimensional space is displayed for visualization.

[0114] Specifically, in one specific embodiment of the present application, the set values of the various main parameters involved can be as shown in Table 2.

[0115] Table 2

[0116] Main parameters Set values Main parameters Set values n_estimators 100 subsample 1.0 learning_rate 0.1 max_features None max_depth 3 loss 'log_loss' min_samples_split 2 random_state None min_samples_leaf 1

[0117] In one specific embodiment of the present application, the specification parameters and constraint configurations of each device in the electric-hydrogen-thermal low-carbon building energy system can be as shown in Table 3:

[0118] Table 3

[0119]

[0120]

[0121] Further, the main training parameters of the adaptive deep deterministic policy gradient algorithm are listed in detail in Table 4 as shown in the following table:

[0122] Table 4

[0123] Parameter Value Parameter Value Policy network learning rate 0.001～0.005 Training episodes 200 Critic network learning rate 0.001～0.005 Experience replay capacity 4000 Decay factor 0.99 Noise mean 0 Sampled sample capacity 96 Soft update coefficient 0.01 Noise variance 0.15 Noise minimum 0.001 Noise maximum 0.35

[0124] Since the adaptive deep deterministic policy gradient algorithm is trained using two types of sample data (high light intensity data and low light intensity data), some main training parameters of the adaptive deep deterministic policy gradient algorithm will change when different sample data are trained, as shown in Table 5:

[0125] Table 5

[0126]

[0127]

[0128] As shown in Table 4, the policy network learning rate and the evaluation network learning rate in the adaptive deep deterministic policy gradient algorithm of the present application are set to a range of values. During the training process, the algorithm automatically selects an appropriate learning rate for training according to the training situation. Compared with the traditional deep deterministic policy gradient algorithm, the present application can greatly improve the learning efficiency and adaptability of the scheduling policy and realize intelligent optimization scheduling of the multi-energy system. Moreover, as shown in Table 5, the present application uses an extreme gradient boosting classification mechanism to train the adaptive deep deterministic policy gradient algorithm based on two types of training data, i.e. high light intensity data and low light intensity data. This can effectively solve the problem of reduced scheduling precision caused by the compromise of the model in the fitting process due to the different load laws of sunny days and cloudy days when using the traditional training method. Therefore, the present application can improve the load data classification accuracy and enhance the generalization and decision-making ability of the model under complex working conditions.

[0129] After the training is completed, the policy network model is saved and simulation optimization scheduling is performed under a typical day scenario. Based on this, the comprehensive scheduling cost (covering system operation cost and sustainable operation cost) of the system within 24 hours, as well as energy consumption, carbon emissions and demand response savings, are obtained, as shown in Table 6:

[0130] Table 6 Comparison of scheduling results of different multi-energy flow scheduling strategies under a typical day scenario

[0131]

[0132] Therefore, the multi-energy flow scheduling strategy of the building energy system based on demand side response is realized, the physical characteristic model and the adaptive deep deterministic policy gradient algorithm are combined, the balance between the system operation cost and the scheduling efficiency is realized through the load classification and the limit gradient boosting mechanism, and the system flexibility and the intelligent scheduling capability are improved.

[0133] In conclusion, according to the multi-energy flow scheduling method of the building energy system based on demand side response, the physical characteristic model and the operation constraint condition of each related device in the building energy system are established, the target function is established based on the demand side response and the balance between the system operation cost and the scheduling cost, the adaptive deep deterministic policy gradient algorithm combined with the limit gradient boosting classification mechanism is constructed based on the target function, the physical characteristic model and the operation constraint condition of each related device, the load data is classified and processed, the adaptive deep deterministic policy gradient algorithm is trained based on the classified and processed data, the corresponding policy network model is acquired, and the multi-energy flow of the building energy system is optimized and scheduled based on the policy network model. Therefore, the robustness of the system operation can be greatly improved, the multi-energy flow is scheduled based on the low-carbon building energy system of electricity-hydrogen-heat, the cost loss of the battery is effectively reduced, and the operation cost of the system is reduced.

[0134] Corresponding to the above embodiment, the application further provides a multi-energy flow scheduling system of a building energy system based on demand side response.

[0135] As shown in Figure 4 The multi-energy flow scheduling system of the building energy system based on demand side response can include a first establishing module 100, a second establishing module 200, a third establishing module 300 and an optimization scheduling module 400.

[0136] The first establishing module 100 is used to establish the physical characteristic model and the operation constraint condition of each related device in the building energy system; the second establishing module 200 is used to establish the target function based on the demand side response and the balance between the system operation cost and the scheduling cost; the third establishing module 300 combines the target function, the physical characteristic model and the operation constraint condition of each related device, and constructs the adaptive deep deterministic policy gradient algorithm combined with the limit gradient boosting classification mechanism; and the optimization scheduling module 400 is used to classify and process the load data, train the adaptive deep deterministic policy gradient algorithm based on the classified and processed data, acquire the corresponding policy network model, and optimize and schedule the multi-energy flow of the building energy system based on the policy network model.

[0137] In an embodiment of the application, the building energy system is a low-carbon building energy system of electricity-heat-hydrogen, which includes photovoltaic power generation equipment, an electrolytic tank, a hydrogen storage tank, a fuel cell, a heat pump and a heat storage tank.

[0138] In an embodiment of the present application, the physical property model of the photovoltaic power generation device is:

[0139]

[0140] wherein, represents the photovoltaic power generation power, η PV represents the photovoltaic efficiency, η inv represents the inverter efficiency, A PV represents the effective illumination area of the photovoltaic power generation device, I0represents the solar radiation intensity;

[0141] The physical property model of the electrolytic tank for consumption of the photovoltaic power generation during the day is:

[0142]

[0143] wherein, H ET is the hydrogen production rate, η F is the Faraday efficiency, n ET is the number of electrolytic tanks, I ET is the working current, and F is the Faraday constant;

[0144] The physical property model of the hydrogen storage tank for storing the hydrogen produced by the electrolytic tank and providing a hydrogen source for fuel cell power generation is:

[0145]

[0146] wherein, TP t H represents the tank pressure of the hydrogen storage tank at time t, represents the tank pressure of the hydrogen storage tank at time t+1, represents the hydrogen flow rate of the hydrogen storage tank at time t, R represents the universal gas constant, T H represents the working temperature of the hydrogen storage tank, V H represents the volume of the hydrogen storage tank, z t represents the hydrogen compression coefficient corresponding to time t;

[0147] The physical property model of the fuel cell is:

[0148]

[0149] wherein, H FC is the hydrogen consumption rate of the fuel cell, I FC is the working current of the fuel cell, n FC is the number of cells, A cell is the reaction area of the cell unit, j is the current density of the fuel cell, is the electric power of the fuel cell, η Tη is the thermal efficiency of the fuel cell E η is the electrical efficiency of the fuel cell, η is the thermal power of the fuel cell;

[0150] The physical property model of the heat pump is:

[0151]

[0152] wherein COP represents the coefficient of performance of the heat pump, h1 represents the first fitting coefficient, h2 represents the second fitting coefficient, h3 represents the third fitting coefficient, ΔT HP represents the difference between the target temperature of the heat pump refrigeration / heat and the current environment temperature, represents the thermal power of the heat pump, represents the electrical power of the heat pump, COP H represents the heating performance coefficient, represents the refrigeration power of the heat pump, COP C represents the cooling performance coefficient;

[0153] The physical property model of the heat storage tank is:

[0154]

[0155] wherein HSD t+1 represents the heat storage state of the heat storage tank at t+1, HSD t represents the heat storage state of the heat storage tank at t, represents the charging power or discharging power of the heat storage tank at t, H c represents the maximum heat storage amount of the heat storage tank, and ΔT is the time step.

[0156] In an embodiment of the present application, the operation constraints include: a safe operation constraint and a power balance constraint, wherein,

[0157] The safe operation constraint is:

[0158]

[0159] TP min ≤ TP t ≤ TP max ,

[0160] wherein, represents the output power of the fuel cell at t, and respectively represent the minimum value and the maximum value of the output power of the fuel cell, represents the power consumption of the heat pump at t, and respectively represent the minimum value and the maximum value of the power consumption of the heat pump, represents the electrolysis power of the electrolyzer at time t, and respectively represent the minimum and maximum values of the electrolysis power of the electrolyzer, represents the hydrogen consumption rate of the fuel cell at time t, and respectively represent the minimum and maximum values of the hydrogen consumption rate of the fuel cell, represents the hydrogen production rate of the electrolyzer at time t, and respectively represent the minimum and maximum values of the hydrogen production rate of the electrolyzer, HSD t represents the heat storage state of the heat storage tank at time t, HSD min and HSD max respectively represent the minimum and maximum values of the heat storage state of the heat storage tank, TP t represents the hydrogen storage state of the hydrogen storage tank at time t, TP min and TP max respectively represent the minimum and maximum values of the hydrogen storage state of the hydrogen storage tank;

[0161] The power balance constraint is:

[0162]

[0163] wherein, represents the output power of the fuel cell at time t, represents the photovoltaic power generation at time t, represents the electrical load of the building energy system at time t, represents the electricity consumption of the heat pump at time t, represents the electrolysis power of the electrolyzer at time t, represents the heat production of the fuel cell at time t, represents the heat production of the heat pump at time t, represents the heat discharge / charge of the heat storage tank at time t, wherein, represents the heat discharge of the heat storage tank at time t when the value is positive, represents the heat charge of the heat storage tank at time t when the value is negative, represents the thermal load of the building energy system at time t, represents the hydrogen production rate of the electrolyzer at time t, represents the hydrogen flow rate of the hydrogen storage tank at time t, represents the hydrogen consumption rate of the fuel cell at time t, represents the hydrogen load of the building energy system at time t.

[0164] In an embodiment of the present application, the objective function is:

[0165]

[0166] wherein, J represents a target function, λ1 represents a first adjustable parameter, λ2 represents a second adjustable parameter, λ3 represents a third adjustable parameter, λ4 represents a fourth adjustable parameter, i represents an i-th device in the building energy system, OC i represents a unit operation cost of the i-th device in the building energy system, P i,t represents an output power of the i-th device at t time, HSD0 represents a heat storage state of the heat storage tank at the beginning of the scheduling period, HSD T represents a heat storage state of the heat storage tank at the end of the scheduling period, TP0 represents a hydrogen storage state of the hydrogen storage tank at the beginning of the scheduling period, TP T represents a hydrogen storage state of the hydrogen storage tank at the end of the scheduling period, T represents the scheduling period, DR i represents a demand response degree reward of the i-th device.

[0167] In an embodiment of the present application, the third establishing module 300 is specifically used for:

[0168] The state space is established by the following formula:

[0169]

[0170] wherein, s t represents the state space, and t represents time;

[0171] The action space is established by the following formula:

[0172]

[0173] wherein, α t represents the action space;

[0174] The reward function is established by the following formula:

[0175]

[0176] wherein, R represents the reward function, λ5 represents a fifth adjustable parameter, k1 represents a first weight parameter, k2 represents a second weight parameter, k3 represents a third weight parameter, C out,t represents an additional increased penalty term, represents a heat imbalance penalty, represents a hydrogen imbalance penalty, represents an electricity imbalance penalty.

[0177] It should be noted that the details not disclosed in the multi-energy flow scheduling system of the building energy system based on demand side response in the embodiments of the present application are referred to the details disclosed in the multi-energy flow scheduling method of the building energy system based on demand side response described above, and will not be described in detail here.

[0178] According to the multi-energy flow scheduling of the building energy system based on demand side response, the physical characteristic model and the operation constraint condition of each related device in the building energy system are established through the first establishing module, the target function is established based on the demand side response and taking balancing the system operation cost and the scheduling cost as the target through the second establishing module, the adaptive deep deterministic policy gradient algorithm combined with the target function, the physical characteristic model and the operation constraint condition of each related device is constructed through the third establishing module, the load data is classified and processed through the optimization scheduling module, the adaptive deep deterministic policy gradient algorithm is trained based on the classified and processed data to obtain the corresponding policy network model, and the multi-energy flow of the building energy system is optimized and scheduled based on the policy network model. Therefore, the adaptive deep deterministic policy gradient algorithm combined with the limit gradient boosting classification mechanism can dynamically adjust the policy according to the demand side response type and the system state change, improve the response ability in the complex building scene, and the operation robustness of the system is higher.

[0179] In the description of the present application, the terms "first", "second" are only for descriptive purpose, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more of the features. The meaning of "plurality" is two or more, unless otherwise specifically limited.

[0180] In the present application, unless otherwise specifically defined and limited, the terms "mounting", "connecting", "connecting", "fixing" and the like should be understood in a broad sense, for example, it can be fixed connection, or detachable connection, or integrated; it can be mechanical connection, or electrical connection; it can be directly connected, or indirectly connected through intermediate medium, or the internal communication of two elements or the interaction relationship between two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0181] In the present application, unless otherwise specifically defined and limited, the first feature is "on" or "under" the second feature, which can be direct contact between the first and second features, or indirect contact between the first and second features through intermediate medium. Moreover, the first feature "above", "over" and "on" the second feature can be directly above or obliquely above the first feature, or only indicate that the horizontal height of the first feature is higher than that of the second feature. The first feature "below", "under" and "under" the second feature can be directly below or obliquely below the first feature, or only indicate that the horizontal height of the first feature is less than that of the second feature.

[0182] In the description of the specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in one or more embodiments or examples. In addition, the person skilled in the art can combine and combine the different embodiments or examples described in the specification and the features of the different embodiments or examples without contradiction.

[0183] In addition, each functional unit in each embodiment of the present application can be integrated in one processing module, or each unit can exist physically, or two or more units can be integrated in one module. The integrated module can be realized in the form of hardware or in the form of a software functional module. The integrated module, if realized in the form of a software functional module and sold or used as an independent product, can also be stored in a computer readable storage medium.

[0184] Although the embodiments of the present application have been shown and described above, it should be understood that the above embodiments are exemplary and should not be construed as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application.

Claims

1. A multi-energy flow scheduling method for a building energy system based on demand-side response, characterized in that: The following steps are involved: S1, establishing a physical characteristic model and operation constraint conditions of each relevant equipment in the building energy system; S2, establishing an objective function based on the demand-side response and aiming to balance system operation cost and dispatch cost; S3, combining the objective function, the physical characteristic models of the relevant devices, and the operating constraints, to construct an adaptive deep deterministic policy gradient algorithm integrating an extreme gradient boosting classification mechanism; S4, classify and process the load data, and train the adaptive deep deterministic policy gradient algorithm based on the classified data to obtain a corresponding policy network model, and optimize the scheduling of the multi-energy flow of the building energy system based on the policy network model.

2. The multi-energy flow scheduling method for a building energy system based on demand-side response according to claim 1 is characterized in that: The building energy system is an electric-heat-hydrogen low-carbon building energy system, which includes: photovoltaic power generation equipment, electrolyzer, hydrogen storage tank, fuel cell, heat pump and heat storage tank.

3. The multi-energy flow scheduling method for a building energy system based on demand-side response according to claim 2 is characterized in that: The physical characteristic model of the photovoltaic power generation equipment is: in, represents photovoltaic power generation, η PV represents photovoltaic efficiency, η inv Indicates the inverter efficiency, A PV It represents the effective illumination area of ​​photovoltaic power generation equipment, and I0 represents the solar radiation intensity; The electrolyzer is used to absorb photovoltaic power generation during the day, and the physical characteristic model is: Among them, H ET is the hydrogen production rate, η F is the Faraday efficiency, n ET is the number of electrolytic cells, I ET is the working current, F is the Faraday constant; The hydrogen storage tank is used to store the hydrogen produced by the electrolyzer and provide a source of hydrogen for the fuel cell to generate electricity. The physical property model is: Among them, TP t H represents the pressure inside the hydrogen storage tank at time t, represents the pressure inside the hydrogen storage tank at time t+1, represents the hydrogen flow rate of the hydrogen storage tank at time t, R represents the universal gas constant, T H Indicates the operating temperature of the hydrogen storage tank, V H represents the volume of the hydrogen storage tank, z t represents the hydrogen compression coefficient corresponding to time t; The physical characteristic model of the fuel cell is: Among them, H FC is the hydrogen consumption rate of the fuel cell, I FC is the operating current of the fuel cell, n FC is the number of batteries, A cell is the reaction area of ​​the battery unit, j is the current density of the fuel cell, is the electrical power of the fuel cell, η T is the thermal efficiency of the fuel cell, η E is the electrical efficiency of the fuel cell, is the thermal power of the fuel cell; The physical characteristic model of the heat pump is: Wherein, COP represents the coefficient of performance of the heat pump, h1 represents the first fitting coefficient, h2 represents the second fitting coefficient, h3 represents the third fitting coefficient, ΔT HP Indicates the difference between the target temperature of the heat pump cooling / heating and the current ambient temperature. represents the thermal power of the heat pump, represents the electrical power of the heat pump, COP H represents the heating performance coefficient, Indicates the cooling power of the heat pump, COP C Indicates the coefficient of performance for cooling; The physical characteristic model of the heat storage tank is: Among them, HSD t+1 Indicates the heat storage state of the heat storage tank at time t+1, HSD t represents the heat storage state of the heat storage tank at time t, It represents the charging power or releasing power of the heat storage tank at time t, H c represents the maximum heat storage capacity of the heat storage tank, and ΔT is the time step.

4. The multi-energy flow scheduling method for a building energy system based on demand-side response according to claim 3 is characterized in that: The operation constraints include: safe operation constraints and power balance constraints, wherein: The safe operation constraints are: HSD min ≤HSD t ≤HSD max , TP min ≤TP t ≤TP max , in, represents the output power of the fuel cell at time t, and represent the minimum and maximum output power of the fuel cell, respectively, represents the electrical power consumed by the heat pump at time t, and represent the minimum and maximum values ​​of the heat pump power consumption, respectively. represents the electrolysis power of the electrolytic cell at time t, and Respectively represent the minimum and maximum values ​​of the electrolytic power of the electrolytic cell, represents the hydrogen consumption rate of the fuel cell at time t, and represent the minimum and maximum values ​​of the hydrogen consumption rate of the fuel cell, respectively, represents the hydrogen production rate of the electrolyzer at time t, and Respectively represent the minimum and maximum hydrogen production rates of the electrolyzer, HSD t Indicates the heat storage state of the heat storage tank at time t, HSD min and HSD max Respectively represent the minimum and maximum values ​​of the heat storage state of the heat storage tank, TP t Indicates the hydrogen storage state of the hydrogen storage tank at time t, TP min and TP max Respectively represent the minimum and maximum values ​​of the hydrogen storage state of the hydrogen storage tank; The power balance constraint is: in, represents the output power of the fuel cell at time t, represents the photovoltaic power generation power at time t, represents the electrical load of the building energy system at time t, represents the electrical power consumed by the heat pump at time t, represents the electrolysis power of the electrolytic cell at time t, represents the heat generation of the fuel cell at time t, represents the heat output of the heat pump at time t, It represents the heat released / charged by the heat storage tank at time t, where: When it is a positive value, it means the heat storage tank releases heat at time t. When it is a negative value, it means the heat storage tank is charged with heat at time t. represents the heat load of the building energy system at time t, represents the hydrogen production rate of the electrolyzer at time t, represents the hydrogen flow rate of the hydrogen storage tank at time t, represents the hydrogen consumption rate of the fuel cell at time t, represents the hydrogen load of the building energy system at time t.

5. The multi-energy flow scheduling method for a building energy system based on demand-side response according to claim 4 is characterized in that: The objective function is: Where J represents the objective function, λ1 represents the first adjustable parameter, λ2 represents the second adjustable parameter, λ3 represents the third adjustable parameter, λ4 represents the fourth adjustable parameter, i represents the i-th device in the building energy system, OC i represents the unit operating cost of the i-th device in the building energy system, P i,t represents the output power of the i-th device at time t, HSD0 represents the heat storage state of the heat storage tank at the beginning of the scheduling cycle, and HSD T It indicates the heat storage state of the heat storage tank at the end of the scheduling cycle, TP0 indicates the hydrogen storage state of the hydrogen storage tank at the beginning of the scheduling cycle, and TP T Indicates the hydrogen storage status of the hydrogen storage tank at the end of the scheduling period, T represents the scheduling period, DR i It represents the demand response degree reward of the i-th device.

6. The multi-energy flow scheduling method for a building energy system based on demand-side response according to claim 5 is characterized in that: Step S3 specifically includes: The state space is established by the following formula: Among them, s t represents the state space, t represents the time; The action space is established by the following formula: Among them, α t represents the action space; The reward function is established by the following formula: Among them, R represents the reward function, λ5 represents the fifth adjustable parameter, k1 represents the first weight parameter, k2 represents the second weight parameter, k3 represents the third weight parameter, and C out,t represents the additional penalty term, represents the thermal imbalance penalty, represents the hydrogen imbalance penalty, Indicates power imbalance penalty.

7. A multi-energy flow scheduling system for building energy systems based on demand-side response, characterized in that: include: A first establishing module, the first establishing module is used to establish a physical characteristic model and operation constraint conditions of each relevant device in the building energy system; a second establishing module, the second establishing module being configured to establish an objective function based on the demand-side response and with the goal of balancing system operation costs and dispatch costs; a third building module, wherein the third building module combines the objective function, the physical characteristic model of each relevant device, and the operating constraints to construct an adaptive deep deterministic policy gradient algorithm integrating an extreme gradient boosting classification mechanism; An optimization scheduling module is used to classify and process load data, and to train the adaptive deep deterministic policy gradient algorithm based on the classified data to obtain a corresponding policy network model, and to optimize and schedule the multi-energy flow of the building energy system based on the policy network model.

Images