Building demand response method based on flexible load response potential evaluation

Abstract

The invention discloses a building demand response method based on flexible load response potential evaluation. The building demand response method mainly comprises the following steps: collecting building user energy consumption data and internal and external related influence factors; constructing a space-time diagram convolutional neural network, effectively capturing time sequence dependence characteristics in a sequence by self-correcting diagram node weights, and realizing real-time load decomposition of energy consumption behaviors of building users; the influence of internal and external multi-dimensional environmental factors is considered, and the main flexible load response potential of the user is evaluated; and finally, constructing a system operator-load aggregator double-level price excitation type demand response model. The invention provides a building demand response method based on flexible load response potential evaluation, which can realize building user demand response strategy formulation considering flexible load response characteristic evaluation.

Description

Technical Field

[0001] The present invention belongs to the technical field of power demand response, and in particular relates to a building demand response method based on flexible load response potential assessment. Background Art

[0002] With the continuous growth of electricity demand and the widespread application of renewable energy, the challenges of load regulation and stability faced by the power grid are becoming increasingly prominent. As one of the important strategies to maintain the balance between supply and demand of smart grids, building demand response aims to achieve the balance and optimization of power grid load by optimizing energy consumption inside buildings. Among them, load aggregators issue price incentives to stimulate customers with flexible loads, thereby encouraging price responsiveness and demand flexibility. Through this means, it can not only promote customers to change their load usage structure and provide economic benefits to customers, but also enhance the reliability of the system and improve market efficiency.

[0003] At present, the demand response methods based on price incentives mainly include peak-valley price difference, time-varying electricity price and real-time electricity price. Among them, the peak-valley price difference is the simplest and easy to be real-time, but its disadvantage is that it may cause users to overuse electricity during low-price periods, increase the load during the valley period, and lack the flexibility of regulation. On this basis, the time-varying electricity price sets the electricity price for different time periods to encourage users to use electricity during periods with lower electricity prices, thereby reducing the load of the power grid during peak hours. Although the flexibility of response has been improved to a certain extent, it cannot be fully applied to all users or emergencies in the network. Therefore, some scholars have proposed the concept of real-time electricity price, which is adjusted according to the supply and demand dynamics of the real-time electricity market, provides more accurate price signals, and encourages users to adjust their electricity consumption according to real-time prices. However, a reasonable real-time electricity price requires real-time monitoring of users' complex energy consumption behaviors and timely adjustment of price incentives. Existing research mainly focuses on assuming users' fixed flexible load energy consumption habits and implementing modeling analysis, and lacks research on demand response strategies based on real-time decomposition of users' energy consumption behaviors. Summary of the invention

[0004] The technical problem to be solved by the present invention is to realize the formulation of building user demand response strategy considering the evaluation of flexible load response characteristics.

[0005] To achieve the above purpose, the present invention adopts the following technical solution: the method involved in the present invention will construct a spatiotemporal graph convolutional network to realize the real-time decomposition of user flexible loads, and evaluate the adjustable potential of temperature control flexible loads in combination with the influencing factors inside and outside the building, and finally construct a two-level demand response model to optimize the price incentive strategy of demand response. Its implementation steps include:

[0006] (1) Collection of load data and internal and external influencing factors of smart buildings: Use smart energy meters to collect and store user power consumption in real time, while also recording external temperature and humidity data and holiday information;

[0007] (2) Data preprocessing: Use the maximum and minimum value normalization operation to preprocess the load data and internal and external influencing factors;

[0008] (3) Build and train a spatiotemporal graph convolutional network model, which includes an input layer, two graph convolutional layers, two temporal convolutional layers, a dropout layer, a fully connected layer, and an output layer. The spatiotemporal graph convolutional network model uses a mean square error loss function to calculate the training effect and perform back propagation and parameter updates to achieve real-time user load decomposition.

[0009] (4) Based on the internal and external temperature data of the building, the temperature response slope is calculated using the change point heat balance model to evaluate the adjustable potential of the temperature control flexible load;

[0010] (5) Taking into account the optimal economic benefits for users and the lowest operating costs for aggregators, a price-incentive multi-objective demand response model is constructed.

[0011] Compared with the existing technology, the beneficial effects are: combining the real-time decomposition of user-side energy consumption behavior, flexible load regulation potential evaluation and two-level demand response strategy, thereby improving the accuracy of demand response, and having certain feasibility and effectiveness; compared with traditional demand response methods, the method proposed in the present invention taps the real-time regulation potential of user flexible loads and has obvious advantages. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A work flow chart of a building demand response method based on flexible load response potential assessment of the present invention;

[0013] Figure 2 This is the structure diagram of the spatiotemporal graph convolutional network proposed in the present invention. DETAILED DESCRIPTION

[0014] The following is a detailed description of a specific implementation of a building demand response method based on flexible load response potential assessment according to the present invention in conjunction with the accompanying drawings.

[0015] like Figure 1 As shown, the building demand response method based on flexible load response potential assessment of the present invention comprises the following steps:

[0016] Step 1: Collecting load data and internal and external influencing factors of smart buildings: Using smart electric energy meters to collect the energy consumption of household users in the building, and collecting meteorological data and holiday data in the same period;

[0017] Smart electricity meters are used to collect energy consumption data of flexible loads in independently operated buildings, mainly including air conditioners, HVAC, heaters, etc. The sampling interval is 1 minute. At the same time, the total energy consumption of building users and indoor and outdoor temperature and humidity are recorded at the same sampling rate.

[0018] Step 2: Data preprocessing: Use the maximum and minimum value normalization operation to preprocess the load data and internal and external influencing factor data.

[0019] Step 3: Build and train a spatiotemporal graph convolutional network model, which includes an input layer, two graph convolutional layers, two temporal convolutional layers, a dropout layer, a fully connected layer, and an output layer. The spatiotemporal graph convolutional network model uses a mean square error loss function to calculate the training effect and perform back propagation and parameter update, thereby realizing real-time user load decomposition.

[0020] The network structure of the spatiotemporal graph convolutional network model is shown in the following table. The current flexible load energy level is cyclically decomposed in units of 30 moments. The network input dimension is 30×5, including user electricity consumption data, indoor temperature data, outdoor temperature data, outdoor humidity data, and holiday data within 30 minutes. The spatial convolution layer extracts the correlation between different moments, and the temporal convolution is used to mine the implicit features within the sequence. The final network output is the flexible load energy consumption at the current moment, with a dimension of 1×1. Among them, the feature extraction method of the graph convolutional network is as follows:

[0021]

[0022] In the formula, H (l) is the node feature matrix of the l-th layer graph convolutional network; Represents the adjacency matrix, which characterizes the association relationship between nodes. This patent adopts a dynamic adjacency matrix to adjust the weights to adapt to the training number through model training and parameter update; is a diagonal matrix where each diagonal element D ii represents the number of connections of node i; W (l) Represents the node feature parameter matrix, which is obtained and corrected through model training; the overall network structure is as follows Figure 2 shown.

[0023] Step 4: Based on the energy decomposition results, the change point heat balance model is used to calculate the demand response potential within the building users, as shown below:

[0024]

[0025] In the formula, represents the power of the thermal control load of the nth user at time t, Represents the set temperature of the user's temperature control load at that moment, represents the outside temperature, Represents the temperature response slope of the user, ε n Represents the measurement noise. The user’s and It is also an important indicator for evaluating the user's thermal load demand response capability. The demand response potential evaluation method is as follows:

[0026]

[0027] in represents the demand response potential of the nth user at time t, It indicates the set temperature of the temperature control load at the lowest indoor temperature, and max{} indicates the maximum value calculation.

[0028] Step 5: Construct a two-tier demand response model of system operator and load aggregator, where the elasticity of user flexible load to price response can be defined as:

[0029]

[0030] in, represents the user's response power, represents the incentive electricity price for the demand response strategy, and Indicates normal electricity price.

[0031] The optimal operating cost model of the system operator can be expressed as:

[0032]

[0033] Among them, F SC represents the operating cost function, T represents the optimization time length, G represents the number of generators, H represents the number of building users, represents the generator's electricity selling price, represents the output power of the generator, represents the market electricity price, represents the electric power demanded by the market, represents the reserve capacity of the generator, and Represents the generator's reserve capacity bid. represents the equivalent reserved capacity of flexible loads participating in demand response, Represents a demand response incentive bid for the reserved capacity of flexible loads. It reduces the operating costs of load aggregators by minimizing the reserved regulation capacity of generators and user flexible loads while ensuring the stability of power network operation.

[0034] Load aggregator maximum profit model:

[0035]

[0036] Among them, F AR represents the profit of the load aggregator, represents the retail price of electricity at time t, NT represents the total time period of the model, NH represents the total number of users, represents the total electricity sold, represents the equivalent flexible load reserved capacity of building users participating in demand response, represents the user's demand response incentive electricity price, represents the power purchased from the electricity market, and Indicates the electricity purchase price.

[0037] User comfort constraints:

[0038]

[0039] in and They are the upper and lower limits of air conditioning operation. Indicates the indoor temperature of building users.

[0040] The supply and demand balance constraint is:

[0041]

[0042] Where NG represents the total number of generators, Indicates the output power of the generator, Indicates the market demand power.

[0043] Electricity sales price constraints:

[0044]

[0045] in represents the retail electricity price of the aggregator at time t, λ RET,min and λ RET,max are the upper and lower bounds of the selling price respectively.

[0046] Node voltage constraints:

[0047] V min ≤V t ≤V max (10)

[0048] V t Represents the node voltage, V min and V max are the upper and lower bounds of the voltage respectively.

[0049] By using CPLEX commercial solving software to solve the constraints and optimization objectives, we can obtain the optimal electricity price incentive strategy for demand response, load aggregator revenue, and system operator cost, thereby assisting in the formulation of building demand response strategies.

[0050] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention and are not intended to limit the present invention. A person skilled in the art should understand that the specific implementation of the present invention can be modified or replaced by equivalents, but these modifications or changes are within the scope of protection of the pending claims.

Claims

1. A building demand response method based on flexible load response potential assessment, characterized in that: The building demand response method comprises the following steps: Step 1: Collect building user energy consumption data and related internal and external influencing factors: Use smart electricity meters to collect energy consumption of building household users, and collect meteorological data and holiday data in the same period; Step 2: Data preprocessing: Use the maximum and minimum value normalization operation to preprocess the load and internal and external influencing factor data; Step 3: Build and train a spatiotemporal graph convolutional network model, which includes an input layer, two graph convolutional layers, two temporal convolutional layers, a dropout layer, a fully connected layer, and an output layer. The spatiotemporal graph convolutional network model uses a mean square error loss function to calculate the training effect and perform back propagation and parameter update, thereby realizing real-time user load decomposition. Step 4: Based on the internal and external temperature data of the building, the temperature response slope is calculated using the change point heat balance model to evaluate the adjustable potential of the temperature control flexible load; Step 5: Considering the optimal economic benefits for users and the lowest operating costs for aggregators, a price-incentive multi-objective demand response model is constructed.

2. The building demand response method based on flexible load response potential assessment according to claim 1 is characterized in that: The feature extraction method of the spatiotemporal graph convolutional network in step 3 is: In the formula, H (l) is the node feature matrix of the l-th layer graph convolutional network; Represents the adjacency matrix, characterizing the association relationship between nodes. A dynamic adjacency matrix is ​​used to adjust the weights to adapt to the training number through model training and parameter update. is a diagonal matrix where each diagonal element D ii represents the number of connections of node i; W (l) Represents the node feature parameter matrix, which is obtained and corrected through model training.

3. The building demand response method based on flexible load response potential assessment according to claim 2 is characterized in that: In step 4, the specific method of calculating the temperature response slope using the change point heat balance model to evaluate the adjustable potential of the temperature control flexible load is: In the formula, represents the power of the thermal control load of the nth user at time t, Represents the set temperature of the user's temperature control load at that moment, T t O represents the outside temperature, Represents the temperature response slope of the user, ε n represents the measurement noise, max{} represents the maximum value calculation; the user's and It is also used as an important indicator for evaluating the user's thermal control load demand response capability; the demand response potential evaluation method is as follows: in represents the demand response potential of the nth user at time t, T t min It indicates the temperature control load setting temperature at the lowest indoor temperature.

Images