Transformer area electric heating power flexible control method, system and device

Abstract

The invention discloses a transformer area electric heating power flexible control method, system and device, and the method comprises the steps: firstly collecting the historical load and other data of a transformer area to train an XGBoost model, then predicting the load of the transformer area based on the XGBoost model, positioning the peak period of power utilization, and carrying out the flexible control in advance through employing an improved ant colony algorithm. An improved ant colony algorithm introduces a pheromone updating mechanism of a self-adaptive elite factor, and differential control strategies are generated for different types of electric heating equipment in combination with a designed flexible control fitness function; and finally, the control instruction is issued to each electric heating device and executed. In addition, by setting the priority of the equipment, the fairness of the control process is ensured. The defects of an existing electric heating control method in the aspects of flexible regulation and pre-regulation are overcome, and the power supply safety of the transformer area can be effectively maintained on the premise that the heating requirement of a user is guaranteed.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of district electric heating load regulation, and relates to a district electric heating power flexible control method, system and device, in particular to a district electric heating power flexible control method, system and device based on XGBoost, graph network and improved ant colony algorithm. BACKGROUND

[0002] At present, distributed electric heating has replaced traditional coal combustion and become the dominant mode of winter heating in northern China. However, the electric heating load presents the characteristics of "large start-up impact, high simultaneous rate and long duration cycle", and the peak power is high under extreme weather, which leads to problems such as peak overload of distribution district, terminal voltage drop and life loss of distribution transformer. Unlike traditional industrial loads, electric heating has obvious thermal inertia, that is, the thermal resistance of building envelope is large, the indoor temperature fluctuation range is wide, and short-time power reduction or suspension of operation of equipment will not significantly affect thermal comfort, thereby providing physical feasibility for load flexible regulation.

[0003] Flexible control refers to the control mode of flexibly adjusting power and time period by using building heat storage and the regulation capacity of electric heating equipment. Fully awakening and utilizing the adjustable potential of distributed electric heating is a key technical path to improve district flexibility, delay infrastructure investment and ensure people's livelihood heating. Therefore, it is urgent to research an electric heating load flexible control scheme that takes into account "power supply safety" and "user comfort", effectively reduces peak load and ensures district power safety on the premise of ensuring residents' warm winter.

[0004] Electric heating control usually uses direct load control (DLC). DLC is a typical means of power demand side management, which is to remotely shut down or adjust part of the user's power equipment by intelligent terminal equipment under the premise of user's prior consent in a certain period such as peak load or emergency state, so as to quickly reduce power demand and ensure stable operation of the system. The traditional optimization algorithms used in DLC electric heating load regulation include linear programming, multi-objective linear programming and dynamic programming algorithm. Linear programming has simple modeling and high efficiency, but it is difficult to consider multiple objectives and has poor adaptability to uncertainty, while multi-objective linear programming and dynamic programming algorithm will face dimension disaster. SUMMARY

[0005] The purpose of the present application is to provide a district electric heating power flexible control method, system and device to solve one or more problems in the prior art.

[0006] Technical scheme: A district electric heating power flexible control method of the present application comprises:

[0007] (1) Collect historical load data of transformer substations and perform feature engineering processing. Construct a sample set based on the historical load data of transformer substations after feature engineering processing and auxiliary information. The auxiliary information includes weather and holidays; using a sample set Train the XGBoost model;

[0008] (2) Obtain the predicted load value of the transformer area using the trained XGBoost model. When the predicted load value of the transformer area Greater than the area load threshold If the condition is met, proceed to step (3); otherwise, do not perform flexible control and maintain the current operating state of the electric heating equipment.

[0009] (3) Determining the optimal flexible control scheme based on graph network and improved ant colony algorithm: The set of all electric heating devices to be controlled in the entire transformer area is modeled as a graph network, where each electric heating device is abstracted as a node in the graph, and the edges between nodes represent the decision paths between electric heating devices; the task of each ant in the ant colony is to construct a path that traverses all nodes, and this path represents a specific device control sequence scheme; when the ant visits each node, it makes a decision based on pheromone and heuristic information; after all ants have completed a round of scheme construction, the scheme is evaluated, selected, and the pheromone is updated according to the adaptive flexible control fitness function; the adaptive flexible control fitness function is divided into two parts: the power grid objective function and the user comfort objective function; the pheromone update adopts the adaptive elite factor pheromone update mechanism; when the maximum number of iterations is reached, the optimal flexible control scheme is output.

[0010] (4) Execute the optimal flexible control scheme and monitor the control effect. If the control is successful, return to step (2) and carry out the next round of load prediction and control process.

[0011] Further, step (1) includes: for the time feature, given a timestamp Extract the following features: 1) Time series The granularity of recording can be set as needed; 2) 1 to 7 represent Monday to Sunday respectively; 3) Where 1 represents a holiday and 0 represents a workday;

[0012] Based on the historical load characteristics of the transformer area, a lag characteristic is constructed, using the target time. The load values ​​at several previous time points are used as characteristics:

[0013] (1)

[0014] in, representing the time lag characteristics; representing the load value at time before time units, , is the maximum time lag order;

[0015] For meteorological features, obtain temperature , relative humidity and wind speed at time

[0016] For sliding window statistical features, construct sliding average features and sliding standard deviation features:

[0017] 1) Sliding average : the average load of the past time window, reflecting the recent average level, which is calculated as:

[0018] (2)

[0019] 2) Sliding standard deviation : the standard deviation of the load within the past time window, reflecting the volatility of the recent load, which is calculated as:

[0020] (3)

[0021] Finally, all the above features are concatenated into a feature vector at time , which together with the target value, i.e. the load at a future time , forms a complete sample , representing the time step of prediction; all samples are arranged in chronological order to form the final sample set , where represents the sample at time , and represents the total number of samples.

[0022] Further, step (1) further comprises: the prediction output of the XGBoost model is jointly determined by the prediction results of regression trees, and for sample , its predicted value is:

[0023] (4)

[0024] where is the a decision tree; is the function space consisting of all possible decision trees;

[0025] Objective function of XGBoost model consists of two parts: a loss function and a regularization term

[0026] (5)

[0027] where, is the parameter of XGBoost model; the loss function uses mean squared error;

[0028] XGBoost regularizes each tree:

[0029] (6)

[0030] where, and are hyperparameters; is the number of leaf nodes of the tree; is the prediction value vector on the leaf nodes.

[0031] Further, step (2) comprises: for an input containing a new feature vector , inputting its features into the trained XGBoost model, and adding the prediction results of all decision trees, i.e. obtaining the substation load prediction value .

[0032] Further, in step (3), according to different regulation modes, the electric heating equipment is divided into three-phase electric heating equipment, variable frequency electric heating equipment and single-phase electric heating equipment, for the three-phase and variable frequency operation electric heating equipment, the power is reduced; for the single-phase operation electric heating equipment, the intermittent heating is realized to reduce the average power according to a certain time length of duty cycle or branch rotation power-off.

[0033] Further, step (3) comprises:

[0034] (301) algorithm initialization

[0035] unified pheromone concentration is given to all edges , the maximum number of ants , the maximum number of iterations , the pheromone evaporation factor , and the importance weight of pheromone , the importance weight of heuristic information ;

[0036] (302) Iterative process of ant colony construction device regulation sequence scheme

[0037] In each iteration , only ants are released; each ant starts from the initial node, and according to the state transition probability, selects the next unvisited node and decides the regulation state of the corresponding electric heating device; when the ant is located at node , the next node to be decided is , it selects node and gives the probability of regulation state :

[0038] (8)

[0039] wherein, is the pheromone concentration of edge at iteration ; is the current unvisited node set of ant ; represents the selected node in the current unvisited node set; is the heuristic information, which is calculated as follows:

[0040] (9)

[0041] wherein, , and are weight coefficients of each heuristic factor, ; and are the set temperature and the actual indoor temperature of the electric heating device corresponding to node , respectively; represents the maximum difference between the set temperature and the actual indoor temperature of all electric heating devices; is the real-time active power of the electric heating device corresponding to node ; is the maximum power of the electric heating device in the transformer area; is the priority weight based on the device type or user contract;

[0042] After each ant completes a full graph traversal, a complete device regulation sequence scheme is generated, which consists of the power and on-off state of all electric heating devices, i.e. ; then, it is determined whether all ants have completed the construction of the device regulation sequence scheme, if yes, step (303) is entered; if not, the next ant is selected The device regulation sequence scheme is constructed;

[0043] (303) Scheme evaluation, selection and pheromone update

[0044] After all the ants complete a round of scheme construction, the adaptive flexible regulation fitness function is used to evaluate the scheme For each scheme Evaluation is carried out;

[0045] (10)

[0046] wherein, and is a weight coefficient, satisfying ;

[0047] The grid target function is:

[0048] (11)

[0049] wherein, is a weight coefficient; represents the peak load of the substation in the regulation period under the scheme ; represents the peak value in the substation load predicted by the XGBoost model;

[0050] The user comfort target function is:

[0051] (12)

[0052] wherein, represents the user with electric heating flexible regulation capacity; is the outdoor temperature of the area where the device is located;

[0053] After the evaluation is completed, the global pheromone is updated; first, the pheromone on all paths is volatilized in proportion:

[0054] (13)

[0055] Subsequently, according to the advantages and disadvantages of each scheme in this round of iteration, the paths passed by the excellent schemes are enhanced; let and respectively represent the fitness value of the optimal solution and the fitness value of the worst solution in one round of iteration, for any solution used for elite enhancement , its adaptive elite factor is:

[0056] (14)

[0057] wherein, is a base weight constant; is a number to prevent denominator from being 0;

[0058] use the optimal solution of this iteration enhance the pheromone:

[0059] (15)

[0060] wherein, is a constant pheromone intensity coefficient;

[0061] Then, it is judged whether the set maximum iteration number is reached, if the maximum iteration number is reached, the global optimal control scheme is output and is issued for execution, if the maximum iteration number is not reached, the next round of iteration is performed, in multiple iterations, the search process will gradually converge to an electric heating flexible control scheme which can balance the total load target of the transformer area and the individualized demand of users .

[0062] Further, in step (4), a priority mechanism is introduced in the control process, the initial priority of the electric heating equipment is divided into several levels; the priority queue of the ordinary user equipment is sorted in descending order, when the transformer area load may be overloaded and needs to be controlled, the electric heating load flexible control is preferentially selected for the equipment with low priority; then the priority of the equipment controlled in the current round is set to the highest and is added to the end of the priority queue; the priority of the equipment not controlled is reduced by one level, if the priority is already the lowest, it is not changed; the special user equipment does not use the above priority rules and always maintains the highest priority; only when all the ordinary user equipment is controlled and the load reduction requirement is still not met, the electric heating flexible control is performed on the special user.

[0063] Further, in step (4), the monitoring control effect comprises: collecting the actual state of each electric heating equipment to form a round instruction execution result report; comparing the expected state of the instruction with the actual state, if the success rate is lower than a threshold value, an instruction retransmission mechanism is started or a new remedial instruction is generated for the electric heating equipment which is not successfully executed; if the overall control effect deviates from the expectation, the deviation is taken as feedback and is input into the ant colony control algorithm in the next round to dynamically adjust the control strategy and realize adaptive closed-loop optimization.

[0064] The electric heating power flexible control system of the transformer area of the application comprises:

[0065] The data acquisition and model training module is used for collecting the historical load data of the transformer area and performing feature engineering processing, and a sample set is constructed based on the historical load data of the transformer area after feature engineering processing and auxiliary information The auxiliary information includes weather, holidays, and the like The XGBoost model is trained;

[0066] The load prediction and judgment module is configured to obtain a load prediction value of the transformer area by using the trained XGBoost model When the load prediction value of the transformer area is greater than a load threshold value of the transformer area The optimal flexible control scheme generation module is executed; otherwise, no flexible control is performed, and the current operation state of the electric heating equipment is maintained unchanged. The optimal flexible control scheme generation module is configured to determine an optimal flexible control scheme based on a graph network and an improved ant colony algorithm: all electric heating equipment to be regulated in the entire transformer area is modeled as a graph network, each electric heating equipment is abstracted as a node in the graph, and the edges between the nodes represent the decision paths between the electric heating equipment; the task of each ant in the ant colony is to build a path that traverses all the nodes, and the path represents a specific device regulation sequence scheme; when the ant visits each node, the ant makes a decision based on pheromone and heuristic information; after all the ants complete a round of scheme construction, scheme evaluation, optimization and pheromone updating are performed according to an adaptive flexible regulation fitness function; the adaptive flexible regulation fitness function is divided into two parts: a power grid objective function and a user comfort objective function; the pheromone updating adopts a pheromone updating mechanism of an adaptive elite factor; when the maximum number of iterations is reached, the optimal flexible control scheme is output.

[0067] The control execution and monitoring module is configured to execute the optimal flexible control scheme generated by the optimal flexible control scheme generation module and monitor the control effect; if the regulation is successful, the load prediction and judgment module is returned to perform the next round of load prediction and regulation process.

[0068] The power flexible control device for electric heating in a transformer area of the application comprises a transformer area master station, a concentrator, a plurality of flexible load control terminal host computers and a plurality of flexible load control terminal slave computers corresponding to each flexible load control terminal host computer, the transformer area master station has a processor and a memory, the memory stores computer instructions, and the processor is configured to execute the computer instructions stored in the memory, when the computer instructions are executed by the processor, the power flexible control device for electric heating in the transformer area implements the steps of the power flexible control method for electric heating in the transformer area, and obtains an optimal flexible control scheme; the concentrator is responsible for communication with the transformer area master station and aggregation of data of all terminal host computers connected thereto; the flexible load control terminal host computer is responsible for management and control of a plurality of flexible load control terminal slave computers; the flexible load control terminal slave computer is installed at the user side, directly interfaces with the electric heating equipment, and performs the final control operation.

[0069] The power flexible control device for electric heating in a transformer area of the application comprises a transformer area master station, a concentrator, a plurality of flexible load control terminal host computers and a plurality of flexible load control terminal slave computers corresponding to each flexible load control terminal host computer, the transformer area master station has a processor and a memory, the memory stores computer instructions, and the processor is configured to execute the computer instructions stored in the memory, when the computer instructions are executed by the processor, the power flexible control device for electric heating in the transformer area implements the steps of the power flexible control method for electric heating in the transformer area, and obtains an optimal flexible control scheme; the concentrator is responsible for communication with the transformer area master station and aggregation of data of all terminal host computers connected thereto; the flexible load control terminal host computer is responsible for management and control of a plurality of flexible load control terminal slave computers; the flexible load control terminal slave computer is installed at the user side, directly interfaces with the electric heating equipment, and performs the final control operation.

[0070] Beneficial effects: Compared with the prior art, the present application has the following remarkable advantages: the present application constructs a complete, efficient and fair flexible control system for the electric heating power of the transformer area by fusing the XGBoost prediction and the improved ant colony algorithm, and remarkable beneficial effects are achieved in preventing the overload of the transformer area, guaranteeing the heating of the users, improving the control efficiency and fairness, etc.

[0071] (1) In view of the problems of insufficient load prediction accuracy and power supply safety of the transformer area, the present application adopts the XGBoost algorithm to accurately predict the total load of the transformer area, combines meteorological data, historical load and other multi-source information, realizes the early identification of the load peak, and on this basis, generates the electric heating power adjustment scheme through the ant colony algorithm, effectively avoids the overload of the transformer area while guaranteeing the heating demand of the users, and improves the power supply safety.

[0072] (2) In view of the problems of insufficient fairness of the control scheme and insufficient protection of special users, the present application introduces the user priority mechanism in the control strategy, distinguishes the ordinary users from the special users such as hospitals and schools, and ensures that the heating of the special users is not affected in the process of flexible regulation. The mechanism is embedded in the constraint condition of the ant colony algorithm, and the fairness of regulation and the actual social demand are considered.

[0073] (3) In view of the problems of slow convergence and low solution quality of the traditional optimization algorithm in the complex scene, the present application improves the ant colony algorithm, optimizes the pheromone update rule and heuristic information design, and strengthens the user selection tendency in the historical best regulation scheme, so as to accelerate the convergence of the algorithm, improve the global optimization ability, and quickly obtain the high-quality flexible control scheme.

[0074] (4) In view of the problem that the comprehensive evaluation system of the regulation scheme is imperfect, the present application designs a weighted fitness function, comprehensively considers the technical feasibility, user satisfaction and system operation efficiency of the control scheme, makes the generated scheme more suitable for the actual application scene, and realizes the multi-objective collaborative optimization.

[0075] (5) In view of the problem of open loop in the control process and lack of feedback adjustment, the present application designs a closed-loop control process, which covers the whole process of instruction issuing, execution and effect monitoring feedback. Through real-time monitoring of the regulation effect and dynamic adjustment of the user priority and the control strategy, continuous optimization is realized, and stable and reliable operation of the system is ensured. BRIEF DESCRIPTION OF DRAWINGS

[0076] Figure 1 is a flow chart of a flexible control method for the electric heating power of the transformer area provided by an embodiment of the present application;

[0077] Figure 2 is a flow chart of the load prediction based on XGBoost and the generation of the flexible control scheme for the electric heating power based on the improved ant colony algorithm in the embodiment of the present application;

[0078] Figure 3 This is a diagram of the hierarchy and communication architecture of the flexible control device for electric heating power in this embodiment of the invention;

[0079] Figure 4 This is a schematic diagram of equipment priority during flexible control of electric heating power in an embodiment of the present invention. Detailed Implementation

[0080] The invention will now be further described with reference to the accompanying drawings.

[0081] Example 1: As Figure 1 and Figure 2 As shown, Example 1 provides a method for flexible control of electric heating power in a transformer substation, including the following steps:

[0082] (1) Collect historical load data of transformer substations and perform feature engineering processing. Construct a sample set based on the historical load data of transformer substations after feature engineering processing and auxiliary information. The auxiliary information includes weather and holidays; using a sample set The XGBoost model is trained. Feature engineering is performed on the historical load data of the transformer substations, including constructing hysteresis features, periodic coding, and moving averages.

[0083] The specific process of data collection is as follows:

[0084] 1) For the data required for load forecasting, historical operating data of the distribution transformer area is collected and frozen at a preset interval of 15 minutes through monitoring terminals deployed in the area. This data includes at least active power, current, and electricity consumption. For advanced distribution transformer areas with distribution automation, the above data is preferentially obtained through feeder terminal units (FTUs) to ensure data accuracy and real-time performance. To achieve accurate load forecasting, external environmental variables are introduced simultaneously; historical meteorological observation data with the same time granularity as the load data is obtained by downloading from the national meteorological data center for the corresponding time period of the distribution transformer area. 2) For the user equipment status data required for electric heating control, flexible load control terminals in the user's meter box are used for collection. Specific data includes the user's electric heating equipment type, voltage, frequency, total active power, total reactive power, and the user-set temperature. Missing values ​​are filled and outliers are handled for all collected data.

[0085] The specific process of feature engineering for feature construction is as follows:

[0086] For time features, given a timestamp Extract the following features: 1) Time series The recording granularity is 15 min, or 0.25 h; 2) 1 to 7 represent Monday to Sunday respectively; 3) where 1 represents a holiday and 0 represents a weekday.

[0087] For the historical load characteristics of the transformer area, lag features are constructed, and the load values at the target time point and several time points before the target time point are used as features:

[0088] (1)

[0089] wherein, represents the lag feature at time ; represents the load value at time before time units, , is the maximum lag order.

[0090] For meteorological features, the temperature , relative humidity , and wind speed at the target time are obtained from the national meteorological data center.

[0091] For sliding window statistical features, sliding average features and sliding standard deviation features are constructed:

[0092] 1) Sliding average : the average load in the past time window, reflecting the recent average level, and the calculation method is:

[0093] (2)

[0094] 2) Sliding standard deviation : the standard deviation of the load in the past time window, reflecting the volatility of the recent load, and the calculation method is:

[0095] (3)

[0096] Finally, all the above features are concatenated into a feature vector at time , and the feature vector and the target value, i.e., the load at a certain future time , together constitute a complete sample , represents the time step of prediction; all samples are arranged in chronological order to form the final sample set , wherein represents the sample at time , and represents the total number of samples.

[0097] ​The specific process of model training is as follows:

[0098] XGBoost is an ensemble model, whose prediction output is jointly determined by the prediction results of regression trees. For a sample , its predicted value is:

[0099] (4)

[0100] where is the th decision tree; is the function space composed of all possible decision trees;

[0101] The objective function of the XGBoost model is composed of two parts: the loss function and the regularization term , aiming to ensure the accuracy of the prediction and the simplicity of the model to prevent overfitting. Its calculation method is:

[0102] (5)

[0103] where is the parameter of the XGBoost model; the loss function uses the mean square error, i.e. ;

[0104] XGBoost regularizes each tree:

[0105] (6)

[0106] where and are hyperparameters that control the penalty strength of the leaf node number and the L2 regularization strength of the leaf node weight, respectively; is the number of leaf nodes of the tree; is the prediction value vector on the leaf node.

[0107] (2) Obtain the substation load prediction value using the trained XGBoost model When the substation load prediction value is greater than the substation load threshold , perform step (3); otherwise, do not perform flexible control and keep the current operating state of the electric heating equipment unchanged.

[0108] After the model training is completed, for a new feature vector The input is taken, its features are fed into the pre-trained XGBoost model, and the prediction results of all decision trees are summed to obtain the predicted load value of the transformer area. .

[0109] (3) Determining the optimal flexible control scheme based on graph network and improved ant colony algorithm: The set of all electric heating devices to be controlled in the entire transformer area is modeled as a graph network, where each electric heating device is abstracted as a node in the graph, and the edges between nodes represent the decision paths between electric heating devices; the task of each ant in the ant colony is to construct a path that traverses all nodes, and this path represents a specific device control sequence scheme; when the ant visits each node, it makes a decision based on pheromone and heuristic information; after all ants have completed a round of scheme construction, the scheme is evaluated, selected, and pheromone is updated according to the adaptive flexible control fitness function; the adaptive flexible control fitness function is divided into two parts: the power grid objective function and the user comfort objective function; the pheromone update adopts the adaptive elite factor pheromone update mechanism; when the maximum number of iterations is reached, the optimal flexible control scheme is output.

[0110] Specifically, the areas within the district awaiting regulation... The collection of electric heating devices is modeled as a graph network. Each device is abstracted as a node. Edges between nodes Representative equipment To the equipment The decision-making path of each ant in the colony. The task is to construct a Hamiltonian path. This path represents a specific equipment control sequence scheme. The essence of the scheme is to determine the state of each device in the path during the control period.

[0111] Based on different control methods, electric heating equipment can be divided into the following three categories: 1) Three-phase electric heating equipment (e.g., distributed electric boilers), where the flexible load control terminal slave unit controls the three-phase load separately, with each phase power accounting for approximately 33% of the total power of the electric heating equipment, and the phase sequence of the water pump being uncontrolled. During control, the load can be reduced in steps of 33% of the total load. 2) Variable frequency electric heating equipment (e.g., air source heat pumps), where the operating frequency of the electric heating equipment is adjusted by the flexible load control terminal slave unit to achieve load adjustment. 3) Single-phase electric heating equipment (e.g., heating film / heating cable, etc.), where the flexible load control terminal slave unit performs intermittent heating by alternately cutting off power to each circuit according to a certain duty cycle or by turning off power in turn, thereby reducing the average power.

[0112] The above three types of equipment have adjustable loads As shown in equation (7), where Indicates the first The power of the electric heating device, The number of phases of the three-phase electric heating device is disconnected. The frequency of the variable frequency electric heating device is represented, The frequency is reduced by regulation. For the three-phase and variable frequency operation of the electric heating device, its power is reduced; for the single-phase operation of the electric heating device, the average power is reduced by the intermittent heating of the certain duty cycle or shunt rotation power-off.

[0113] (7)

[0114] (301) Algorithm initialization

[0115] Assign a uniform pheromone concentration to all edges , set the maximum number of ants , the maximum number of iterations , the pheromone evaporation factor , and the importance weight of pheromone , the importance weight of heuristic information ; and respectively control the relative influence of pheromone and heuristic information.

[0116] (302) Iterative process of ant constructing device regulation sequence scheme

[0117] In each iteration , Only ants are released to construct solutions in parallel; each ant starts from the initial node, selects the next unvisited node according to the state transition probability and decides the regulation state of the corresponding electric heating device; when the ant is located at node , the next node to be decided is , it selects node and gives the probability of regulation state :

[0118] (8)

[0119] where, is the pheromone concentration of edge at iteration , representing the reinforcement degree of historical excellent scheme to this decision; is the current set of unvisited nodes of ant ; represents the selected node in the set of unvisited nodes. is heuristic information, which guides the ants to make better decisions based on localized user and device data, and the calculation method is:

[0120] (9)

[0121] wherein, , and are the weight coefficients of each heuristic factor, ; and are the set temperature and actual indoor temperature of the electric heating device corresponding to the node , respectively; represents the maximum difference between the set temperature and the actual indoor temperature of all electric heating devices; is the real-time active power of the electric heating device corresponding to the node ; is the maximum power of the electric heating device in the transformer area; is the priority weight based on the device type or user contract (for example, the weight of key users such as hospitals and nursing homes is higher).

[0122] When the ant completes a full graph traversal, a complete device control sequence scheme is generated, which consists of the power and on-off state of all electric heating devices, i.e. ; then it is judged whether all ants have completed the device control sequence scheme construction, if yes, it goes to step (303); if not, the next ant is selected to construct the device control sequence scheme.

[0123] (303) Scheme evaluation, optimization and pheromone update

[0124] After all ants complete a round of scheme construction, each scheme is evaluated according to the adaptive flexible control fitness function . In order to follow the principle of flexible control in load control, the adaptive flexible control fitness function as shown in formula (10) is designed in the present application , which is divided into two parts, the first part is the power grid objective function, which promotes load smoothing and peak shaving, and the second part is the user comfort objective function, which considers the indoor temperature and priority of the user.

[0125] (10)

[0126] wherein, and are weight coefficients, satisfying ;

[0127] Grid objective function is:

[0128] (11)

[0129] wherein, is a weight coefficient, used to balance the importance of load control accuracy and peak reduction; denotes the peak load of the substation in the regulation period under the scheme . denotes the peak value in the substation load predicted by the XGBoost model.

[0130] User comfort objective function is:

[0131] (12)

[0132] wherein, denotes the user with flexible regulation capability of electric heating; is the outdoor temperature of the area where the device is located.

[0133] After the evaluation is completed, the global pheromone is updated; first, the pheromone on all paths is volatilized in proportion:

[0134] (13)

[0135] Subsequently, according to the pros and cons of each scheme in this round of iteration, the paths passed by the excellent schemes are enhanced in pheromone; the elite ant strategy with adaptive elite factor is adopted, the better the quality of the solution, the greater the elite weight. Let and denote the fitness value of the optimal solution and the fitness value of the worst solution in one round of iteration, respectively, for any solution used for elite enhancement , its adaptive elite factor is:

[0136] (14)

[0137] wherein, is a basic weight constant; is a very small number, for example, 1e-4, to prevent the denominator from being 0; the fraction measures the relative quality of the current solution among all solutions in this iteration.

[0138] The optimal solution of this iteration is used to enhance the pheromone:

[0139] (15)

[0140] wherein, is a constant pheromone intensity coefficient;

[0141] Subsequently, it is determined whether the maximum number of iterations set is reached. If the maximum number of iterations is reached, the globally optimal control scheme is output and executed. If the maximum number of iterations is not reached, the next round of iteration is performed.

[0142] Through this positive feedback mechanism of "survival of the fittest", in multiple iterations, the search process will gradually converge to an electric heating flexible control scheme that can balance the total load target of the transformer area and the individualized needs of users .

[0143] (4) Execute the optimal flexible control scheme and monitor the control effect. If the control is successful, return to step (2) for the next round of load forecasting and control process.

[0144] In the control process, a priority mechanism is introduced. The initial priority of the electric heating equipment is divided into several levels. The priority queue of the ordinary user equipment is sorted in descending order. When the load of the transformer area may be overloaded and needs to be controlled, the electric heating load flexible control is preferentially selected for the equipment with low priority. Then the priority of the equipment controlled in the current round is set to the highest and added to the end of the priority queue. The priority of the equipment not controlled is reduced by one level, and if its priority is already the lowest, it remains unchanged. The special user equipment does not use the above priority rules and always maintains the highest priority. Only when all the ordinary user equipment is controlled and the load reduction requirement is still not met, the electric heating flexible control is performed on the special user.

[0145] Monitoring the control effect includes: collecting the actual state of each electric heating equipment to form a report on the execution result of this round of instructions; comparing the expected state of the instructions with the actual state. If the success rate is lower than the threshold, the instruction retransmission mechanism is started or new remedial instructions are generated for the electric heating equipment that has not successfully executed; if the overall control effect deviates from the expectation, the deviation is taken as feedback and input into the ant colony control algorithm in the next round to dynamically adjust the control strategy and realize adaptive closed-loop optimization.

[0146] Embodiment 2: A transformer area electric heating power flexible control system, comprising:

[0147] A data acquisition and model training module is configured to acquire historical load data of the transformer area and perform feature engineering processing, construct a sample set based on the historical load data of the transformer area after feature engineering processing and auxiliary information , the auxiliary information includes weather, holidays; and train an XGBoost model using the sample set ;

[0148] The load forecasting and judgment module is used to obtain the load forecast values ​​for the transformer area using a trained XGBoost model. When the predicted load value of the transformer area Greater than the area load threshold If the optimal flexible control scheme generation module is executed, then flexible control will not be performed, and the current operating state of the electric heating equipment will remain unchanged.

[0149] The optimal flexible control scheme generation module is used to determine the optimal flexible control scheme based on graph networks and an improved ant colony algorithm. It models all the electric heating devices to be controlled within the entire distribution area as a graph network, where each device is abstracted as a node, and the edges between nodes represent decision paths between devices. Each ant in the colony is tasked with constructing a path traversing all nodes, representing a specific device control sequence scheme. When visiting each node, the ants make decisions based on pheromones and heuristic information. After all ants have completed one round of scheme construction, the scheme is evaluated, optimized, and pheromones are updated according to an adaptive flexible control fitness function. The adaptive flexible control fitness function consists of two parts: a power grid objective function and a user comfort objective function. Pheromones are updated using an adaptive elite factor pheromone update mechanism. When the maximum number of iterations is reached, the optimal flexible control scheme is output.

[0150] The control execution and monitoring module is used to execute the optimal flexible control scheme generated by the optimal flexible control scheme generation module and monitor the control effect; if the regulation is successful, it returns to the load forecasting and judgment module to carry out the next round of load forecasting and regulation process.

[0151] Example 3: As Figure 3 As shown, a flexible control device for electric heating power in a transformer substation includes a substation master station, a concentrator, several flexible load control terminal hosts, and several flexible load control terminal slaves corresponding to each flexible load control terminal host. The substation master station, located on the transformer side, is the center for generating and deciding control commands. It has a processor and a memory. The memory stores computer instructions, and the processor executes the computer instructions stored in the memory. When the computer instructions are executed by the processor, the flexible control device for electric heating power in the substation implements the steps of the flexible control method for electric heating power in the substation as described in Embodiment 1, thereby obtaining the optimal flexible control scheme.

[0152] The concentrator is responsible for communication with the district master station and aggregation of all terminal host data under its association. The flexible load control terminal host is usually installed at the aggregation point of buildings or units, etc., and is responsible for managing and controlling multiple flexible load control terminal slaves. The flexible load control terminal slave is installed at the user side and directly interfaces with the electric heating equipment to perform the final control operation. The user electric heating equipment, i.e. the controlled object, such as electric heating boilers, air source heat pumps, direct heating electric heaters, etc.

[0153] The specific process of instruction issuing and execution is as follows:

[0154] 1) Control instruction generation and encapsulation

[0155] The district master station generates an optimal electric heating regulation scheme, which is a set of N device regulation instructions, as shown in formula (16).

[0156] (16)

[0157] wherein, represents the unique identifier of the electric heating equipment in the district, represents a specific control command, which consists of a command word (1 byte) and a parameter value (1 byte), and the format is . According to different device types, the settings are shown in Table 1. represents a unified planned execution timestamp, which is used to realize quasi-synchronous control. The district master station encapsulates the instruction set into a master station downlink instruction frame, the structure of which is shown in Table 2.

[0158] Table 1 Design details of control instructions for different types of electric heating equipment

[0159]

[0160] Table 2 Format of district electric heating flexible load control master station downlink instruction frame

[0161]

[0162] The instruction is issued to the target concentrator through wireless virtual private network or optical fiber, etc.

[0163] 2) Instruction acceptance and forwarding

[0164] After receiving the master station downlink instruction frame, the concentrator first performs CRC check to ensure data integrity. After passing the check, the instruction set is parsed, and according to the device identifier in it, the instruction is routed and distributed to each flexible load control terminal host under its association. The concentrator generates a concentrator downlink instruction frame, the structure of which is similar to that of the master station frame, but the target address is changed to the terminal host ID.

[0165] 3) Instruction decoding and slave addressing

[0166] The terminal host receives and checks the instructions from the concentrator. Then it decodes the instruction set and extracts all instructions belonging to the slave devices within its jurisdiction. The terminal host generates a specific host downlink control frame for each target slave, which contains more detailed execution parameters, and issues it to the corresponding flexible load control terminal slave through the LoRa gateway.

[0167] 4) Local execution and state feedback

[0168] After receiving the final control instructions, the slave drives its internal relay or power control module to perform corresponding switching or power regulation operations on the user's electric heating equipment.

[0169] 5) Device priority adjustment

[0170] In order to follow the principle of "fair management" of electric heating flexible control, a priority mechanism is introduced in the control process. As shown in the figure, the initial priority of the device is evaluated according to factors such as outdoor temperature, user set temperature, electric heating equipment type, etc., and is divided into 0-5 levels in total. The priority queue of ordinary user devices is sorted in descending order, and when the load of the transformer area may be overloaded and needs to be regulated, the devices with low priority are preferentially selected, such as Figure 4 devices with priority 0 for electric heating load flexible control. Then the priority of the current round of regulated devices is set to 5 and added to the end of the priority queue. The priority of the unregulated devices is reduced by 1, and if it is already 0, it remains unchanged. Special user (such as hospital, school, etc.) devices do not use the above priority rules and always maintain the highest priority. Only when all ordinary user devices are regulated and still do not meet the load reduction requirements, will the special user be executed for electric heating flexible control. Figure 4

[0171] The specific process of control effect monitoring is as follows:

[0172] After executing the instructions, the slave will immediately collect the actual state of the device (such as relay opening and closing state, real-time power and key operating parameters such as loop current, voltage, and generate a slave uplink state frame, as shown in Table 3.

[0173] Table 3 Flexible control terminal slave uplink instruction frame format

[0174]

[0175] ​​The state frame is uploaded along the original path, first, the terminal slave sends the device state to the corresponding terminal master; second, the terminal master collects the state feedback of all slaves under it, packs it into a master uplink state report frame, and sends it to the concentrator; finally, the concentrator finally collects all the reports of the terminal master, forms a complete round of instruction execution result report, and reports to the substation master station.

[0176] After the substation master station receives the execution result report, it performs closed loop verification. It compares the expected state of the instruction with the actual state reported by the slave. If the success rate is lower than the threshold, the master station can immediately start the retransmission mechanism, or generate new remedial instructions for the devices that have not been successfully executed. If the overall control effect deviates from the expectation, the master station can use this deviation as feedback, input it into the optimization algorithm of the next round, dynamically adjust the control strategy, and realize adaptive closed loop optimization.

[0177] The memory can include computer system readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory. The device can further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage media can be used for reading and writing non-removable, non-volatile magnetic media, usually called "hard drives." Program / utility programs, having a set (at least one) of program modules, can be stored in, for example, memory (RAM, ROM, etc.), such as those having a set (at least one) of program modules, including an operating system, one or more application programs, other program modules, and program data, each or a combination thereof, can include implementation of a network environment. Program modules typically perform the functions and / or methodologies of embodiments of the present invention described herein.

[0178] The processor performs various functional applications and data processing by running programs stored in the memory, such as implementing the method provided in embodiment 1 of the present invention.

Claims

1. A method for flexible control of electric heating power in a transformer substation, characterized in that, include: (1) Collect historical load data of transformer substations and perform feature engineering processing. Construct a sample set based on the historical load data of transformer substations after feature engineering processing and auxiliary information. The auxiliary information includes weather and holidays; using a sample set Train the XGBoost model; (2) Obtain the predicted load value of the transformer area using the trained XGBoost model. When the predicted load value of the transformer area Greater than the area load threshold If the condition is met, proceed to step (3); otherwise, do not perform flexible control and maintain the current operating state of the electric heating equipment. (3) Determine the optimal flexible control scheme based on graph network and improved ant colony algorithm: Model all electric heating equipment to be controlled in the entire transformer area as a graph network, where each electric heating equipment is abstracted as a node in the graph, and the edges between nodes represent the decision paths between electric heating equipment; Each ant in the colony has the task of constructing a path that traverses all nodes, which represents a specific device control sequence scheme. When visiting each node, the ants make decisions based on pheromones and heuristic information. After all the ants have completed a round of scheme construction, the scheme is evaluated, the best one is selected, and the pheromones are updated according to the adaptive flexible control fitness function. The adaptive flexible control fitness function is divided into two parts: the power grid objective function and the user comfort objective function; the pheromone update adopts the adaptive elite factor pheromone update mechanism; when the maximum number of iterations is reached, the optimal flexible control scheme is output. (4) Execute the optimal flexible control scheme and monitor the control effect. If the control is successful, return to step (2) and carry out the next round of load prediction and control process.

2. The method for flexible control of electric heating power in a transformer substation according to claim 1, characterized in that, Step (1) includes: For the time feature, given a timestamp Extract the following features: 1) Time series The granularity of recording can be set as needed; 2) 1 to 7 represent Monday to Sunday respectively; 3) Where 1 represents a holiday and 0 represents a workday; Based on the historical load characteristics of the transformer area, a lag characteristic is constructed, using the target time. The load values ​​at several previous time points are used as characteristics: (1) in, Indicates time The lag characteristic; Indicates at time Before Load value per time unit , The maximum lag order; For meteorological characteristics, obtain Temperature of Time relative humidity and wind speed ; For sliding window statistical features, construct the moving average feature and the moving standard deviation feature: 1) Moving average :past The average load over a time window reflects the recent average level, and its calculation method is as follows: (2) 2) Sliding standard deviation :past The standard deviation of load within a time window reflects the recent load volatility, and it is calculated as follows: (3) Finally, all the above features are pieced together to form a time frame. eigenvectors This eigenvector is related to the target value, i.e., the load at a future time. Together they constitute a complete sample , This indicates the time step for prediction; all samples are arranged in chronological order to form the final sample set. ,in Indicates the first A sample at each moment, This represents the total number of samples.

3. The method for flexible control of electric heating power in a transformer substation according to claim 2, characterized in that, Step (1) also includes: the predicted output of the XGBoost model is generated by... The prediction results of each regression tree collectively determine the outcome for a given sample. Its predicted value for: (4) in, It is the first A decision tree; It is the function space consisting of all possible decision trees; The objective function of the XGBoost model By loss function and regularization term It consists of two parts: (5) in, These are the parameters of the XGBoost model; the loss function. Use mean squared error; XGBoost performs regularization on each tree: (6) in, and It's a hyperparameter; This is the number of leaf nodes in the tree; It is the predicted value vector at the leaf node.

4. The method for flexible control of electric heating power in a transformer substation according to claim 3, characterized in that, Step (2) includes: for a given feature vector containing a new feature vector The input is taken, its features are fed into the pre-trained XGBoost model, and the prediction results of all decision trees are summed to obtain the predicted load value of the transformer area. .

5. The method for flexible control of electric heating power in a transformer substation according to claim 1, characterized in that, In step (3), electric heating equipment is divided into three-phase electric heating equipment, variable frequency electric heating equipment and single-phase electric heating equipment according to different control methods. For three-phase and variable frequency electric heating equipment, its power is reduced. For single-phase electric heating equipment, intermittent heating can be achieved by switching off power to different circuits at a certain duty cycle for a certain duration, thereby reducing the average power consumption.

6. The method for flexible control of electric heating power in a transformer substation according to claim 1, characterized in that, Step (3) includes: (301) Algorithm Initialization Assign a uniform pheromone concentration to all edges. Set the maximum number of ants Maximum number of iterations pheromone volatile factors And the importance weight of pheromones The importance weight of heuristic information ; (302) Iterative process of ant construction device control sequence scheme In each iteration middle, Only one ant was released; each ant Starting from the initial node, the ant sequentially selects the next unvisited node based on the state transition probability and decides the control state of the corresponding electric heating equipment; when the ant is located at the node... The next decision-making node is When, it selects a node And assign a control state probability for: (8) in, It is the edge In iteration The concentration of pheromones at that time; It's an ant. The set of currently unvisited nodes; This indicates the node currently selected from the set of unvisited nodes; It is heuristic information, and the calculation method is as follows: (9) in, , and These are the weight coefficients of each heuristic factor. ; and These are nodes The corresponding set temperature of the electric heating equipment and the actual indoor temperature; This represents the maximum difference between the set temperature of all electric heating devices and the actual indoor temperature. It is a node The real-time active power of the corresponding electric heating equipment; This is the maximum power of the electric heating equipment within the transformer area; It is based on the priority weight of device type or user contract; Each ant After completing a full graph traversal, a complete equipment control sequence scheme is generated. This scheme consists of the power and on / off status of all electric heating equipment, i.e. Then, determine whether all ants have completed the construction of the device control sequence scheme. If yes, proceed to step (303); otherwise, select the next ant. Construct equipment control sequence schemes; (303) Scheme evaluation, selection of the best option and pheromone update After all ants have completed one round of scheme construction, the fitness function is adjusted according to the adaptive flexibility. For each scheme Conduct an assessment; (10) in, and Let be the weighting coefficient, satisfying ; Power grid objective function for: (11) in, These are the weighting coefficients; In the plan The peak load of the transformer substation during the control period; This represents the peak value of the load in the transformer area predicted by the XGBoost model; User comfort objective function for: (12) in, This indicates users who have the ability to flexibly control their electric heating systems. For equipment Outdoor temperature in the area; After the evaluation is completed, a global pheromone update is performed; first, pheromones on all paths evaporate proportionally: (13) Subsequently, based on the merits and demerits of each scheme in this iteration, pheromone enhancement is applied to the paths traversed by the superior schemes; let... and Let represent the fitness values ​​of the best and worst solutions in a single iteration, respectively. For any solution used for elite enhancement... Its adaptive elite factor for: (14) in, It is a basic weight constant; To prevent the denominator from being 0; Use the optimal solution from this iteration. Enhance pheromones: (15) in, It is a constant pheromone intensity coefficient; The system then checks if the maximum number of iterations has been reached. If it has, the globally optimal control scheme is output and executed. If not, the next iteration begins. Through multiple iterations, the search process gradually converges to a flexible electric heating control scheme that balances the total load target of the distribution area with the personalized needs of users. .

7. The method for flexible control of electric heating power in a transformer substation according to claim 1, characterized in that, In step (4), a priority mechanism is introduced during the control process. The initial priority of electric heating equipment is divided into several levels. The priority queue of ordinary user equipment is sorted in descending order. When the load of the transformer area may be overloaded and needs to be adjusted, the equipment with the lower priority is selected first for flexible control of electric heating load. Then, the priority of the equipment being adjusted in the current round is set to the highest and added to the end of the priority queue. The priority of the equipment not being adjusted is reduced by one level. If its priority is already the lowest level, it will not change. Special user equipment does not use the above priority rules and always maintains the highest priority. Only when the load reduction requirement is still not met after adjusting all ordinary user equipment will the flexible control of electric heating be performed on special users.

8. The method for flexible control of electric heating power in a transformer substation according to claim 1, characterized in that, In step (4), monitoring the control effect includes: collecting the actual status of each electric heating device and forming a report on the execution results of the current round of instructions; comparing the expected state of the instructions with the actual state, and if the success rate is lower than the threshold, then starting the instruction retransmission mechanism or generating new remedial instructions for the electric heating devices that have not been successfully executed; if the overall control effect deviates from the expectation, then this deviation is used as feedback and input into the ant colony control algorithm of the next round to dynamically adjust the control strategy and achieve adaptive closed-loop optimization.

9. A flexible control system for the power of electric heating in a transformer substation, characterized in that, include: The data acquisition and model training module is used to collect historical load data of transformer substations and perform feature engineering processing. Based on the feature-engineered historical load data of transformer substations and auxiliary information, a sample set is constructed. The auxiliary information includes weather and holidays; using a sample set Train the XGBoost model; The load forecasting and judgment module is used to obtain the load forecast values ​​for the transformer area using a trained XGBoost model. When the predicted load value of the transformer area Greater than the area load threshold At that time, the optimal flexible control scheme generation module is executed; Otherwise, flexible control will not be implemented, and the current operating status of the electric heating equipment will remain unchanged; The optimal flexible control scheme generation module is used to determine the optimal flexible control scheme based on graph networks and improved ant colony algorithm: the set of all electric heating devices to be controlled in the entire transformer area is modeled as a graph network, where each electric heating device is abstracted as a node in the graph, and the edges between nodes represent the decision paths between electric heating devices; Each ant in the colony has the task of constructing a path that traverses all nodes, which represents a specific device control sequence scheme. When visiting each node, the ants make decisions based on pheromones and heuristic information. After all the ants have completed a round of scheme construction, the scheme is evaluated, the best one is selected, and the pheromones are updated according to the adaptive flexible control fitness function. The adaptive flexible control fitness function is divided into two parts: the power grid objective function and the user comfort objective function. The pheromone update adopts an adaptive elite factor pheromone update mechanism; when the maximum number of iterations is reached, the optimal flexible control scheme is output. The control execution and monitoring module is used to execute the optimal flexible control scheme generated by the optimal flexible control scheme generation module and monitor the control effect; if the regulation is successful, it returns to the load forecasting and judgment module to carry out the next round of load forecasting and regulation process.

10. A flexible power control device for electric heating in a transformer substation, characterized in that, The system includes a distribution area master station, a concentrator, several flexible load control terminal hosts, and several flexible load control terminal slaves corresponding to each flexible load control terminal host. The distribution area master station has a processor and a memory. The memory stores computer instructions, and the processor executes the computer instructions stored in the memory. When the computer instructions are executed by the processor, the distribution area electric heating power flexible control device implements the steps of the distribution area electric heating power flexible control method as described in any one of claims 1 to 8, and obtains the optimal flexible control scheme. The concentrator is responsible for communicating with the distribution area master station and aggregating data from all its downstream terminal hosts. The flexible load control terminal hosts are responsible for managing and controlling multiple flexible load control terminal slaves. The flexible load control terminal slaves are installed on the user side, directly interface with the electric heating equipment, and perform the final control operation.

Images