Virtual power plant scheduling method, system and device considering electricity price and storage medium

Abstract

The invention relates to the technical field of data processing, and particularly provides a virtual power plant scheduling method, system and device considering electricity price, and a storage medium, and the method comprises the steps: collecting multi-source monitoring data at regular intervals, the multi-source monitoring data comprising electricity price data, load data, meteorological data and power data; constructing a space-time tensor according to the acquisition time and the monitoring position of the multi-source monitoring data; inputting the space-time tensor into a prediction model to obtain probability distribution of load demand, distributed power supply output and market electricity price in a set time span; and solving an optimal scheduling scheme according to the probability distribution, the established minimum economic cost objective function and the minimum load fluctuation objective function. According to the method, the space-time tensor is constructed by fusing the multi-source data such as the electricity price, the load and the weather, the prediction precision is remarkably improved, and the problem of scheduling deviation caused by dependence of single data in a traditional method is solved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of data processing, and particularly relates to a virtual power plant scheduling method, system, device and storage medium considering electricity price. BACKGROUND

[0002] With the acceleration of energy structure transformation, the virtual power plant (VPP) becomes a key carrier for integrating distributed resources. The traditional scheduling method has significant defects: first, most models use deterministic electricity price prediction, which is difficult to adapt to the real-time volatility of the electricity market, resulting in economic optimization failure; second, the strong randomness of load and distributed power output causes scheduling deviation, and the existing technology relies on single historical data statistics without fusing spatiotemporal correlation factors such as weather; third, the objective function often considers economic cost or load stability in isolation, lacking a multi-objective collaborative optimization mechanism. SUMMARY

[0003] In view of the above deficiencies of the prior art, the application provides a virtual power plant scheduling method, system, device and storage medium considering electricity price to solve the above technical problems.

[0004] In a first aspect, the application provides a virtual power plant scheduling method considering electricity price, comprising: periodically collecting multi-source monitoring data, the multi-source monitoring data including electricity price data, load data, weather data and power source data; constructing a spatiotemporal tensor according to the collection time and monitoring position of the multi-source monitoring data; inputting the spatiotemporal tensor into a prediction model to obtain the probability distribution of load demand, distributed power output and market electricity price within a set time span; solving an optimal scheduling scheme according to the probability distribution, and a constructed minimum economic cost objective function and minimum load fluctuation objective function.

[0005] In an optional implementation, the spatiotemporal tensor is constructed according to the collection time and monitoring position of the multi-source monitoring data, comprising: constructing an adjacency matrix of monitoring nodes according to the power grid topology; generating a time series feature matrix for the multi-source monitoring data using a time sliding window mechanism; integrating the time series feature matrix and the adjacency matrix into a spatiotemporal tensor.

[0006] In an optional implementation, the adjacency matrix of monitoring nodes is constructed according to the power grid topology, comprising: determining the edge weight between corresponding nodes in the power grid topology according to the electrical distance and load correlation between the nodes; constructing an initial adjacency matrix according to the number of nodes of the power grid topology; According to the edge weight between nodes in the power grid topology, the corresponding element in the initial connection matrix is filled, and an adjacency matrix is obtained after filling.

[0007] In an optional implementation, the prediction model comprises: a time feature extraction layer for capturing local time features of a spatio-temporal tensor through a ConvLSTM network; a spatial feature aggregation layer for obtaining spatial enhanced features by performing graph convolution on the local time features of each time step and the adjacency matrix of the spatio-temporal tensor; an output layer for decomposing the spatial enhanced features into a feature sequence of each node, and for each node, the feature sequence is input into a hybrid density network, which outputs probability distribution parameters of load demand, power generation output and market electricity price based on a time step-by-step manner; the hybrid density network uses a Gaussian mixture model to generate a conditional probability distribution for each output variable, including a mixing coefficient, a mean and a variance; each output variable is processed through an independent output head to model the correlation between variables.

[0008] In an optional implementation, the method further comprises: constructing a first objective function for minimizing economic cost and a second objective function for minimizing load fluctuation, respectively; the first objective function comprises:

[0009] wherein, is the electricity purchase price of the t period; is the electricity purchase quantity of the t period, and the calculation method is total load plus charging and discharging power of the energy storage system minus power generation minus power reduction; is the demand response subsidy of the t period; is the total reduction power; the second objective function comprises:

[0010] represents the power change amount of the adjacent period; represents the maximum power change rate allowed; represents the out-of-limit penalty strength; T represents the monitoring period length; a weighted sum of the first objective function and the second objective function is taken as a comprehensive objective function.

[0011] In an optional implementation, the method further comprises: constructing a constraint function system, the constraint function system comprising a power balance constraint, an adjustable load constraint, an energy storage constraint and a power generation power constraint.

[0012] In an optional embodiment, the optimal scheduling scheme is solved according to the probability distribution, and a constructed minimum economic cost objective function and a minimum load fluctuation objective function, including: an optimal particle position vector is solved according to the comprehensive objective function and the probability distribution by using a quantum particle swarm optimization algorithm; the particle position vector is discretely converted into an economic pheromone map and a safety pheromone map; an economic optimal solution is solved according to the economic pheromone map and the first objective function by using a first ant colony in the adaptive dual ant colony algorithm; a safety optimal solution is solved according to the safety pheromone map and the second objective function by using a second ant colony in the adaptive dual ant colony algorithm; the economic optimal solution and the safety optimal solution are integrated to obtain the optimal scheduling scheme.

[0013] In a second aspect, the present application provides a virtual power plant scheduling system considering electricity price, including: a collection module, configured to periodically collect multi-source monitoring data, the multi-source monitoring data including electricity price data, load data, weather data and power source data; a preprocessing module, configured to construct a space-time tensor according to the collection time and monitoring position of the multi-source monitoring data; a prediction module, configured to input the space-time tensor into a prediction model to obtain the probability distribution of load demand, distributed power output and market electricity price within a set time span; a solving module, configured to solve an optimal scheduling scheme according to the probability distribution, and a constructed minimum economic cost objective function and a minimum load fluctuation objective function.

[0014] In a third aspect, a device is provided, including: a memory, configured to store a virtual power plant scheduling program considering electricity price; a processor, configured to implement the steps of the virtual power plant scheduling method considering electricity price provided in the first aspect when the virtual power plant scheduling program considering electricity price is executed.

[0015] In a fourth aspect, a computer readable storage medium is provided, and the storage medium stores a virtual power plant scheduling program considering electricity price, which implements the steps of the virtual power plant scheduling method considering electricity price provided in the first aspect when the virtual power plant scheduling program considering electricity price is executed by a processor.

[0016] The application has the beneficial effects that the virtual power plant scheduling method, system, device and storage medium considering electricity price are provided, time-space tensors are constructed by fusing multi-source data such as electricity price, load and weather, the prediction accuracy is significantly improved, and the scheduling deviation problem caused by single data dependence in the traditional method is solved; the probability distribution prediction model is adopted, the dynamic fluctuation characteristics of the power market are effectively captured, and the electricity price sensitivity and economic optimization capability are enhanced; the minimum economic cost and load fluctuation double target are innovatively optimized, the operation cost is reduced, and the power grid stability is ensured. In addition, the introduction of the time-space correlation analysis technology quantifies the coupling relationship between the distributed power output and the weather factors, further improves the robustness and adaptability of the scheduling scheme, and provides more scientific technical support for the virtual power plant participating in the power market. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, for those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative labor.

[0018] Figure 1 is a schematic flow chart of the method of one embodiment of the present application.

[0019] Figure 2 is a schematic block diagram of the system of one embodiment of the present application.

[0020] Figure 3 is a structural schematic diagram of the equipment provided by the embodiment of the present application. DETAILED DESCRIPTION

[0021] In order to make the person skilled in the art better understand the technical solutions in the present application, the technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings of the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor should belong to the scope of protection of the present application.

[0022] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs. The terms used in the specification of the present application are only for the purpose of describing the specific embodiments and are not intended to limit the present application.

[0023] The method for scheduling a virtual power plant considering electricity price provided by the embodiment of the application is executed by a computer device, and accordingly, the virtual power plant scheduling system considering electricity price runs in the computer device.

[0024] Figure 1 is a schematic flow chart of the method of one embodiment of the application. In the flow chart, Figure 1 The execution subject can be a virtual power plant scheduling system considering electricity price. The order of steps in the flow chart can be changed, and some steps can be omitted according to different requirements.

[0025] As Figure 1 shown, the method comprises the following steps: S1. periodically collecting multi-source monitoring data, the multi-source monitoring data comprising electricity price data, load data, weather data and power source data; S2. constructing a space-time tensor according to the collection time and monitoring position of the multi-source monitoring data; S3. inputting the space-time tensor into a prediction model to obtain the probability distribution of load demand, distributed power output and market electricity price within a set time span; S4. solving an optimal scheduling scheme according to the probability distribution, a constructed minimum economic cost objective function and a minimum load fluctuation objective function.

[0026] In one embodiment of the application, based on step S1, a possible embodiment will be given below to non-restrictively describe the specific implementation scheme.

[0027] Real-time electricity price data: obtained from a power market (such as PJM) through an API, and the data format is (timestamp, price); Load data: collected from a smart meter, containing (timestamp, node_id, active_power, reactive_power); Weather data: obtained through a weather API, including (timestamp, temperature, irradiance, humidity, wind_speed); Distributed power source data: obtained through a SCADA system, containing (timestamp, device_id, pv_power, wind_power, battery_soc).

[0028] Time alignment: all data are unified to the same timestamp sequence (such as 15-minute intervals), and small missing sections are filled by using linear interpolation.

[0029] Outlier processing: the 3σ principle is used to detect outliers for load and power generation data, and the average value of adjacent time periods is used to replace the outliers.

[0030] Normalization: Perform maximum and minimum normalization on each feature.

[0031] The distributed time series database InfluxDB is used to build a data storage platform. With timestamps as index keys, multi-source data is partitioned and stored according to the three-level directory of "data source-region-time". Data compression algorithms (such as LZ4) are used to reduce storage costs. At the same time, a tiered storage strategy for hot and cold data is established to improve data query efficiency.

[0032] In an embodiment of the present invention, based on step S2, a possible embodiment will be given below to illustrate its specific implementation scheme in a non-limiting manner.

[0033] S201. Construct an adjacency matrix of monitoring nodes according to the power grid topology.

[0034] (1) Determine the edge weights between corresponding nodes in the power grid topology based on the electrical distance and load correlation between the nodes.

[0035] Electrical distance calculation: Using the Thevenin equivalent principle, the electrical distance d between nodes i and j in the power grid is calculated. ij Defined as the Thevenin equivalent impedance Z between two nodes eq The modulus value, d ij = |Z eq In the specific calculation, the node voltage and branch current data of the power grid under normal operation are obtained through power flow calculation. The network simplification algorithm is used to gradually eliminate the non-i and non-j nodes to obtain the equivalent impedance between the two nodes.

[0036] Load correlation analysis: Use the Pearson correlation coefficient to calculate the load correlation r between node i and node j ij Collect the active power P of two nodes for a period of time (such as one sampling point every 15 minutes for one week) i (t) and P j (t), which is calculated as follows:

[0037] Among them, P i and P j are the average active power of node i and node j respectively, and n is the number of sampling points.

[0038] Comprehensive weight determination: In order to balance the impact of electrical distance and load correlation on edge weight, the edge weight wij between node i and node j is determined by weighted summation, that is: w ij =αd ij +β(1−r ij ) Wherein, a and β are weight coefficients, and a + β = 1. Through simulation experiments or expert experience, a and β can be optimized.

[0039] (2) Construct an initial adjacency matrix according to the number of nodes of the power grid topology.

[0040] According to the number of nodes N of the power grid topology, an N×N two-dimensional zero matrix is constructed as an initial adjacency matrix A0. The rows and columns of the matrix correspond to the N nodes in the power grid, and the elements A0(i,j) in the matrix represent the initial connection state between node i and node j. Before filling the edge weight, all elements are 0, that is, A0(i,j)=0,∀i,j∈{1,2,⋯,N}.

[0041] (3) According to the edge weight between nodes in the power grid topology, fill in the corresponding elements in the initial connection matrix, and obtain the adjacency matrix after filling.

[0042] Traverse all node pairs (i,j) in the power grid topology. If there is a physical connection between node i and node j, the calculated edge weight w ij is assigned to the corresponding elements A0(i,j) and A0(j,i) in the initial adjacency matrix A0 (considering the undirected nature of the power grid topology), that is, A(i,j)=A(j,i)=w ij ; If there is no physical connection between the two nodes, the corresponding elements remain 0. After the above filling process, the adjacency matrix A is finally obtained, which fully describes the connection relationship and the correlation strength between nodes in the power grid topology.

[0043] S202. Generate a time series feature matrix for multi-source monitoring data using a time sliding window mechanism.

[0044] A fixed-length sliding window mechanism is used, with a window length of T sampling periods and a sliding step of S periods (S≤T). For a time series data x(t) of a certain data source, in the nth window, the data segment can be represented as x n =[x(nT−T+1),x(nT−T+2),⋯,x(nT)].

[0045] Basic statistical features: Calculate the mean μ, standard deviation σ, maximum value max, minimum value min, and kurtosis kurtosis of the data in each sliding window. Trend features: Linear trend fitting of the data in the window is performed using the least squares method, and the slope and intercept are extracted as trend features; the first-order difference and second-order difference of adjacent data points are calculated to represent the data change rate and acceleration. Frequency domain features: Discrete Fourier transform (DFT) is performed on the window data, and the top N main frequency components with the largest energy proportion are selected as the frequency domain features.

[0046] After feature extraction of multi-source data, the data is horizontally spliced according to the type of data source. The feature matrix of all time windows is stacked in time sequence to form a time series feature matrix set:

[0047] Wherein, N represents a node, and T represents a time step.

[0048] S203. The time series feature matrix is integrated with the adjacency matrix to form a space-time tensor.

[0049] The integrated data structure is:

[0050] The tensor product of the time series feature matrix and the adjacency matrix is calculated to obtain a space-time tensor.

[0051] In an embodiment of the present application, based on step S3, a possible embodiment will be given below to non-restrictively describe the specific implementation scheme.

[0052] The prediction model comprises: The time feature extraction layer captures the local time features of the space-time tensor through the ConvLSTM network; the time feature extraction layer adopts the ConvLSTM network, and the structure of the ConvLSTM network is specially designed for capturing dynamic features of space-time data. The network is stacked by multiple ConvLSTM units, and each unit contains a convolution gating mechanism and a recurrent memory unit. When processing the space-time tensor, the network slides along the time dimension to extract features in a local time window. Through convolution operation, the local time dependence in time series data can be effectively extracted, such as the intra-day fluctuation rule of load curve and the time-of-use trend of electricity price. At the same time, the memory unit of LSTM is responsible for storing the feature information of historical time, so that the network can capture the dynamic changes of long time scale, such as week cycle, seasonal cycle and other time features. The network parameters are optimized by the back propagation algorithm to minimize the error between the prediction result and the true value, thereby enhancing the adaptability and generalization ability of the network to time series data.

[0053] The spatial feature aggregation layer obtains spatial enhanced features by performing graph convolution on the local time features of each time step and the adjacency matrix of the spatio-temporal tensor. The spatial feature aggregation layer is based on a graph convolution network (GCN) and combines the adjacency matrix in the spatio-temporal tensor and the local time features output by the time feature extraction layer to achieve feature enhancement in the spatial dimension. Specifically, for the local time features of each time step, the local time features are input as node feature of the graph convolution network, and the connection relationship between nodes is defined by the power grid topology adjacency matrix. The graph convolution operation fully excavates the spatial correlation between nodes in the power grid, such as the mutual influence of node loads and the linkage change of regional electricity prices, by aggregating the feature information of adjacent nodes. In the convolution process, a multi-layer graph convolution structure is adopted, and each layer transforms the node features by using different convolution kernel parameters to gradually extract higher-level spatial feature representations. After the multi-layer graph convolution processing, the output spatial enhanced features not only contain the local information of each node, but also fuse the spatial topological structure information of the whole network, providing richer feature representation for subsequent prediction tasks.

[0054] The output layer is configured to decompose the spatial enhanced features into a feature sequence of each node. For each node, the feature sequence is input into a hybrid density network, which outputs probability distribution parameters of load demand, power generation output and market electricity price based on a time step-by-step manner. The hybrid density network uses a Gaussian mixture model to generate a conditional probability distribution for each output variable, including a mixing coefficient, a mean and a variance. Each output variable is processed by an independent output head to model the correlation between variables. The output layer first decomposes the spatial enhanced features obtained by the spatial feature aggregation layer according to the nodes of the power grid to obtain a feature sequence corresponding to each node. In order to facilitate the processing of the Gaussian mixture model (GMM), a flattening operation is performed on the feature sequence of each node to convert the multi-dimensional feature vector into a one-dimensional vector. The GMM fits the distribution mode of the feature sequence to a mixed form of multiple Gaussian distributions by learning. Each Gaussian distribution corresponds to a set of mean, covariance and weight parameters, respectively representing the central tendency, dispersion and mixing proportion of the data. Specifically, the output of the prediction model includes: Main output: a three-dimensional probability distribution tensor p (500x96x3): Three-dimensional tensor storage structure "spatial dimension (node)": 500 "time dimension (time period)": 96 "feature dimension (3 types x 5 parameters)": 15.

[0055] Explanatory output: example matrix for a specific node:

[0056] The essence of the explanatory output is a slice display of the three-dimensional tensor p in the node and time dimensions.

[0057] In the training stage, the expectation maximization (EM) algorithm is used to iteratively optimize the parameters of the GMM to maximize the likelihood function of the training data. Finally, the model outputs the probability distribution of load demand, power generation and market electricity price within a certain time span through the GMM, providing comprehensive uncertainty information for power system operation, supporting risk assessment and optimization decision.

[0058] When the spatio-temporal tensor is input, the prediction model processes the data in the order of time feature extraction layer, spatial feature aggregation layer and output layer, and finally generates the probability distribution prediction result of each variable within the future time span through the layer-by-layer feature transformation and information fusion of the network, effectively reflecting the dynamic changes and uncertainties in the operation of the power system.

[0059] In an embodiment of the present application, based on step S4, a possible embodiment will be given below to non-restrictively illustrate the specific implementation thereof.

[0060] A first objective function for minimizing economic cost and a second objective function for minimizing load fluctuation are constructed respectively; The first objective function includes:

[0061] wherein, is the electricity purchase price at t period; is the electricity purchase quantity at t period, and the calculation method is total load plus charging and discharging power of the energy storage system minus power generation minus power reduction; is the demand response subsidy at t period; is the total reduction power; The second objective function includes:

[0062] represents the power change amount of adjacent periods; represents the maximum power change rate allowed; represents the out-of-limit penalty strength; T represents the monitoring period length; The weighted sum of the first objective function and the second objective function is taken as the comprehensive objective function.

[0063] A constraint function system is constructed: Power balance constraint:

[0064] wherein, is the electricity purchase quantity at t period; is the total reduction power; is the energy storage charging and discharging power at t time; is the power generation at t time, i.e. the distributed power output.

[0065] Adjustable load constraints:

[0066] represents the maximum allowed load of node i (determined by equipment capacity), ensuring that the electrical equipment operates within a safe range.

[0067] Energy storage constraints:

[0068]

[0069] represents the lower limit of SOC; represents the upper limit of SOC; represents the charging and discharging efficiency; represents the rated capacity.

[0070] Generation power constraints:

[0071] represents the lower limit value of generation power, represents the upper limit value of generation power.

[0072] The weighted sum of the first objective function and the second objective function is taken as the comprehensive objective function f:

[0073] S401. Solve the optimal particle position vector according to the comprehensive objective function and the probability distribution by using a quantum particle swarm optimization algorithm.

[0074] The quantum particle swarm optimization algorithm enhances the global search ability of the traditional particle swarm optimization algorithm by introducing a quantum behavior mechanism. In step S401, the particle position vector is first defined, and its dimension corresponds to the decision variables in the scheduling problem, such as the active power of each generator, the charging and discharging power of the energy storage system, etc. The comprehensive objective function integrates economic indicators (such as generation cost, electricity purchase cost) and safety indicators (such as node voltage deviation, line load rate), and adjusts the relative importance of the two through weight coefficients. At the initialization of the algorithm, a swarm of particles is randomly generated in the solution space, each particle having a position vector and a velocity vector. In the iteration process, the particles are updated according to the individual optimal position (pbest) and the global optimal position (gbest). Unlike the traditional particle swarm algorithm, QPSO introduces the concept of quantum bits, adjusts the particle position through the quantum rotation gate, and makes the particle appear at any position in the solution space with a certain probability in the quantum state, effectively avoiding falling into local optimum. After each iteration, the comprehensive objective function value of the particle corresponding to the solution is calculated, and pbest and gbest are updated. When the maximum number of iterations is reached or the convergence condition (such as the change of gbest in continuous iterations being less than a threshold value) is met, the optimal particle position vector is output as the initial solution for subsequent processing.

[0075] S402. Discretely convert the particle position vector into an economic pheromone map and a security pheromone map.

[0076] The continuous optimal particle position vector is converted into a discrete economic pheromone map and a security pheromone map. The conversion process is based on the physical topology of the power system, mapping each node and branch to an element in the pheromone map. For the economic pheromone map, according to the economic indicators such as generation cost, purchase cost, etc. corresponding to the decision variables in the particle position vector, different pheromone concentrations are assigned to the elements in the map according to the proportional relationship. The higher the pheromone concentration, the better the path or node is in economic dispatch. The construction of the security pheromone map is based on the security indicators such as node voltage deviation and line load rate. By setting a security threshold, the decision variables in the particle position vector are converted into security state evaluation values, which are then mapped to the concentration distribution of the pheromone map. For example, for nodes with excessive voltage deviation, their pheromone concentration in the security pheromone map is reduced; while for branches in good operating condition, their pheromone concentration is increased, so as to intuitively reflect the security status of the system.

[0077] The output of QPSO is a multi-dimensional decision vector, representing the complete dispatch strategy: x* = [DR, P ch , P dis , DG curt ,...] Where DR ∈ ^(N×T) is the demand response decision matrix, with element value: [-1, 1] representing the load reduction ratio; P ch ∈ ^(E×T) is the energy storage charging power matrix; P dis ∈ ^(E×T) is the energy storage discharging power matrix; DG curt ∈ ^(G×T) is the distributed generator limited generation decision.

[0078] Discretize the continuous vector into decision paths: (Node 1, t1)→(Node 2, t1)→...→(Node N, tT).

[0079] S403. Solve the economic optimal solution according to the economic pheromone map and the first objective function by using the first ant colony in the adaptive double ant colony algorithm.

[0080] The first ant colony searches based on the economic pheromone map, aiming to solve the economic optimal solution. The ants of the first ant colony move in the pheromone map according to the probability transition rule, which combines the pheromone concentration and the heuristic function. The heuristic function is usually designed based on the economic cost difference between nodes, guiding the ants to preferentially select paths with lower costs. As the ants move, the pheromone is updated according to the quality of the path, exhibiting a "positive feedback" mechanism: the pheromone concentration of the path with lower economic cost increases, attracting more ants to choose, and vice versa. At the same time, the algorithm introduces an adaptive adjustment mechanism to dynamically adjust the pheromone evaporation coefficient and the number of ants according to the search process. In the early stage of search, the pheromone evaporation coefficient is increased to speed up the global search speed of the algorithm; as the iteration proceeds, the evaporation coefficient is reduced to enhance the local search ability of the algorithm. Through continuous iteration, the first ant colony finally converges to the optimal path in the economic pheromone map, corresponding to the economic optimal scheduling scheme of the power system.

[0081] Economic pheromone update:

[0082] η represents the path efficiency coefficient (based on the initial solution of QPSO); 𝔼[C] represents the expected cost = purchase cost - response income; Var[C] represents the risk cost (from the prediction layer Σ_price).

[0083] S404. Solve the safety optimal solution according to the safety pheromone map and the second objective function by using the second ant colony in the adaptive double ant colony algorithm.

[0084] The second ant colony searches based on the safety pheromone map, guided by the second objective function (mainly considering system safety constraints). Similar to the first ant colony, the ants of the second ant colony move in the map according to the pheromone concentration and the safety-related heuristic function. When designing the heuristic function, the focus is on safety factors such as node voltage stability and line overload risk, and the path with large safety margin is preferentially selected. In terms of pheromone update strategy, the second ant colony also adopts adaptive adjustment. For safety-critical nodes and branches, a higher pheromone enhancement coefficient is set to intensify the search in these areas. Through repeated iteration, the second ant colony finds the optimal path in the safety pheromone map and obtains the optimal scheduling scheme that satisfies the safety constraints.

[0085] Security information update:

[0086] γ represents a security violation penalty factor; V grid represents a node voltage out-of-limit amount; Q2 represents a positive correlation with grid topology complexity.

[0087] S405. Integrating the economic optimal solution and the security optimal solution to obtain an optimal scheduling scheme.

[0088] The economic optimal solution and the security optimal solution are integrated by using a weighted fusion strategy. According to the current operating state and scheduling demand of the power system, the economic weight and the security weight are determined by expert experience or an optimization algorithm. The decision variables (such as unit output, load distribution, etc.) in the economic optimal solution and the security optimal solution are weighted and summed according to the weights to obtain the final optimal scheduling scheme. To ensure the feasibility of the scheme, the integrated scheduling scheme is subjected to constraint checking, including power balance constraint, voltage constraint, line capacity constraint, etc. If the scheme does not satisfy the constraint conditions, the decision variables are fine-tuned by a heuristic adjustment algorithm until an optimal scheduling scheme satisfying all constraints is obtained, thereby providing decision support for the economic and secure operation of the power system.

[0089] In some embodiments, the virtual power plant scheduling system considering electricity price can include a plurality of functional modules composed of computer program segments. The computer programs of each program segment in the virtual power plant scheduling system considering electricity price can be stored in the memory of the computer device and executed by at least one processor to perform the functions of the virtual power plant scheduling system considering electricity price (see detailed description). Figure 1

[0090] In this embodiment, the virtual power plant scheduling system considering electricity price can be divided into a plurality of functional modules according to the functions it performs, as shown in Figure 2 The module referred to in the present application refers to a series of computer program segments that can be executed by at least one processor and can complete a fixed function, which are stored in the memory. In this embodiment, the functions of each module will be described in detail in subsequent embodiments.

[0091] The acquisition module is configured to periodically acquire multi-source monitoring data, and the multi-source monitoring data includes electricity price data, load data, weather data, and power supply data. The preprocessing module is configured to construct a spatiotemporal tensor according to the acquisition time and monitoring position of the multi-source monitoring data. The prediction module is configured to input the spatiotemporal tensor into a prediction model to obtain the probability distribution of load demand, distributed power output, and market electricity price within a set time span. ​A solution module is configured to solve the optimal scheduling scheme according to the probability distribution and the constructed minimum economic cost objective function and minimum load fluctuation objective function.

[0092] Figure 3 The virtual power plant scheduling method considering electricity price provided by the embodiments of the present application can be applied to equipment. Those skilled in the art can understand that the equipment structure involved in the embodiments of the present application does not constitute a limitation on the equipment, and the equipment can include more or fewer components than the illustration, or combine certain components, or different component arrangements. In the embodiments of the present application, the equipment includes but is not limited to a laptop computer, a desktop computer, a workstation, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The equipment can also represent various forms of mobile devices, such as a personal digital processor, a cellular phone, a smart phone, a wearable device, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples, and are not intended to limit the implementation of the embodiments of the present application described herein and / or claimed.

[0093] The device 300 can include a processor 310, a memory 320, and a communication unit 330. These components communicate through one or more buses. Those skilled in the art can understand that the structure of the server shown in the figure does not constitute a limitation on the present application, and it can be a bus structure or a star structure, and can include more or fewer components than the illustration, or combine certain components, or different component arrangements.

[0094] The memory 320 can be used to store the execution instructions of the processor 310. The memory 320 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic storage, a flash memory, a magnetic disk or an optical disk. When the execution instructions in the memory 320 are executed by the processor 310, the device 300 can execute part or all of the steps in the following method embodiments.

[0095] The processor 310 is the control center of the storage device, connects various parts of the entire electronic device by using various interfaces and lines, and performs various functions of the electronic device and / or processes data by running or executing software programs and / or modules stored in the memory 320 and calling data stored in the memory. The processor can be composed of an integrated circuit (IC), for example, can be composed of a single packaged IC, or can be composed of a plurality of packaged ICs connected together. For example, the processor 310 can only include a central processing unit (CPU). In the embodiments of the present application, the CPU can be a single operation core or can include multiple operation cores.

[0096] The communication unit 330 is configured to establish a communication channel, so that the storage device can communicate with other devices. The communication unit 330 receives user data sent by other devices or sends user data to other devices.

[0097] The present application also provides a computer storage medium, wherein the computer storage medium can store a program, and the program can include some or all steps in the embodiments of the present application when executed. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0098] Those skilled in the art can clearly understand that the technology in the embodiments of the present application can be realized by means of software and necessary general hardware platforms. Based on such understanding, the technical solutions in the embodiments of the present application can be embodied in the form of a software product, which is stored in a storage medium such as a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc. The software product includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a second device, a network device, etc.) to execute all or part of the steps of the method described in the embodiments of the present application.

[0099] In the present specification, the same or similar parts among various embodiments can be referred to each other. In particular, for the device embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the description in the method embodiments.

[0100] In several embodiments provided by the present application, it should be understood that the disclosed system and method can be implemented in other manners. For example, the system embodiments described above are merely schematic. For example, the division of the modules is only a logical function division. There can be another division manner for the actual implementation. For example, a plurality of modules or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between different modules can be indirect couplings or communication connections through some interfaces, and electrical, mechanical or other forms.

[0101] The modules illustrated as separated components can or can not be physically separated, and the components illustrated as modules can or can not be physical modules, i.e., can be located in one place, or can be distributed to a plurality of network modules. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment.

[0102] In addition, each functional module in each embodiment of the present application can be integrated into a processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0103] Although the present application has been described in detail by referring to the preferred embodiments thereof, it is to be understood that the present application is not limited to the embodiments described above. Rather, it should be appreciated that those skilled in the art, upon consideration of the disclosure, can make modifications and / or improvements to the embodiments of the application without deviating from the spirit and scope of the application. Any modifications and / or improvements made to the embodiments of the application by those skilled in the art are to be considered within the scope of the application.

Claims

1. A virtual power plant scheduling method considering electricity prices, characterized in that: include: Regularly collect multi-source monitoring data, including electricity price data, load data, meteorological data, and power supply data; Construct a spatiotemporal tensor based on the acquisition time and monitoring location of multi-source monitoring data; Inputting the spatiotemporal tensor into a prediction model to obtain a probability distribution of load demand, distributed power generation output, and market electricity price within a set time span; According to the probability distribution and the constructed objective function of minimizing economic cost and minimizing load fluctuation, the optimal scheduling plan is solved.

2. The method according to claim 1, characterized in that Construct a spatiotemporal tensor based on the acquisition time and monitoring location of multi-source monitoring data, including: Construct an adjacency matrix of monitoring nodes based on the power grid topology; Use the time sliding window mechanism to generate a time series feature matrix for multi-source monitoring data; The temporal feature matrix and the adjacency matrix are integrated into a spatiotemporal tensor.

3. The method according to claim 2, characterized in that Construct an adjacency matrix of monitoring nodes based on the power grid topology, including: Determine the edge weights between corresponding nodes in the power grid topology based on the electrical distance and load correlation between the nodes; Construct an initial adjacency matrix based on the number of nodes in the power grid topology; According to the edge weights between nodes in the power grid topology, corresponding elements in the initial connection matrix are filled, and an adjacency matrix is ​​obtained after the filling is completed.

4. The method according to claim 1, wherein The prediction model includes: The temporal feature extraction layer captures the local temporal features of the spatiotemporal tensor through the ConvLSTM network; The spatial feature aggregation layer obtains spatial enhancement features by performing graph convolution on the local temporal features of each time step and the adjacency matrix of the spatiotemporal tensor; The output layer is used to decompose the spatial enhancement features into a feature sequence for each node. For each node, the feature sequence is input into a mixture density network, which outputs the probability distribution parameters of load demand, power output and market electricity price based on a time stepping method; the mixture density network uses a Gaussian mixture model to generate a conditional probability distribution for each output variable, including the mixing coefficient, mean and variance; each output variable is processed by an independent output head to model the correlation between variables.

5. The method according to claim 1, wherein The method further comprises: The first objective function of minimizing economic cost and the second objective function of minimizing load fluctuation are constructed respectively; The first objective function includes: in, is the electricity purchase price during period t; is the amount of electricity purchased during period t, calculated as the total load plus the charging and discharging power of the energy storage system minus the generated power minus the power reduction; is the demand response subsidy for period t; is the total power reduction; The second objective function includes: Indicates the power change in adjacent time periods; Indicates the maximum allowable power change rate; Indicates the intensity of the penalty for exceeding the limit; T indicates the duration of the monitoring cycle; The weighted sum of the first objective function and the second objective function is taken as the comprehensive objective function.

6. The method according to claim 5, characterized in that The method further comprises: A constraint function system is constructed, wherein the constraint function system includes power balance constraint, adjustable load constraint, energy storage constraint and power generation constraint.

7. The method according to claim 6, characterized in that According to the probability distribution and the constructed objective function of minimizing economic cost and minimizing load fluctuation, the optimal dispatching solution is solved, including: Using a quantum particle swarm optimization algorithm to solve the optimal particle position vector according to the comprehensive objective function and the probability distribution; Discretely converting the particle position vector into an economic pheromone graph and a safety pheromone graph; The first ant colony in the adaptive dual ant colony algorithm is used to solve the economic optimal solution according to the economic pheromone graph and the first objective function; The second ant colony in the adaptive dual ant colony algorithm is used to solve the optimal security solution based on the security pheromone graph and the second objective function; Integrate the economic optimal solution and the safety optimal solution to obtain the optimal scheduling plan.

8. A virtual power plant dispatching system considering electricity prices, characterized in that: include: The acquisition module is used to regularly collect multi-source monitoring data, including electricity price data, load data, meteorological data and power supply data; A preprocessing module, used to construct a spatiotemporal tensor based on the acquisition time and monitoring location of multi-source monitoring data; A prediction module, configured to input the spatiotemporal tensor into a prediction model to obtain a probability distribution of load demand, distributed power output, and market electricity price within a set time span; The solution module is used to solve the optimal scheduling solution according to the probability distribution and the constructed objective function of minimizing economic cost and minimizing load fluctuation.

9. A device, characterized in that include: a memory for storing a virtual power plant scheduling program taking electricity prices into consideration; A processor is configured to implement the steps of the virtual power plant scheduling method considering electricity prices as described in any one of claims 1 to 7 when executing the virtual power plant scheduling program considering electricity prices.

10. A computer-readable storage medium storing a computer program, characterized in that: The readable storage medium stores a virtual power plant scheduling program that considers electricity prices. When the virtual power plant scheduling program that considers electricity prices is executed by a processor, the steps of the virtual power plant scheduling method that considers electricity prices as described in any one of claims 1 to 7 are implemented.

Images